realizable plan vs. objective
bpeters
At the OBI workshop in January this year, it was thought that both
plans and objectives are needed in the description of experiments.
Where and how to place them within the BFO hierarchy has been a
continuous subject of debate. Briefly, and skipping the ''generically
dependent' --> concretized step, we have agreement that
Objective" 'is a information entity that specifies a desired endpoint
in a process'.
Example: 'Determine sugar level in blood sample'.
Plan: 'is a realizable information entity that specifies a series of
steps to be carried out to accomplish an objective'.
Example: 'ship sample to lab X, pay them $ to measure blood sugar
level.
Example: 'use instrument Y to measure blood sugar level.
The question to this group: should objectives themselves be considered
realizable or not?
At the OBI workshop, I argued strongly that objectives are not
"realizable'. I just recently comprehended that what I was concerned
about is that objectives are not 'executable'. That is, without having
a plan, the process to achieve an objective is not well defined. But
is this that required to make objectives executable? It now seems to
me that 'roles' are not executed in that sense either.
It seems to me that there are processes like 'planning' that have a
clear objectives but can never be fully planned in advance. For
these, I would like to put forward the notion of 'objective driven
processes' which are processes initiated by one or more actors in
order to achieve an objective. 'planned processes' would be a subclass
of these. I am still unclear if an 'objective driven process' is a
realization of the objective, and more general what 'realization of'
entails.
Any help would be very appreciated!
Bjoern
Cristian Cocos
> in a process'.
> Example: 'Determine sugar level in blood sample'.
I made several postings a long while ago on the OBI DENRIE list trying to
correct what appears to me to be a fundamental misunderstanding, though I
have yet to hear back from any of the developers. Here's hoping for a flame
or two :-)
A confusion that runs rampant on the DENRIE branch is the confusion between
a thing on the one hand, and a proposition ("information") *about* that
thing on the other. Objectives *themselves* are in no way informational
entities; they are either states of affairs (the closest BFO comes to that
is "process boundary" (span:ProcesBoundary)), or *roles of* states of
affairs (... yes, I know, "role" only applies to independent continuants).
The *proposition* that states the objective/state of affairs in question
could, at best, be regarded as an informational entity.
CC
Barry Smith
>First time posting, long time lurking.
>
>At the OBI workshop in January this year, it was thought that both
>plans and objectives are needed in the description of experiments.
>Where and how to place them within the BFO hierarchy has been a
>continuous subject of debate. Briefly, and skipping the ''generically
>dependent' --> concretized step, we have agreement that
>
>Objective" 'is a information entity that specifies a desired endpoint
>in a process'.
>Example: 'Determine sugar level in blood sample'.
Not sure that the objective is not itself the endpoint, and the
specification of the objective the information entity.
>Plan: 'is a realizable information entity that specifies a series of
>steps to be carried out to accomplish an objective'.
>Example: 'ship sample to lab X, pay them $ to measure blood sugar
>level.
>Example: 'use instrument Y to measure blood sugar level.
>
>The question to this group: should objectives themselves be considered
>realizable or not?
>
>At the OBI workshop, I argued strongly that objectives are not
>"realizable'. I just recently comprehended that what I was concerned
>about is that objectives are not 'executable'. That is, without having
>a plan, the process to achieve an objective is not well defined. But
>is this that required to make objectives executable? It now seems to
>me that 'roles' are not executed in that sense either.
>
>It seems to me that there are processes like 'planning' that have a
>clear objectives but can never be fully planned in advance. For
>these, I would like to put forward the notion of 'objective driven
>processes' which are processes initiated by one or more actors in
>order to achieve an objective. 'planned processes' would be a subclass
>of these. I am still unclear if an 'objective driven process' is a
>realization of the objective, and more general what 'realization of'
>entails.
I do not object to the idea of distingishing 'objective-driven
processes' with 'planned objective-driven processes' as a subtype.
(Could there be planned processes that are not objective driven?))
Not all objective-driven processes involve realizing their
objectives-- only completed objective-driven processes do so.
However, specifications of objective-driven processes are, I think,
realizable (and this independently of whether they are planned).
Realizables, as Bjorn notes, do not always involve realization of plans.
BS
Barry Smith
>First time posting, long time lurking.
>
>At the OBI workshop in January this year, it was thought that both
>plans and objectives are needed in the description of experiments.
>Where and how to place them within the BFO hierarchy has been a
>continuous subject of debate. Briefly, and skipping the ''generically
>dependent' --> concretized step, we have agreement that
>
>Objective" 'is a information entity that specifies a desired endpoint
>in a process'.
>Example: 'Determine sugar level in blood sample'.
Not sure that the objective is not itself the endpoint, and the
specification of the objective the information entity.
>Plan: 'is a realizable information entity that specifies a series of
>steps to be carried out to accomplish an objective'.
>Example: 'ship sample to lab X, pay them $ to measure blood sugar
>level.
>Example: 'use instrument Y to measure blood sugar level.
>
>The question to this group: should objectives themselves be considered
>realizable or not?
>
>At the OBI workshop, I argued strongly that objectives are not
>"realizable'. I just recently comprehended that what I was concerned
>about is that objectives are not 'executable'. That is, without having
>a plan, the process to achieve an objective is not well defined. But
>is this that required to make objectives executable? It now seems to
>me that 'roles' are not executed in that sense either.
>
>It seems to me that there are processes like 'planning' that have a
>clear objectives but can never be fully planned in advance. For
>these, I would like to put forward the notion of 'objective driven
>processes' which are processes initiated by one or more actors in
>order to achieve an objective. 'planned processes' would be a subclass
>of these. I am still unclear if an 'objective driven process' is a
>realization of the objective, and more general what 'realization of'
>entails.
I do not object to the idea of distingishing 'objective-driven
processes' with 'planned objective-driven processes' as a subtype.
(Could there be planned processes that are not objective driven?))
Not all objective-driven processes involve realizing their
objectives-- only completed objective-driven processes do so.
However, specifications of objective-driven processes are, I think,
realizable (and this independently of whether they are planned).
Realizables, as Bjorn notes, do not always involve realization of plans.
BS
>Any help would be very appreciated!
>
>Bjoern
>
>
>
>
bpeters
I take back from this that I need to take care to talk about
'objective specifications' as information entities, not 'objectives'.
Christian: I am no taker for a DENRIE branch flame, as I am writing
this as representative of the 'plans and planned processes branch'.
That points out one of our current problems: the OBI development
structure is relying on having a clean separation between the
branches. This poses problems when terms like 'objective' are
discussed that fall between branches; they can get neglected for quite
a while.
Barry: I agree with everything else you wrote. You posed one side
question: "Could there be planned processes that are not objective
driven?". Currently we are stating that every plan has an objective,
which answers this as no. However, clearly one can think of a process
specification that can be realized without an apparent objective, such
as 'shape a chicken beak with your hands. Open and close it four
times'. For OBI, where planned processes are encountered as
experimental designs, we have not encountered the need for a process
specification without an objective. For BFO, having that option
probably makes sense, but I think it would not need to be considered a
plan.
I am now integrating this into the OWL representation of OBI, to see
the impact on other terms, and will gather feedback from the OBI
developers. There will likely be more questions to you arising in this
process, and we will definitely want to get approval from you once we
believe to have a workable model.
Bjoern
Cristian Cocos
What I meant was that I was hoping to at least be flamed *myself*--as
opposed to ignored. Certainly in no way would I dare to flame anyone.
> as I am writing
> this as representative of the 'plans and planned processes branch'.
As a matter of fact, my critique regarding the way "objective" was being
construed did make its way onto the obi-plan-branch mailing list last year
around mid-October, in reply to Chris Stoeckert's "Do objectives belong to
the Plan branch?" question.
CC
Barry Smith
I think your point (about the need to distinguish between objective
and representation of objective) is a good one -- I made the same
point myself a couple of days ago, before I read your message.
The question now is: what is an objective.
Terminal boundary of a process is, I think, a good start, and allows
us to draw on a lot of good work by the verb aspect people (Vendler,
Galton ...) on the distinction between activities (walking, climbing)
and achievements (completing a race, reaching the top).
But of course not every terminal boundary of a process is an
objective; so we still need to say more
BS
Phillip Lord
Surely an objective is where you want the terminal boundary to a process to
be, rather that the actual termination of it.
It's a bit like the distinction that is made between a function and a role;
the function might be a intrinsic part of the entity, while an role is not
intrinsic to the entity.
So an objective is what is socially considered to be an desired ending to
a process; so, it is an objective for Eurovision to vote for a song, but it's
not an objective for Eurovision to vote based on the historical politics of
the countries involved.
Which means that Eurovision 09 has an objective, even thought it hasn't begun
yet.
Phil
>>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
Barry> Christian, I think your point (about the need to distinguish between
Barry> objective and representation of objective) is a good one -- I made
Barry> the same point myself a couple of days ago, before I read your
Barry> message. The question now is: what is an objective. Terminal boundary
Barry> of a process is, I think, a good start, and allows us to draw on a
Barry> lot of good work by the verb aspect people (Vendler, Galton ...) on
Barry> the distinction between activities (walking, climbing) and
Barry> achievements (completing a race, reaching the top). But of course not
Barry> every terminal boundary of a process is an objective; so we still
Barry> need to say more BS
Barry>
--
Phillip Lord, Phone: +44 (0) 191 222 7827
Lecturer in Bioinformatics, Email: philli...@newcastle.ac.uk
School of Computing Science, http://homepages.cs.ncl.ac.uk/phillip.lord
Claremont Tower Room 909, skype: russet_apples
Newcastle University, msn: m...@russet.org.uk
NE1 7RU
Cristian Cocos
> and representation of objective) is a good one
Conflating things with information/propositions *about* things is, as a
matter of fact, a mistake that I came across all too often in my dealings
with the OBI, especially on the DENRIE branch. It is in no way specific to
"objective." My repeated attempts to point this out however, have fallen on
deaf ears.
> The question now is: what is an objective.
> Terminal boundary of a process is, I think, a good start
It does seem to me to be a good start, although so far
"snap:ProcessBoundary" has been presented to me as more akin to "event" than
to "state," whereas I am tempted to regard "objective" as more akin to the
latter. Truth be told, "event" and "state" are not really that alien from
each other, as it is customary to assimilate an event to a pair (state,
moment in time): event = (state, t).
CC
Matthew Pocock
> > I think your point (about the need to distinguish between objective
> > and representation of objective) is a good one
>
> matter of fact, a mistake that I came across all too often in my dealings
> with the OBI, especially on the DENRIE branch. It is in no way specific to
> "objective." My repeated attempts to point this out however, have fallen on
> deaf ears.
I have some sympathy with you on this one.
> > The question now is: what is an objective.
> > Terminal boundary of a process is, I think, a good start
>
> It does seem to me to be a good start, although so far
> "snap:ProcessBoundary" has been presented to me as more akin to "event"
> than to "state," whereas I am tempted to regard "objective" as more akin to
> the latter. Truth be told, "event" and "state" are not really that alien
> from each other, as it is customary to assimilate an event to a pair
> (state, moment in time): event = (state, t).
Objectives seem to be statements about matcing vs non-matching states of the
world. The objective is met if the world is in a particular state range
during a particular time interval.
Objective: to weigh my Stryer 3rd edition
Plan: a specification (rarefication) of a process
put the item to be weighed on scales ; read off the weight ; report the
reading.
Intent:
* I will weigh my Stryer 3rd edition.
* I will follow the weighing plan.
* I will weigh my Stryer 3rd edition by following the weighing plan.
Process: my Stryer 3rd edition was placed on scales; I read off the weight ;
It was 10.2kg.
The process followed the plan.
The process met the objective.
The process achieves the intent.
It is easy to get these three things mixed up when talking becasue each of
these are often stated in terms of the others. Plans are often stated in
terms of sub-objectives (walk to the shop ; buy shopping ; return home) and
objectives can have both static and dynamic requirements (know the weight of
my Stryer vs have exercised for half an hour). Intent can be to meet an
objective or to follow a plan, or to meet an objective by following a plan
and probably other things as well.
Matthew
> CC
Alan Ruttenberg
On May 28, 2008, at 8:57 AM, Matthew Pocock wrote:
> Objectives seem to be statements about matcing vs non-matching
> states of the
> world. The objective is met if the world is in a particular state
> range
> during a particular time interval.
>
> Objective: to weigh my Stryer 3rd edition
>
> Plan: a specification (rarefication) of a process
> put the item to be weighed on scales ; read off the weight ; report
> the
> reading.
Matthew, do you distinguish what you have written as the steps of the
plan from objectives? It seems to me each could be written as an
objective, with a more specific plan for it. The delineation here
would seem to be conscious versus unconscious. Certainly "report the
reading", for instance, involves a series of steps, for instance
getting some paper, getting a pen, writing it down. And "getting a
pen": Looking around the room. Walking towards the pen, picking up the
pen, walking back to the paper. And "walking towards the pen" taking a
step. Avoiding falling over the laundry basket, taking the next step,
etc.
The model of goals leading to plans with subgoals, recursively,
matches well one of the foundations of artificial intelligence http://en.wikipedia.org/wiki/General_Problem_Solver
> Intent:
> * I will weigh my Stryer 3rd edition.
> * I will follow the weighing plan.
> * I will weigh my Stryer 3rd edition by following the weighing plan.
Do the intents not include the plan steps? Does one not intend to
"report the reading".
> Process: my Stryer 3rd edition was placed on scales; I read off the
> weight ;
> It was 10.2kg.
"It was 10.2 kg" doesn't sound like a process.
> The process followed the plan.
> The process met the objective.
> The process achieves the intent.
Christian's characterization of objectives as process boundaries is
the first one I've heard that puts objectives outside the head, so I
like that. Concur re: Specification of objectives versus objectives.
Bjoern, you coming to the conclusion that objectives are realizable is
a relief as I can put a piece of thinking that was held upside down in
consideration of your thinking right side up and back in place.
-Alan
Cristian Cocos
> On Behalf Of Phillip Lord
> Sent: Tuesday, May 27, 2008 19:21
> Surely an objective is where you want the terminal boundary to a
> process to be, rather that the actual termination of it.
My very first attempt to demystify objectives proclaimed them as *roles* of
states of affairs. That, however, didn't bode well with the view that roles
only apply to independent continuants (IC)--though, of course, one can
always attempt to extend the range of the "hasRole" relation to include
other stuff besides ICs.
Otherwise, as I see it, there are two major obstacles to construing
objectives as a subclass of process boundaries simpliciter:
1. as I was pointing out earlier, span:ProcessBoundary is *not quite* an
exact synonym for "state of affairs," but closer to "event"--which, though
related to a state, is qualitatively something else;
2. as you very well point out, the term "objective" connotes potentiality to
a much larger extent than other BFO universals (like "Process," say)--to
such an extent that it just cannot be ignored: it is not only *actual*
states of affairs that are objectives, but *potential* (or hypothetical)
states of affairs too.
Given these difficulties, my suggestion would be to just *not* reflect
"objective" anywhere in the taxonomic tree, but capture it solely via the
objectiveOf/hasObjective relation, along the lines of:
StateOfAffairs objectiveOf Action
Action hasObjective StateOfAffairs
CC
Phillip Lord
CC> Otherwise, as I see it, there are two major obstacles to construing
CC> objectives as a subclass of process boundaries simpliciter:
CC> 1. as I was pointing out earlier, span:ProcessBoundary is *not quite* an
CC> exact synonym for "state of affairs," but closer to "event"--which,
CC> though related to a state, is qualitatively something else;
I wasn't sure that I got that, but I'll take your word for it.
CC> 2. as you very well point out, the term "objective" connotes
CC> potentiality to
CC> a much larger extent than other BFO universals (like "Process," say)--to
CC> such an extent that it just cannot be ignored: it is not only *actual*
CC> states of affairs that are objectives, but *potential* (or hypothetical)
CC> states of affairs too.
Well, both role and function are potentials as well. Most sperm don't
fertilise anything. Doctors don't necessarily see any patients (during the
golf season).
CC> Given these difficulties, my suggestion would be to just *not* reflect
CC> "objective" anywhere in the taxonomic tree, but capture it solely via
CC> the objectiveOf/hasObjective relation, along the lines of:
CC> StateOfAffairs objectiveOf Action
CC> Action hasObjective StateOfAffairs
A StateOfAffairs not being an objective? I realise that there is not clean
separation between a StateOfAffairs that is an Objective and one which is not.
The solution is to invent a word like "Fiat" and apply it to occurrents.
Phil
Daniel Schober
sorry, just saw this discussion now... I was posting this to OBI dev earlier:
As a concrete example we can look at a
low
level objective from an OBI use case ( Yes, it might be difficult to
decide at what level we
need to model objectives -only for top level abstractions or for each
process steps...):
"Prepare [state of affairs:]
Percoll
solutions of the following densities (1.064, 1.075, 1.081, and 1.087 g/ml) [Action/Process:]
by diluting Percoll medium with 10× PBS and distilled water according
to the manufacturer’s one-step procedure."
So each small protocoll part can have its
own objective. Here it seems the Objective/State of affairs is
to have 4
Percoll solutions of defined density.
The Action/Process to archieve this objective is the dilution
of a medium with buffers and water according to
PA/manufacturers_one-step_procedure .
So in my knowledgebase I would represent:
Objective/State
of
affairs:
Fact 1: Percoll
solution has_density 1.064 g/ml ...
Fact 4: Percoll
solution has_density 1.087 g/ml
Action/Process: [Christian,
where
is the problem you have with Processes as 'Actions'?]
Process/manufacturers_one-step_procedure: (modeled as composite
steps)
Percoll medium dilute_with solution/10xPBS
Percoll medium dilute_with destilled water
...
I agree with CC: span:ProcessBoundary is *not quite* an exact synonym for "state of affairs,"
I am not sure how to define such boundaries... In case you mean it is an ending number of assertions or steps, then this is in conflict with the assumption that objectives are fractal and can always be represented more granular via subobjectives as can sub-processes. I dont think span:ProcessBoundary will bring us any further since to me objectives are sets of states/ facts. A state can then trigger a process of cause, but is not the process itself.
Matthews mail hints that a look at the BDI models could be valuable here: http://en.wikipedia.org/wiki/BDI_software_agent#Intentions
and from page 54 onward in http://books.google.co.uk/books?id=JYcznFCN3xcC&pg=PA585&lpg=PA585&dq=bdi+architecture&source=web&ots=IF-SoGKrUy&sig=Kg7ab5eWLbQ916TgKEzDedAZ-AQ&hl=en#PPP1,M1
There is even a book on process modeling for KBs: http://csli-publications.stanford.edu/site/1575861925.html
Christian, is your notion of 'state of affairs' the same as the one from Wittgenstein (e.g. from http://www.formalontology.it/wittgensteinl.htm )? Am I right to correlate Wittgenstein terminology to our OWL DL speak in the following way ?
state of affairs= set of assertions/triples/facts
Object =atomic class
complex item = composite/defined class (but also set of assertions?)
I would like to know because we used the term
'states of affairs' in OBI before, e.g. see my
obi-dev mail from 02/11/2007 13:46
(
http://groups.google.com/group/obi-developer/msg/e1c08ba4bcc4773c?hl=en&
)
Cheers,
Daniel Schober
-- __________________________________________________________________________________________ Dr. Daniel Schober NET Project - Ontologist The European Bioinformatics Institute email: sch...@ebi.ac.uk EMBL Outstation - Hinxton direct: +44 (0)1223 494410 Wellcome Trust Genome Campus fax: +44 (0)1223 494 468 Cambridge CB10 1SD, UK Room: A3-141 (extension building) Project page: www.ebi.ac.uk/net-project Personal page: http://www.ebi.ac.uk/Information/Staff/person_maint.php?s_person_id=734 Former home page: http://www.bioinf.mdc-berlin.de/%7Eschober/
bpeters
proposal I take that we need a StateOfAffairs, Action, and a relation
objectiveOf, none of which currently exist in BFO. This is motivated
by not wanting to call objectives information, but not really wanting
to call them states either and putting them in a relation. I am afraid
if plans and their relation to objectives is dealt with in a similar
way, this will lead to major BFO revisions. Given the fact that we
need to use something for development of OBI now (which can always be
revisited later), how horrible is it to stick to
'objective specification' is a realizable information entity that
specifies a desired endpoint (process boundary) of a process'
'plan specification': 'is a realizable information entity that
specifies a series of
steps (sub-objectives) to be carried out to accomplish an objective'.
These can be concretized and realized by an actor. We would work
within what we have and understand.
To support this, I would argue that there *are* specifications, and
that they are information entities. I understand that it is important
to not confuse these with the entities they specify, but that won't
make them go away. Their relationships should instead be made
explicit: A specification applies_to a kind of entity each of which
can comform_with the specification or not. An objective specification
applies_to process boundaries. Some process_boundaries conform_with
the objective specification, others don't.
Cristian Cocos
From: bfo-d...@googlegroups.com
[mailto:bfo-d...@googlegroups.com] On Behalf Of Daniel Schober
Sent: Thursday, May 29, 2008 17:00
To: bfo-d...@googlegroups.com
Cc: OBI Developers
Subject: [bfo-discuss] Re: realizable plan vs. objective
Action/Process: [Christian, where is the problem you have with Processes as 'Actions'?]
“Action” is, so far, something that I am having trouble construing in BFO terms. “Action” connotes intentionality and agency to an extent that simple “process” does not. “Process,” on the other hand, is something objective: in order to specify a process one doesn’t need to point out a (rational) agent, whereas for an action one does. I find very odd to speak of a process that has an objective without essentially tying it to a rational agent. What is the objective of the Earth’s rotation? The objective of the Big Bang [or is that regarded as an event, not a process]? Of course, one way to go about it is to say that not all processes have objectives. The trouble is that sometimes one and the same process can be regarded at the same time as both someone’s actions, or as a process simpliciter. In the first case it would have an objective, in the other--none. Processes with multiple agents can have multiple diverging (even contradictory) objectives. Perhaps one way to go about it would be to construe an action as a pair (process, agent)—actions would then appear as results of perspectives/ways of regarding processes. Either way, this is not something BFO could easily accommodate in its present guise.
As a result of this, just to get things going and avoid all this ontological nightmare, I ultimately find it not that horrible (to take Bjoern’s word) if we stuck to “Process” instead of “Action” for the time being—hence speak of “Process hasObjective” etc.
As Christian, I see 'states of affairs' as a set of assertions/facts within the
knowledgebase. So, the range and domain of the assertion Christian proposes (
Action hasObjective StateOfAffairs ) is here broken down into many
assertions. Is that right?
My main beef was with the difference between proposition (“assertion”) on one hand, and whatever the proposition stands for (“represents”) on the other. Colloquial language is sublimely sloppy in this respect, as most of the times it conflates the two. I think we, for OBI purposes, should not. As such, I have to correct your above statement: states of affairs (or just simply “states”) are whatever the set of assertions within the knowledgebase refer to—in case, of course, they are declarative, factual assertions.
A state can then trigger a process of cause …
I’d rather go with “event can trigger a (causal) process” rather than state.
Christian, is your notion of 'state of affairs' the same as the one from
Wittgenstein (e.g. from http://www.formalontology.it/wittgensteinl.htm
)?
My notion of a state of affairs comes from Physics, and is essentially that of a state of a physical system—see, e.g., (Phenomenological) Thermodynamics. This has been abstracted and generalized by Systems Theory. As for Wittgenstein’s stance, as a non-Wittgenstein scholar I’d have to lend credence to what others say about the way he speaks about states of affairs (see, e.g., here, par. 2.1, where they discuss his picture theory of propositions), and I’d have to sort of agree that his and my notions may be the same.
Am
I right to correlate Wittgenstein terminology to our OWL DL speak in the
following way ?
state of affairs= set of assertions/triples/facts
Object =atomic class
complex item = composite/defined class (but also set of assertions?)
If by “=” you mean “represents,” I’d sort of agree with the spirit of these “equalities,” although I may need further explanations. I’d also have to suggest to move “fact” onto the other side of the equality though, as for Wittgenstein the set of facts is a subset of the set of states of affairs. Otherwise, strictly identifying these would be a mistake—though quite harmless in day to day computer science life, save for dealing with explicit references to informational entities (as on the DENRIE branch), where making a distinction b/w the two categories becomes crucial.
C
Barry Smith
>From: bfo-d...@googlegroups.com
>[mailto:bfo-d...@googlegroups.com] On Behalf Of Daniel Schober
>Sent: Thursday, May 29, 2008 17:00
>To: bfo-d...@googlegroups.com
>Cc: OBI Developers
>Subject: [bfo-discuss] Re: realizable plan vs. objective
>Action/Process: [Christian, where is the problem
>you have with Processes as 'Actions'?]
>"Action" is, so far, something that I am having
>trouble construing in BFO terms. "Action"
>connotes intentionality and agency to an extent
>that simple "process" does not. "Process," on
>the other hand, is something objective: in order
>to specify a process one doesn't need to point
>out a (rational) agent, whereas for an action one does.
an action is one type of process (one which
stands in the agent relation e.g. to some human being)
each action is a process
the fact that we can specify an action by
describing it in terms of intentionality is interesting
but not different, in principle, from the fact
that we can describe a process of oxidation in terms of oxygen
everything is objective, here
there are, it is true, subjective view of actions
e.g. I find my current action of typing this email satisfying
but then I might find also a process of oxidation satisfying
subjective views are additional entities, and to
this extent they, too, are objective
BS
>I find very odd to speak of a process that has
>an objective without essentially tying it to a
>rational agent. What is the objective of the
>Earth's rotation? The objective of the Big Bang
>[or is that regarded as an event, not a
>process]? Of course, one way to go about it is
>to say that not all processes have objectives.
>The trouble is that sometimes one and the same
>process can be regarded at the same time as both
>someone's actions, or as a process simpliciter.
>In the first case it would have an objective, in
>the other--none. Processes with multiple agents
>can have multiple diverging (even contradictory)
>objectives. Perhaps one way to go about it would
>be to construe an action as a pair (process,
>agent)-actions would then appear as results of
>perspectives/ways of regarding processes. Either
>way, this is not something BFO could easily accommodate in its present guise.
>As a result of this, just to get things going
>and avoid all this ontological nightmare, I
>ultimately find it not that horrible (to take
>Bjoern's word) if we stuck to "Process" instead
>of "Action" for the time being-hence speak of "Process hasObjective" etc.
>
>As Christian, I see 'states of affairs' as a set
>of assertions/facts within the knowledgebase.
>So, the range and domain of the assertion
>Christian proposes ( Action hasObjective
>StateOfAffairs ) is here broken down into many assertions. Is that right?
>
>My main beef was with the difference between
>proposition ("assertion") on one hand, and
>whatever the proposition stands for
>("represents") on the other. Colloquial language
>is sublimely sloppy in this respect, as most of
>the times it conflates the two. I think we, for
>OBI purposes, should not. As such, I have to
>correct your above statement: states of affairs
>(or just simply "states") are whatever the set
>of assertions within the knowledgebase refer
>to-in case, of course, they are declarative, factual assertions.
>
>A state can then trigger a process of cause ...
>I'd rather go with "event can trigger a (causal) process" rather than state.
>
>Christian, is your notion of 'state of affairs'
>the same as the one from Wittgenstein (e.g.
>from
><http://www.formalontology.it/wittgensteinl.htm>http://www.formalontology.it/wittgensteinl.htm
>)?
>My notion of a state of affairs comes from
>Physics, and is essentially that of a state of a
>physical system-see, e.g., (Phenomenological)
>Thermodynamics. This has been abstracted and
>generalized by Systems Theory. As for
>Wittgenstein's stance, as a non-Wittgenstein
>scholar I'd have to lend credence to what others
>say about the way he speaks about states of
>affairs (see, e.g.,
><http://plato.stanford.edu/entries/wittgenstein>here,
>par. 2.1, where they discuss his picture theory
>of propositions), and I'd have to sort of agree
>that his and my notions may be the same.
>Am I right to correlate Wittgenstein terminology
>to our OWL DL speak in the following way ?
>state of affairs= set of assertions/triples/facts
>Object =atomic class
>complex item = composite/defined class (but also set of assertions?)
>If by "=" you mean "represents," I'd sort of
>agree with the spirit of these "equalities,"
>although I may need further explanations. I'd
>also have to suggest to move "fact" onto the
>other side of the equality though, as for
>Wittgenstein the set of facts is a subset of the
>set of states of affairs. Otherwise, strictly
>identifying these would be a mistake-though
Barry Smith
>Given these difficulties, my suggestion would be to just *not* reflect
>"objective" anywhere in the taxonomic tree, but capture it solely via the
>objectiveOf/hasObjective relation, along the lines of:
>
>StateOfAffairs objectiveOf Action
>
>Action hasObjective StateOfAffairs
>
>CC
I agree with the spirit of Christian's final proposal here.
However, we will be able to allow something approximating to the
second of these only -- not every state of affairs, not even every
state of affairs of a specific type T will be such that all its
instances exist only as the objective of some action (which is what
would be required for an all some relation objective_of)
A further problem is that we do not have a category 'state of
affairs' in BFO. I agree that a state of affairs is not in general
identifiable with the boundary of a process. But how, then, should
'state of affairs' be defined? I have failed to find a way of
answering this question that is in keeping with BFO principles -- the
reason being (roughly) that states of affairs seem too closely tied
to their specifications to be acceptable as first-order entities in
their own right.
Perhaps, if we follow Bjoern's line of reasoning, and include in OBI
a type 'specification of objective', we can avoid this thorn.
BS
Barry Smith
>Hi all,
>sorry, just saw this discussion now... I was posting this to OBI dev earlier:
>
>As a concrete example we can look at a low level objective from an
>OBI use case ( Yes, it might be difficult to decide at what level we
>need to model objectives -only for top level abstractions or for
>each process steps...):
>
>"Prepare [state of affairs:] Percoll solutions of the following
>densities (1.064, 1.075, 1.081, and 1.087 g/ml) [Action/Process:] by
>diluting Percoll medium with 10 PBS and distilled water according to
>the manufacturer's one-step procedure."
>
>So each small protocoll part can have its own objective. Here it
>seems the Objective/State of affairs is to have 4 Percoll solutions
>of defined density.
>The Action/Process to archieve this objective is the dilution of a
>medium with buffers and water according to
>PA/manufacturers_one-step_procedure .
>
>So in my knowledgebase I would represent:
>
>Objective/State of affairs:
> Fact 1: Percoll solution has_density 1.064 g/ml ...
> Fact 4: Percoll solution has_density 1.087 g/ml
The objective cannot be a fact, since of course many objectives are
never realized.
Also the objective is (presumably) to have some specific percoll
solution in some specific place have the mentioned density.
>Action/Process: [Christian, where is the problem you have with
>Processes as 'Actions'?]
> Process/manufacturers_one-step_procedure: (modeled as composite steps)
> Percoll medium dilute_with solution/10xPBS
> Percoll medium dilute_with destilled water
> ...
>As Christian, I see 'states of affairs' as a set of assertions/facts
>within the knowledgebase.
Assertions and facts are ontologically two quite different things.
>So, the range and domain of the assertion Christian proposes (
>Action hasObjective StateOfAffairs ) is here broken down into many
>assertions. Is that right?
>I agree with CC: span:ProcessBoundary is *not quite* an exact
>synonym for "state of affairs,"
>I am not sure how to define such boundaries... In case you mean it
>is an ending number of assertions or steps, then this is in conflict
>with the assumption that objectives are fractal and can always be
>represented more granular via subobjectives as can sub-processes.
This would make the specification of every objective infinitely
complex. I think you need to take granularity more seriously.
Granularity means that for each specific purpose we overlook, in
specifying our objectives, and in specifying our processes, the
levels of finer detail not relevant to that purpose.
> I dont think span:ProcessBoundary will bring us any further since
> to me objectives are sets of states/ facts. A state can then
> trigger a process of cause, but is not the process itself.
State = state of affairs?
And what is a 'process of cause' when it's at home?
>Matthews mail hints that a look at the BDI models could be valuable
>here: http://en.wikipedia.org/wiki/BDI_software_agent#Intentions
>and from page 54 onward in
><http://books.google.co.uk/books?id=JYcznFCN3xcC&pg=PA585&lpg=PA585&dq=bdi+architecture&source=web&ots=IF-SoGKrUy&sig=Kg7ab5eWLbQ916TgKEzDedAZ-AQ&hl=en#PPP1,M1>http://books.google.co.uk/books?id=JYcznFCN3xcC&pg=PA585&lpg=PA585&dq=bdi+architecture&source=web&ots=IF-SoGKrUy&sig=Kg7ab5eWLbQ916TgKEzDedAZ-AQ&hl=en#PPP1,M1
>There is even a book on process modeling for KBs:
><http://csli-publications.stanford.edu/site/1575861925.html>http://csli-publications.stanford.edu/site/1575861925.html
>
>Christian, is your notion of 'state of affairs' the same as the one
>from Wittgenstein (e.g.
>from
><http://www.formalontology.it/wittgensteinl.htm>http://www.formalontology.it/wittgensteinl.htm
>)? Am I right to correlate Wittgenstein terminology to our OWL DL
>speak in the following way ?
>state of affairs= set of assertions/triples/facts
W distinguishes very clearly between assertions on the one side (=
the side of human beings using language, or the side of the
meaning/logical content) and states of affairs / facts on the other
side. Facts are what make assertions true. Facts, for W, divide into
atomic (corresponding, I suppose, to triples), and molecular facts
(corresponding to logical compounds of triples). All facts obtain.
States of affairs are sometimes described as possible atomic facts,
i.e. they are atomic facts which may or may not obtain. It is not
clear that BFO has room for states of affairs since BFO deals only
with what it exists (no unicorns, leprechauns, perpetual motion
machines, fractured lips, absent nipples, etc.), and so to make room
for states of affairs in BFO would be to make room for existing
entities which do not obtain.
>Object =atomic class
>complex item = composite/defined class (but also set of assertions?)
Does not sound like W to me.
>I would like to know because we used the term 'states of affairs' in
>OBI before, e.g. see my obi-dev mail from 02/11/2007 13:46
>(
><http://groups.google.com/group/obi-developer/msg/e1c08ba4bcc4773c?hl=en&>http://groups.google.com/group/obi-developer/msg/e1c08ba4bcc4773c?hl=en&
>)
There you say:
>So in short a 'role' is a particular manifestation that occurs only
>under certain circumstances (natural, social or institutional contexts
>(Wittgensteins 'state of affairs' I believe)
This is not a correct reading of W's state of affairs.
Incidentally, my scepticism as concerns finding a place for states of
affairs in BFO, rests on a multi-year attempt to find a coherent
treatment of them:
http://ontology.buffalo.edu/smith/articles/logsvh.html
http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html
http://ontology.buffalo.edu/smith/articles/hloa.html
http://ontology.buffalo.edu/smith/articles/Questions.pdf
Barry
>Cheers,
> Daniel Schober
>
>Cristian Cocos wrote:
>>>
>Institute email: <mailto:sch...@ebi.ac.uk>sch...@ebi.ac.uk
>EMBL Outstation - Hinxton direct: +44 (0)1223 494410
>Wellcome Trust Genome Campus fax: +44 (0)1223 494 468
>Cambridge CB10 1SD, UK Room: A3-141 (extension building)
>
>Project page: <http://www.ebi.ac.uk/net-project>www.ebi.ac.uk/net-project
>
>Personal
>page:
><http://www.ebi.ac.uk/Information/Staff/person_maint.php?s_person_id=734>http://www.ebi.ac.uk/Information/Staff/person_maint.php?s_person_id=734
>Former home page:
><http://www.bioinf.mdc-berlin.de/%7Eschober/>http://www.bioinf.mdc-berlin.de/%7Eschober/
>
>
>
Barry Smith
>We are working with BFO as it currently exists in OWL. From CC's
>proposal I take that we need a StateOfAffairs, Action, and a relation
>objectiveOf, none of which currently exist in BFO. This is motivated
>by not wanting to call objectives information, but not really wanting
>to call them states either and putting them in a relation. I am afraid
>if plans and their relation to objectives is dealt with in a similar
>way, this will lead to major BFO revisions. Given the fact that we
>need to use something for development of OBI now (which can always be
>revisited later), how horrible is it to stick to
>
>'objective specification' is a realizable information entity that
>specifies a desired endpoint (process boundary) of a process'
Drop 'realizable', I think.
An objective specification is not a process or plan specification,
and even the latter are not clearly a realizables, in contrast to: plan.
>'plan specification': 'is a realizable information entity that
>specifies a series of
>steps (sub-objectives) to be carried out to accomplish an objective'.
Here, too, drop 'realizable' and drop 'sub-objective'
Try to define 'step'.
>These
rather: what they specify
>can be concretized and realized by an actor. We would work
>within what we have and understand.
>
>To support this, I would argue that there *are* specifications, and
>that they are information entities.
I agree
>I understand that it is important
>to not confuse these with the entities they specify, but that won't
>make them go away. Their relationships should instead be made
>explicit:
Exactly
Barry Smith
>Surely an objective is where you want the terminal boundary to a process to
>be, rather that the actual termination of it.
>
>It's a bit like the distinction that is made between a function and a role;
>the function might be a intrinsic part of the entity, while an role is not
>intrinsic to the entity.
>
>So an objective is what is socially considered to be an desired ending to
>a process; so, it is an objective for Eurovision to vote for a song, but it's
>not an objective for Eurovision to vote based on the historical politics of
>the countries involved.
>
>Which means that Eurovision 09 has an objective, even thought it hasn't begun
>yet.
I agree with this.
I believe that it lends more weight to Bjoern's suggestion that we
include, not objective, in BFO, but rather objective specification.
It is the latter that really exists. If and when the plan associated
with the objective specification is realized, then we might also say
that the objective is realized (shorthand for something more
complicated involving objective specification); sometimes this will
involve the terminal boundary of a process of a specific sort (e.g.
the liquid is brought to a certain temperature at the conclusion of a
heating process); sometimes not (the objective might have been, e.g.
to PREVENT the liquid ever reaching a certain temperature).
Conclusion: objectives do not form a coherent universal on the side
of reality; hence no objectives in BFO. Just objective specifications.
BS
Phillip Lord
>> g/ml ... Fact 4: Percoll solution has_density 1.087 g/ml
Barry> The objective cannot be a fact, since of course many objectives are
Barry> never realized. Also the objective is (presumably) to have some
Barry> specific percoll solution in some specific place have the mentioned
Barry> density.
Barry
There are many functions and many roles which are never realised also. If
objectives are socially defined, then they have an existence regardless of
whether they actually come to pass?
>> In case you mean it is an ending number of assertions or steps, then this
>> is in conflict with the assumption that objectives are fractal and can
>> always be represented more granular via subobjectives as can
>> sub-processes.
Barry> This would make the specification of every objective infinitely
Barry> complex.
This does not follow. I can specify a program with an infinite number of steps
with a single function call. Or, take the specification of the (infinitely
long) integers which can be done in two lines. In fact, all specifications of
the infinite that I have seen in full are not infinite.
I think this is a clear reason why an objective and it's specification need to
be consider separately.
Barry> I think you need to take granularity more seriously. Granularity
Barry> means that for each specific purpose we overlook, in specifying our
Barry> objectives, and in specifying our processes, the levels of finer
Barry> detail not relevant to that purpose.
"Not relevant" does not mean that it is not there.
Barry> Incidentally, my scepticism as concerns finding a place for states of
Barry> affairs in BFO, rests on a multi-year attempt to find a coherent
Barry> treatment of them:
Barry> http://ontology.buffalo.edu/smith/articles/logsvh.html
Barry> http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html
Barry> http://ontology.buffalo.edu/smith/articles/hloa.html
Barry> http://ontology.buffalo.edu/smith/articles/Questions.pdf
Okay. So, we can't do objectives?
Phil
Phillip Lord
>> So an objective is what is socially considered to be an desired ending to
>> a process; so, it is an objective for Eurovision to vote for a song, but
>> it's not an objective for Eurovision to vote based on the historical
>> politics of the countries involved.
>>
>> Which means that Eurovision 09 has an objective, even thought it hasn't
>> begun yet.
Barry> I agree with this. I believe that it lends more weight to Bjoern's
Barry> suggestion that we include, not objective, in BFO, but rather
Barry> objective specification. It is the latter that really exists. If and
Barry> when the plan associated with the objective specification is
Barry> realized, then we might also say that the objective is realized
Barry> (shorthand for something more complicated involving objective
Barry> specification); sometimes this will involve the terminal boundary of
Barry> a process of a specific sort (e.g. the liquid is brought to a certain
Barry> temperature at the conclusion of a heating process); sometimes not
Barry> (the objective might have been, e.g. to PREVENT the liquid ever
Barry> reaching a certain temperature). Conclusion: objectives do not form a
Barry> coherent universal on the side of reality; hence no objectives in
Barry> BFO. Just objective specifications. BS
I think that this is like arguing that the political boundaries of the UK do
not exist because they are socially defined, and that only maps of the
boundaries of the UK exist.
An objective is a socially defined boundary for a process -- in this case a
desired boundary.
My worry with ignoring objectives and using specifications is that it means
"specification" becomes a catch-all category; it could be argued, for example,
that we could just put the term "concept" underneath specification and then
stick all of our other terms under that; this would save the requirement for
on representing coherent universals on the side of reality. But it wouldn't be
very useful.
Phil
Barry Smith
On the contrary. The boundaries exist because the maps exist (and
because all the necessary treaties exist have been ratified)
The objective will be realized if the objective specification is
fulfilled. But since objectives can fall under so many different
sorts of categories (unlike spatial boundaries) there is no coherent
universal deserving the name 'objective'.
>An objective is a socially defined boundary for a process -- in this case a
>desired boundary.
The process has an end, which is a process boundary. But of course
many processes which are initiated in order to realize plans and
which are associated with objective specifications do not end with
the realization of any objective. Many of them fail. (Compare: many
intended maps of national boundaries do not correspond to any
national boundaries.)
>My worry with ignoring objectives and using specifications is that it means
>"specification" becomes a catch-all category; it could be argued, for example,
>that we could just put the term "concept" underneath specification and then
>stick all of our other terms under that; this would save the requirement for
>on representing coherent universals on the side of reality. But it wouldn't be
>very useful.
I understand your qualms. But I think whether the use of
'specification' will be useful is an empirical question.
BFO and OBI are being built, remember, on empirical grounds.
BS
>Phil
>
>
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
> >> Objective/State of affairs: Fact 1: Percoll solution has_density 1.064
> >> g/ml ... Fact 4: Percoll solution has_density 1.087 g/ml
>
> Barry> The objective cannot be a fact, since of course many objectives are
> Barry> never realized. Also the objective is (presumably) to have some
> Barry> specific percoll solution in some specific place have the mentioned
> Barry> density.
>
>Barry
>
>There are many functions and many roles which are never realised also. If
>objectives are socially defined, then they have an existence regardless of
>whether they actually come to pass?
There are clear cases where unrealized functions and roles exist
nonetheless -- e.g. the functions of almost all of your sperm; the
role of poor McX, who is appointed to a professorship and dies 1 minute later.
There are no clear cases where unrealized objectives exist.
> >> In case you mean it is an ending number of assertions or
> steps, then this
> >> is in conflict with the assumption that objectives are fractal and can
> >> always be represented more granular via subobjectives as can
> >> sub-processes.
>
> Barry> This would make the specification of every objective infinitely
> Barry> complex.
>
>This does not follow. I can specify a program with an infinite number of steps
>with a single function call. Or, take the specification of the (infinitely
>long) integers which can be done in two lines. In fact, all specifications of
>the infinite that I have seen in full are not infinite.
Good. I retract the mentioned argument.
>I think this is a clear reason why an objective and it's specification need to
>be consider separately.
Of course they are different. And it may be that we need to accept a
category of objective. But following the principle of low hanging
fruit, let us start with objective specification and see how far it
takes us. (It will not take us all the way if e.g. dogs have objectives.)
> Barry> I think you need to take granularity more seriously. Granularity
> Barry> means that for each specific purpose we overlook, in specifying our
> Barry> objectives, and in specifying our processes, the levels of finer
> Barry> detail not relevant to that purpose.
>
>"Not relevant" does not mean that it is not there.
Of course not.
> Barry> Incidentally, my scepticism as concerns finding a place
> for states of
> Barry> affairs in BFO, rests on a multi-year attempt to find a coherent
> Barry> treatment of them:
>
> Barry> http://ontology.buffalo.edu/smith/articles/logsvh.html
> Barry> http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html
> Barry> http://ontology.buffalo.edu/smith/articles/hloa.html
> Barry> http://ontology.buffalo.edu/smith/articles/Questions.pdf
>
>Okay. So, we can't do objectives?
It is at least very hard. So lets do the easier case (objective
specification) first.
BS
>Phil
>
>
Phillip Lord
Barry> A further problem is that we do not have a category 'state of
Barry> affairs' in BFO. I agree that a state of affairs is not in general
Barry> identifiable with the boundary of a process. But how, then, should
Barry> 'state of affairs' be defined? I have failed to find a way of
Barry> answering this question that is in keeping with BFO principles -- the
Barry> reason being (roughly) that states of affairs seem too closely tied
Barry> to their specifications to be acceptable as first-order entities in
Barry> their own right. Perhaps, if we follow Bjoern's line of reasoning,
Barry> and include in OBI a type 'specification of objective', we can avoid
Barry> this thorn. BS
I'm not sure that just saying "specification" rather than "state of affairs"
changes much.
The point with objective is surely that it is an intention about a process. I
can't see why this should not be expressed "in keeping with BFO". It seems to
be fairly analogous to "role". A "doctor" for example who intends (or is
intended to) do medicine.
Is a role a specification of the things that a continuant should do?
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> Barry> A further problem is that we do not have a category 'state of
> Barry> affairs' in BFO. I agree that a state of affairs is not in general
> Barry> identifiable with the boundary of a process. But how, then, should
> Barry> 'state of affairs' be defined? I have failed to find a way of
> Barry> answering this question that is in keeping with BFO
> principles -- the
> Barry> reason being (roughly) that states of affairs seem too closely tied
> Barry> to their specifications to be acceptable as first-order entities in
> Barry> their own right. Perhaps, if we follow Bjoern's line of reasoning,
> Barry> and include in OBI a type 'specification of objective', we can avoid
> Barry> this thorn. BS
>
>I'm not sure that just saying "specification" rather than "state of affairs"
>changes much.
One belongs to the realm of specifications, the other to the realm of
that about which specifications are formulated.
It is like the difference between 'Phillip Lord' and Phillip Lord.
The latter would exist even if there were no language.
>The point with objective is surely that it is an intention about a process.
The issue is: what is the it here?
Feel free to propose a definition.
>I
>can't see why this should not be expressed "in keeping with BFO". It seems to
>be fairly analogous to "role". A "doctor" for example who intends (or is
>intended to) do medicine.
'Fairly analogous', perhaps, but then: can you propose a definition.
Our computers cannot reason with 'fairly analogous'.
>Is a role a specification of the things that a continuant should do?
The specification of a role will typically specify things which its
bearer can and should do.
BS
>Phil
>
>
>
bpeters
> My worry with ignoring objectives and using specifications is that it means
> "specification" becomes a catch-all category; it could be argued, for example,
> that we could just put the term "concept" underneath specification and then
> stick all of our other terms under that; this would save the requirement for
> on representing coherent universals on the side of reality. But it wouldn't be
> very useful.
thought, so I may overlook obvious problems in my line of reasoning,
or repeat well known points of view. Still...
I don't think Phil and Alan's problem is unavoidable. The key lies in
instances and relations. Nobody confuses a dog with the specification
of what a dog is. An instance of a dog is a physical entity that bites
me and that I run away from. An instance of a specification of a dog
is an information entity that I read, communicate, and apply to
entities (in order to identify dogs and to start running early).
In this sense, I don't see how the class dog could end up as a
specification (via concept as suggested by Phil). I believe the
textual definition of the class dog might be an instance of a
specification. However, a universal is not the same as its textual
definition.
Daniel Schober
see inline....
The emphasis of my statement lies in the fact that objectives usually consist of multiple assertions and that these represent a state of affairs that someone wants to be true (fact) at some future time. So this depends on time handling - which is something that BFO has not yet tackled (nor documented?) in a sufficient way (E.g. Do you want to model possible future worlds within the ontology, or states the world would have been in when some processes would have been differrent?, ...)
If you want the ontology to be used in a simulation approach (e.g. forward chaining on objective states and verifying if they can be met) the outcome/endpoint would be these assertions as facts/endpoints.
For GandrKB I had experimented with JESS to model processes and simulate time (see below).
Of cause I concur that states are assertions as long as they are not realized. But this counts only as long as they are not modeled as objectives (which already implies that they are not necessarily valid at all times)).
Action/Process: [Christian, where is the problem you have with Processes as 'Actions'?] Process/manufacturers_one-step_procedure: (modeled as composite steps) Percoll medium dilute_with solution/10xPBS Percoll medium dilute_with destilled water ... As Christian, I see 'states of affairs' as a set of assertions/facts within the knowledgebase.Assertions and facts are ontologically two quite different things.
We need to be able to capture a state of affairs that is potentially true (a potential fact) at some time.
So, the range and domain of the assertion Christian proposes ( Action hasObjective StateOfAffairs ) is here broken down into many assertions. Is that right? I agree with CC: span:ProcessBoundary is *not quite* an exact synonym for "state of affairs," I am not sure how to define such boundaries... In case you mean it is an ending number of assertions or steps, then this is in conflict with the assumption that objectives are fractal and can always be represented more granular via subobjectives as can sub-processes.This would make the specification of every objective infinitely complex. I think you need to take granularity more seriously.
Granularity means that for each specific purpose we overlook, in specifying our objectives, and in specifying our processes, the levels of finer detail not relevant to that purpose.
I dont think span:ProcessBoundary will bring us any further since to me objectives are sets of states/ facts. A state can then trigger a process of cause, but is not the process itself.State = state of affairs? And what is a 'process of cause' when it's at home?
I believe without a clarification on how BFO models time (e.g. via a temporal logics approach like concurrent metatem (
http://books.google.com/books?id=JYcznFCN3xcC&printsec=frontcover&dq=weiss+gerhard&sig=azkE4-65Hs-J01tJiQdRMODBOt8#PPA69,M1 ) there will always be reoccuring discussions on these issues.
Object =atomic class complex item = composite/defined class (but also set of assertions?)Does not sound like W to me.I would like to know because we used the term 'states of affairs' in OBI before, e.g. see my obi-dev mail from 02/11/2007 13:46 ( <http://groups.google.com/group/obi-developer/msg/e1c08ba4bcc4773c?hl=en&>http://groups.google.com/group/obi-developer/msg/e1c08ba4bcc4773c?hl=en& )There you say:So in short a 'role' is a particular manifestation that occurs only under certain circumstances (natural, social or institutional contexts (Wittgensteins 'state of affairs' I believe)This is not a correct reading of W's state of affairs.
Incidentally, my scepticism as concerns finding a place for states of affairs in BFO, rests on a multi-year attempt to find a coherent treatment of them: http://ontology.buffalo.edu/smith/articles/logsvh.html http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html http://ontology.buffalo.edu/smith/articles/hloa.html http://ontology.buffalo.edu/smith/articles/Questions.pdf Barry
Cheers,
Daniel Schober.
P.S.: The Eurovision Song Contest has always been just a big political propaganda event... It was intended as a mass media advertisement for the 'EU' that none of the peoples really wanted. Panem et circenses....
------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Obi-devel mailing list Obi-...@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/obi-devel
-- __________________________________________________________________________________________ Dr. Daniel Schober NET Project - Ontologist The European Bioinformatics Institute email: sch...@ebi.ac.uk
EMBL Outstation - Hinxton direct: +44 (0)1223 494410 Wellcome Trust Genome Campus fax: +44 (0)1223 494 468
Cambridge CB10 1SD, UK Room: A3-141 (extension building) Project page: www.ebi.ac.uk/net-project Personal page: http://www.ebi.ac.uk/Information/Staff/person_maint.php?s_person_id=734
Barry Smith
You need to be much more careful with punctuation if we are to
resolve these subtle issues. Do you mean:
To say of something that it is an 'objective' means to say that some
agent wants certain assertions (representing states of affairs) to
become true in the future. This is what makes (what? the assertions?)
objectives.
'Assertion' is not quite the right word, since an assertion is
already believed by the relevant agent to be true.
So let's substitute 'assertable sentence'.
But now there are lots of assertable sentences which I want to become
true in the future which are not objectives. E.g. I want it to become
true that the sun will rise tomorrow, and the day after, and the day
after, etc.
Thus we need to add:
that some agent wants .... to become true in the future as a result
of the agents efforts.
All well and good. But this still does not tell us what an objective
is, ontologically.
It does, this, tell us what an objective specification is. The
beginnings of a definition might be:
objective specification = an assertable sentence that some agent
wants to become true as a result of his own efforts.
I think this still needs to embrace a dimension involving plans and
their realizations.
>The emphasis of my statement lies in the fact that objectives
>usually consist of multiple assertions
statements of objectives!!!!
specifications of objectives!!!!
please do not confuse representations with the entities they represent
>and that these represent a state of affairs that someone wants to be
>true (fact) at some future time.
If we do not have a good definition of 'state of affairs' then adding
the jargon bringeth no progress,
>So this depends on time handling - which is something that BFO has
>not yet tackled (nor documented?) in a sufficient way (E.g. Do you
>want to model possible future worlds within the ontology, or states
>the world would have been in when some processes would have been
>differrent?, ...)
BFO has embraced the principle of low hanging fruit. Possible worlds
(etc.) are hard to deal with. It is not clear that they exist.
Objective specifications are easy (or easier) to deal with. And it is
clear that they exist. The OBO coverage domain is full of them.
> If you want the ontology to be used in a simulation approach (e.g.
> forward chaining on objective states and verifying if they can be
> met) the outcome/endpoint would be these assertions as facts/endpoints.
Assertions are not facts/endpoints!!!!! Use-mention confusion again!
And not all objective specifications are specifications of endpoints.
The objective might be to STOP a process from coming to an end, e.g.
>For GandrKB I had experimented with JESS to model processes and
>simulate time (see below).
BFO is not a model, though if people want to use it for that they are
of course welcome to do so.
It is a representation of reality, analogous to a scientific theory.
>Of cause I concur that states are assertions
NO NO NO
>as long as they are not realized. But this counts only as long as
>they are not modeled as objectives (which already implies that they
>are not necessarily valid at all times)).
Bringing in the dimension of modeling just adds an extra layer of
difficulty. The problem of working out what entities like states of
affairs are is hard enough.
BS
>(<http://herzberg.ca.sandia.gov/jess/>http://herzberg.ca.sandia.gov/jess/)
>to define forward-chaining production rules that trigger the
>modification of a KB based according to rules that define input
>states and the action on the KB (e.g. remove instance from Class B,
>create instance in ClassB') to be carried out when the rule fires
>(its input facts are found in the KB). The cycles updating the KB
>would correspond to steps in time. The rules define processes and
>actions and the Ontology plus instances models the world state at a
>given time step. I believe the granularity we have to model
>processes is dependent on how we want to make use of them. I still
>have not seen a complete use case of an application that uses the
>ontology (and instaces).
>I believe without a clarification on how BFO models time (e.g. via a
>temporal logics approach like concurrent metatem (
><http://books.google.com/books?id=JYcznFCN3xcC&printsec=frontcover&dq=weiss+gerhard&sig=azkE4-65Hs-J01tJiQdRMODBOt8#PPA69,M1>http://books.google.com/books?id=JYcznFCN3xcC&printsec=frontcover&dq=weiss+gerhard&sig=azkE4-65Hs-J01tJiQdRMODBOt8#PPA69,M1
>) there will always be reoccuring discussions on these issues.
>>
>>
>>
>>>
>>>Matthews mail hints that a look at the BDI models could be valuable
>>>here:
>>><http://en.wikipedia.org/wiki/BDI_software_agent#Intentions>http://en.wikipedia.org/wiki/BDI_software_agent#Intentions
>><http://ontology.buffalo.edu/smith/articles/logsvh.html>http://ontology.buffalo.edu/smith/articles/logsvh.html
>>http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html
>><http://ontology.buffalo.edu/smith/articles/hloa.html>http://ontology.buffalo.edu/smith/articles/hloa.html
>><http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/>http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
>>_______________________________________________
>>Obi-devel mailing list
>><mailto:Obi-...@lists.sourceforge.net>Obi-...@lists.sourceforge.net
>>https://lists.sourceforge.net/lists/listinfo/obi-devel
>>
>
>
>--
>__________________________________________________________________________________________
>
>Dr. Daniel Schober
>
>NET Project - Ontologist
>
>The European Bioinformatics
Phillip Lord
>> I think that this is like arguing that the political boundaries of the UK
>> do not exist because they are socially defined, and that only maps of the
>> boundaries of the UK exist.
Barry> On the contrary. The boundaries exist because the maps exist (and
Barry> because all the necessary treaties exist have been ratified)
This isn't true. Boundaries of all sorts exist without a formal specification.
In most cases, the maps and international treaties simply formalised existing
territorial claims.
If you want a biological example, take the territorial boundary of an
predatory animal, or group of animals.
Barry> The objective will be realized if the objective specification is
Barry> fulfilled. But since objectives can fall under so many different
Barry> sorts of categories (unlike spatial boundaries) there is no coherent
Barry> universal deserving the name 'objective'.
>> An objective is a socially defined boundary for a process -- in this case
>> a desired boundary.
Barry> The process has an end, which is a process boundary.
This is an assumption which is not true of processes in general, although I
agree it is a probably reasonable assumption within biomedicine.
Of course, in many cases we don't know that the process has ended.
Barry> But of course many processes which are initiated in order to realize
Barry> plans and which are associated with objective specifications do not
Barry> end with the realization of any objective. Many of them fail.
Barry> (Compare: many intended maps of national boundaries do not correspond
Barry> to any national boundaries.)
Indeed, it is true that many of them fail. But many of them do not. That not
all objectives come to pass (and are not therefore real) surely does not hold
for objectives from the past which actually come to pass.
>> My worry with ignoring objectives and using specifications is that it
>> means "specification" becomes a catch-all category; it could be argued,
>> for example, that we could just put the term "concept" underneath
>> specification and then stick all of our other terms under that; this
>> would save the requirement for on representing coherent universals on the
>> side of reality. But it wouldn't be very useful.
Barry> I understand your qualms. But I think whether the use of
Barry> 'specification' will be useful is an empirical question. BFO and OBI
Barry> are being built, remember, on empirical grounds. BS
The question is not whether specification will be useful, but whether
objective will be useful. It appears that it already has been. Apart from the
worry that we can have a specification for something that doesn't exist or
can't be placed in the ontology.
When you are defining "Specification of Objective" surely this suggest the
need for a concept of objective.
Phil
Phillip Lord
b> Phil wrote:
>> My worry with ignoring objectives and using specifications is that it
>> means "specification" becomes a catch-all category; it could be argued,
>> for example, that we could just put the term "concept" underneath
>> specification and then stick all of our other terms under that; this
>> would save the requirement for on representing coherent universals on the
>> side of reality. But it wouldn't be very useful.
b> Alan voiced a very similar concern. I am new + naive to this line of
b> thought, so I may overlook obvious problems in my line of reasoning, or
b> repeat well known points of view. Still...
b> I don't think Phil and Alan's problem is unavoidable. The key lies in
b> instances and relations. Nobody confuses a dog with the specification of
b> what a dog is. An instance of a dog is a physical entity that bites me
b> and that I run away from. An instance of a specification of a dog is an
b> information entity that I read, communicate, and apply to entities (in
b> order to identify dogs and to start running early).
b> In this sense, I don't see how the class dog could end up as a
b> specification (via concept as suggested by Phil). I believe the textual
b> definition of the class dog might be an instance of a specification.
b> However, a universal is not the same as its textual definition.
Bjoern
My suggestion that we use "concept" in this way was a reduction to absurdity.
Sorry if my poor wording confused you.
It's clear that a specification of the dog and dog are not the same thing.
What worries me is an ontology which suggests that the former exists but the
latter does not.
Phil
Phillip Lord
>> There are many functions and many roles which are never realised also. If
>> objectives are socially defined, then they have an existence regardless
>> of whether they actually come to pass?
Barry> There are clear cases where unrealized functions and roles exist
Barry> nonetheless -- e.g. the functions of almost all of your sperm; the
Barry> role of poor McX, who is appointed to a professorship and dies 1
Barry> minute later. There are no clear cases where unrealized objectives
Barry> exist.
The objective of a surgical proceedure to increase the life expectancy of the
patient.
So, the surgeon manages to kill the patient; clearly the objective is not
met; this does not mean that there are not occasions when the patient does
manage to survive.
Barry> Incidentally, my scepticism as concerns finding a place
>> for states of
Barry> affairs in BFO, rests on a multi-year attempt to find a coherent
Barry> treatment of them:
>>
Barry> http://ontology.buffalo.edu/smith/articles/logsvh.html
Barry> http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html
Barry> http://ontology.buffalo.edu/smith/articles/hloa.html
Barry> http://ontology.buffalo.edu/smith/articles/Questions.pdf
>>
>> Okay. So, we can't do objectives?
Barry> It is at least very hard. So lets do the easier case (objective
Barry> specification) first. BS
Okay. When you are defining "specification of objective", how are you going to
define objective?
Phil
Phillip Lord
>> The point with objective is surely that it is an intention about a
>> process.
Barry> The issue is: what is the it here? Feel free to propose a definition.
>> I can't see why this should not be expressed "in keeping with BFO". It
>> seems to be fairly analogous to "role". A "doctor" for example who
>> intends (or is intended to) do medicine.
Barry> 'Fairly analogous', perhaps, but then: can you propose a definition.
Barry> Our computers cannot reason with 'fairly analogous'.
We haven't talked at all about the computability of definitions; it's a
somewhat separate issue and not one which seems to be frequently considered
here. The point of my discussion was to understand the requirements for the
definition.
A Objective is a state of affairs which is not essential to a process but
which reflects the desired outcome for that process by the actors who
initiated it.
Just a starter for 10; I'm sure it could be improved.
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> There are many functions and many roles which are never
> realised also. If
> >> objectives are socially defined, then they have an existence regardless
> >> of whether they actually come to pass?
>
> Barry> There are clear cases where unrealized functions and roles exist
> Barry> nonetheless -- e.g. the functions of almost all of your sperm; the
> Barry> role of poor McX, who is appointed to a professorship and dies 1
> Barry> minute later. There are no clear cases where unrealized objectives
> Barry> exist.
>
>The objective of a surgical proceedure to increase the life expectancy of the
>patient.
>
>So, the surgeon manages to kill the patient; clearly the objective is not
>met; this does not mean that there are not occasions when the patient does
>manage to survive.
And then the objective is realized.
The issue is to find a clear case of an unrealized objective which exists.
In my view this is a bit like trying to find a clear case of a
cancelled performance, or of a declined handshake, or of an absent nipple.
> Barry> Incidentally, my scepticism as concerns finding a place
> >> for states of
> Barry> affairs in BFO, rests on a multi-year attempt to find a coherent
> Barry> treatment of them:
> >>
> Barry> http://ontology.buffalo.edu/smith/articles/logsvh.html
> Barry> http://ontology.buffalo.edu/smith/articles/cogsvh/cogsvh.html
> Barry> http://ontology.buffalo.edu/smith/articles/hloa.html
> Barry> http://ontology.buffalo.edu/smith/articles/Questions.pdf
> >>
> >> Okay. So, we can't do objectives?
>
> Barry> It is at least very hard. So lets do the easier case (objective
> Barry> specification) first. BS
>
>
>Okay. When you are defining "specification of objective", how are you going to
>define objective?
For these purposes we can take any good dictionary definition
e.g.
the goal intended to be attained (and which is believed to be attainable)
Something worked toward or striven for; a goal
All such definitions are likely to rest on the use of terms like
'goal' -- which are near-synonyms of 'objective' -- I doubt that we
can get much further, leading me to believe that 'objective' should
be treated as a primitive'. The issue for us here is whether it
should be included in BFO, either as a term in its own right (for
which it is surely too specific, being restricted to organisms with
intentions) or as a child of an existing term. Since objectives can
include entities of so many ontologically different sorts, the latter
cannot work. But 'specification of objective', in contrast, can be
included as a child of 'information object'
BS
>Phil
>
>
Phillip Lord
>> The objective of a surgical proceedure to increase the life expectancy of
>> the patient.
>>
>> So, the surgeon manages to kill the patient; clearly the objective is not
>> met; this does not mean that there are not occasions when the patient
>> does manage to survive.
Barry> And then the objective is realized. The issue is to find a clear case
Barry> of an unrealized objective which exists. In my view this is a bit
Barry> like trying to find a clear case of a cancelled performance, or of a
Barry> declined handshake, or of an absent nipple.
So you are saying that surgery only has the objective of increasing life span
AFTER the surgery has completed? Or if it succeeds?
>> Okay. When you are defining "specification of objective", how are you
>> going to define objective?
Barry> For these purposes we can take any good dictionary definition e.g.
Barry> the goal intended to be attained (and which is believed to be
Barry> attainable) Something worked toward or striven for; a goal
Barry> All such definitions are likely to rest on the use of terms like
Barry> 'goal' -- which are near-synonyms of 'objective' -- I doubt that we
Barry> can get much further, leading me to believe that 'objective' should
Barry> be treated as a primitive'. The issue for us here is whether it
Barry> should be included in BFO, either as a term in its own right (for
Barry> which it is surely too specific, being restricted to organisms with
Barry> intentions) or as a child of an existing term. Since objectives can
Barry> include entities of so many ontologically different sorts, the latter
Barry> cannot work. But 'specification of objective', in contrast, can be
Barry> included as a child of 'information object' BS
Again, I don't understand this. If objective means, say, 5 different things,
then how is "specification of objective" not also a conflation of 5 different
things.
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> The objective of a surgical proceedure to increase the life
> expectancy of
> >> the patient.
> >>
> >> So, the surgeon manages to kill the patient; clearly the
> objective is not
> >> met; this does not mean that there are not occasions when the patient
> >> does manage to survive.
>
> Barry> And then the objective is realized. The issue is to find a
> clear case
> Barry> of an unrealized objective which exists. In my view this is a bit
> Barry> like trying to find a clear case of a cancelled performance, or of a
> Barry> declined handshake, or of an absent nipple.
>
>So you are saying that surgery only has the objective of increasing life span
>AFTER the surgery has completed? Or if it succeeds?
I am saying that there is a clear sense in which the objective exists
only after a successful process is completed
Before that we have an intention to reach some objective, and perhaps
a specification thereof
The phrase 'intention to reach the objective so-and-so' then bears
ontological commitment only on the 'intention' clause (and similarly
with the corresponding phrase referring to a specification).
Mother: What would you like for Christmas, Phillip?
Phillip: a trip to Disney World, please
The putative referent of 'Phillip's trip' does not exist until (a)
Christmas arrives, (b) plan successfully completed.
> >> Okay. When you are defining "specification of objective", how are you
> >> going to define objective?
>
> Barry> For these purposes we can take any good dictionary definition e.g.
> Barry> the goal intended to be attained (and which is believed to be
> Barry> attainable) Something worked toward or striven for; a goal
>
> Barry> All such definitions are likely to rest on the use of terms like
> Barry> 'goal' -- which are near-synonyms of 'objective' -- I doubt that we
> Barry> can get much further, leading me to believe that 'objective' should
> Barry> be treated as a primitive'. The issue for us here is whether it
> Barry> should be included in BFO, either as a term in its own right (for
> Barry> which it is surely too specific, being restricted to organisms with
> Barry> intentions) or as a child of an existing term. Since objectives can
> Barry> include entities of so many ontologically different sorts,
> the latter
> Barry> cannot work. But 'specification of objective', in contrast, can be
> Barry> included as a child of 'information object' BS
>
>
>Again, I don't understand this. If objective means, say, 5 different things,
>then how is "specification of objective" not also a conflation of 5 different
>things.
Believe in numbers, do you?
You think we can count emails? apples? trips to Disney World? prime
factors of 81? lakes in Finland?
If yes, does this mean that there are 5 different kinds of numbers?
BS
>Phil
>
>
Barry Smith
Since 'state of affairs' isn't in BFO either, perhaps you can tell us
what you mean by that?
Also, 'desired outcome' sounds like just another word for 'objective'
BS
Barry Smith
We have adopted, remember, the principle of low hanging fruit.
"Specification of objective" is low hanging fruit. "Objective" is
hard. So let's work with the former first. No one is suggesting that
(realized) objectives do not exist.
BS
>Phil
>
>
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> I think that this is like arguing that the political
> boundaries of the UK
> >> do not exist because they are socially defined, and that only
> maps of the
> >> boundaries of the UK exist.
>
> Barry> On the contrary. The boundaries exist because the maps exist (and
> Barry> because all the necessary treaties exist have been ratified)
>
>This isn't true. Boundaries of all sorts exist without a formal specification.
Of course. But not boundaries between nations, methinks. (Not even
the boundaries of nations bounded by coastlines (of which the UK post
1922, is not an example).
>In most cases, the maps and international treaties simply formalised existing
>territorial claims.
'Simply' is a good word.
>If you want a biological example, take the territorial boundary of an
>predatory animal, or group of animals.
You introduced the phrase 'poltitical boundaries'
> Barry> The objective will be realized if the objective specification is
> Barry> fulfilled. But since objectives can fall under so many different
> Barry> sorts of categories (unlike spatial boundaries) there is no coherent
> Barry> universal deserving the name 'objective'.
>
>
>
>
> >> An objective is a socially defined boundary for a process --
> in this case
> >> a desired boundary.
>
> Barry> The process has an end, which is a process boundary.
>
>This is an assumption which is not true of processes in general, although I
>agree it is a probably reasonable assumption within biomedicine.
>
>Of course, in many cases we don't know that the process has ended.
I agree
> Barry> But of course many processes which are initiated in order to realize
> Barry> plans and which are associated with objective specifications do not
> Barry> end with the realization of any objective. Many of them fail.
> Barry> (Compare: many intended maps of national boundaries do not
> correspond
> Barry> to any national boundaries.)
>
>Indeed, it is true that many of them fail. But many of them do not. That not
>all objectives come to pass (and are not therefore real) surely does not hold
>for objectives from the past which actually come to pass.
I agree
> >> My worry with ignoring objectives
that is certainly not what we are doing
>and using specifications is that it
> >> means "specification" becomes a catch-all category; it could be argued,
> >> for example, that we could just put the term "concept" underneath
> >> specification and then stick all of our other terms under that; this
> >> would save the requirement for on representing coherent
> universals on the
> >> side of reality. But it wouldn't be very useful.
> Barry> I understand your qualms. But I think whether the use of
> Barry> 'specification' will be useful is an empirical question. BFO and OBI
> Barry> are being built, remember, on empirical grounds. BS
>
>
>The question is not whether specification will be useful, but whether
>objective will be useful. It appears that it already has been.
Not within a formal ontology, as far as I know. That is, I know of no
successful use of a formal ontology which contains a universal like
'objective'.
>Apart from the
>worry that we can have a specification for something that doesn't exist or
>can't be placed in the ontology.
Relax.
>When you are defining "Specification of Objective" surely this suggest the
>need for a concept of objective.
Yes. It is, as I argued earlier, (a) likely primitive (i.e. definable
only with the help of terms like 'goal', or 'desired outcome'), (b)
likely not an ontological coherent category (just as 'pet' is not a
biologically coherent category).
BS
>Phil
>
>
>
bpeters
> It's clear that a specification of the dog and dog are not the same thing.
> What worries me is an ontology which suggests that the former exists but the
> latter does not.
>
> Phil
straight.
I would now argue further that there are specifications for which
there is no corresponding entity in reality. I can specify a unicorn.
That specification exists. The physical object meeting these
specifications doesn't.
If the above is acceptable, then having objective specifications
without a corresponding 'real' objective seems to me be the correct
implementation.
From a gut feeling, it seems to me that there are more cognitive
information entities for which this is the right implementation. A
careful implementation of the relations between such information
entities and physical reality will be necessary. For specifications,
that seems to be 'applies to' and 'conforms with'. I do also believe
that this is necessary to describe scientific discovery.
Phillip Lord
>> So you are saying that surgery only has the objective of increasing life
>> span AFTER the surgery has completed? Or if it succeeds?
Barry> I am saying that there is a clear sense in which the objective exists
Barry> only after a successful process is completed Before that we have an
Barry> intention to reach some objective, and perhaps a specification
Barry> thereof The phrase 'intention to reach the objective so-and-so' then
Barry> bears ontological commitment only on the 'intention' clause (and
Barry> similarly with the corresponding phrase referring to a
Barry> specification). Mother: What would you like for Christmas, Phillip?
Barry> Phillip: a trip to Disney World, please
Barry> The putative referent of 'Phillip's trip' does not exist until (a)
Barry> Christmas arrives, (b) plan successfully completed.
Okay. I think that this is wrong. Objective is an intention to reach a point.
Mother: What will you like to be when you grow up
Barry: I would like to be a philosopher who sends lots of emails.
The objective here is to become a philosopher. Now consider a view point here;
these statements may have happened in the past, or may be happening now. In
either case, an objective is being expressed, regardless of whether or not
Barry becomes a philosopher.
Take these statements:
I plan to make a journey so that I can reach Birmingham.
I plan to make a journey so that I can reach Xanadu.
I plan to make a journey so that I can reach Atlantis.
I think that all three of these statements express an objective. Atlantis
never existed (probably). Xanadu did exist in the past. Birmingham does exist.
Barry> All such definitions are likely to rest on the use of terms like
Barry> 'goal' -- which are near-synonyms of 'objective' -- I doubt that we
Barry> can get much further, leading me to believe that 'objective' should
Barry> be treated as a primitive'. The issue for us here is whether it
Barry> should be included in BFO, either as a term in its own right (for
Barry> which it is surely too specific, being restricted to organisms with
Barry> intentions) or as a child of an existing term. Since objectives can
Barry> include entities of so many ontologically different sorts,
>> the latter
Barry> cannot work. But 'specification of objective', in contrast, can be
Barry> included as a child of 'information object' BS
>>
>>
>> Again, I don't understand this. If objective means, say, 5 different
>> things, then how is "specification of objective" not also a conflation of
>> 5 different things.
Barry> Believe in numbers, do you?
I think that they are a useful model, yes. I've never managed to find out
whether you believe they exist or not.
Barry> You think we can count emails? apples? trips to Disney World? prime
Barry> factors of 81? lakes in Finland? If yes, does this mean that there
Barry> are 5 different kinds of numbers? BS
It depends on how you think of numbers. If you think of them entities to which
you can apply a set of functions, then indeed there are many different types
of numbers even if the functions can be specified generically -- that is, you
can't add apples and oranges.
My point is that if objective means many different things then you will want
to describe these things, so that it is clear what kind of specification we
are talking about.
Phil
Phillip Lord
>> We haven't talked at all about the computability of definitions; it's a
>> somewhat separate issue and not one which seems to be frequently
>> considered here. The point of my discussion was to understand the
>> requirements for the definition.
>>
>> A Objective is a state of affairs which is not essential to a process but
>> which reflects the desired outcome for that process by the actors who
>> initiated it.
Barry> Since 'state of affairs' isn't in BFO either, perhaps you can tell us
Barry> what you mean by that? Also, 'desired outcome' sounds like just
Barry> another word for 'objective' BS
Barry
You can't define all words on pain of infinite regress.
Desired outcome is different from objective because it specifies the
intentionality -- that some actor, person, social process is involved. Still,
I think I can try for a better definition based on your helpful comments.
An Objective is a some result or end that is not essential to a process but
which reflects the desire for that process by some natural, social or
institution context.
Phil
Phillip Lord
Barry> On the contrary. The boundaries exist because the maps exist (and
Barry> because all the necessary treaties exist have been ratified)
>>
>> This isn't true. Boundaries of all sorts exist without a formal
>> specification.
Barry> Of course. But not boundaries between nations, methinks. (Not even
Barry> the boundaries of nations bounded by coastlines (of which the UK post
Barry> 1922, is not an example).
I think that there are nations where no one is quite sure whether the boundary
is. Nowadays, of course, it's relatively easy to specify a boundary. In the
past, it was not, because our mapping technology wasn't advanced enough.
Nations existed, I would suggest, before accurate mapping technology.
>> In most cases, the maps and international treaties simply formalised
>> existing territorial claims.
Barry> 'Simply' is a good word.
Probably a poor choice of words in the context:-)
>> If you want a biological example, take the territorial boundary of an
>> predatory animal, or group of animals.
Barry> You introduced the phrase 'poltitical boundaries'
Yes, as an example of a boundary. I think that territorial boundaries have
some similar properties.
>> The question is not whether specification will be useful, but whether
>> objective will be useful. It appears that it already has been.
Barry> Not within a formal ontology, as far as I know. That is, I know of no
Barry> successful use of a formal ontology which contains a universal like
Barry> 'objective'.
Well, it's good to be ground-breaking.
More seriously, I might agree that it's not appropriate in BFO itself. But
having some idea of where it fits in BFO probably is appropriate.
>> Apart from the worry that we can have a specification for something that
>> doesn't exist or can't be placed in the ontology.
Barry> Relax.
>> When you are defining "Specification of Objective" surely this suggest
>> the need for a concept of objective.
Barry> Yes. It is, as I argued earlier, (a) likely primitive (i.e. definable
Barry> only with the help of terms like 'goal', or 'desired outcome'), (b)
Barry> likely not an ontological coherent category (just as 'pet' is not a
Barry> biologically coherent category). BS
Pet is a role no? And as such an coherent universal on the side of reality?
Phil
Barry Smith
>
>My point is that if objective means many different things then you will want
>to describe these things, so that it is clear what kind of specification we
>are talking about.
If and when we ever reach that stage, of analyzing the sorts of
entities that get specified in specifications of objectives, then
this sort of analysis could indeed be carried out. One large subgroup
will likely be: terminal boundaries of processes.
BS
Barry Smith
We have for every x, if x is a role then there is some y such that y bears x.
'Objective' could not designate a role because it could not satisfy this axiom.
BS
Phillip Lord
>> Pet is a role no? And as such an coherent universal on the side of
>> reality?
Barry> We have for every x, if x is a role then there is some y such that y
Barry> bears x. 'Objective' could not designate a role because it could not
Barry> satisfy this axiom. BS
Again, I am extending an analogy. If pet can be a role which inheres in an
animal, then an objective can be something which inheres in a process.
I think that the cornerstone difficulty here, is that RealizableEntity
(shouldn't this be RelisableContinuant BTW) inheres only in Continuants. So,
considering informally the requirements, it seems to me that:
1) Objective is something related to a process. It's processes that have
objectives.
2) Objective is defined by some agent.
3) Objective is a hope for something that might not actually come to pass,
even though we want it to. The hope is there whether it actually happens.
4) An objective is not intrinsic to the process. The same process may or may
not have an objective.
Likewise,
1) A role is related to a continuant
2) A role is defined by some agent
3) A role is a hope or possibility; i.e. you can be a surgeon without ever
actually doing surgery. The hope is there whether it actually happens.
4) A role is not intrinsic to the continuant. The same continuant may or may
not have a role.
It seems that a similar solution to that used to represent a "role" works for
"objective".
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> Pet is a role no? And as such an coherent universal on the side of
> >> reality?
>
> Barry> We have for every x, if x is a role then there is some y such that y
> Barry> bears x. 'Objective' could not designate a role because it could not
> Barry> satisfy this axiom. BS
>
>Again, I am extending an analogy. If pet can be a role which inheres in an
>animal, then an objective can be something which inheres in a process.
If you want the objective to inhere in the process before it is
concluded, and including processes that are interrupted (as well as
processes that fail to reach their objectives for other reasons) then
this would mean that the objective would not exist in the future, as
seems to be required (the objective is something like an aimed for
end point, isn't it).
This is okay; but then the implied analysis is not of the objective
as such, but rather of something like the objective as it exists in
the mind of the agent, i.e. again, something like the objective
specification (or a hope, as you have it below).
>I think that the cornerstone difficulty here, is that RealizableEntity
>(shouldn't this be RelisableContinuant BTW) inheres only in Continuants. So,
>considering informally the requirements, it seems to me that:
>
>1) Objective is something related to a process. It's processes that have
> objectives.
>2) Objective is defined by some agent.
>3) Objective is a hope for something that might not actually come to pass,
> even though we want it to. The hope is there whether it actually happens.
BFO can do hope. No problem.
>4) An objective is not intrinsic to the process. The same process may or may
> not have an objective.
>
I think this is just wrong. Where continuants have non-intrinsic
parts and qualities (you can be suntanned, or not), all parts and
qualities of processes are intrinsic to them. Your process of walking
up the hill is a different process according to whether you are doing
to reach the top or not.
>Likewise,
>
>1) A role is related to a continuant
>2) A role is defined by some agent
>3) A role is a hope or possibility; i.e. you can be a surgeon without ever
> actually doing surgery. The hope is there whether it actually happens.
>4) A role is not intrinsic to the continuant. The same continuant may or may
> not have a role.
This all holds for roles (which are always of continuants); but there
is no analogy for occurrents, unfortunately.
The reason is that continuants can continue to exist while they take
on and lose roles.
Processes cannot do this; as soon as a process-role-analogue is in
the picture, the process becomes a DIFFERENT PROCESS.
BS
Phillip Lord
>> Again, I am extending an analogy. If pet can be a role which inheres in
>> an animal, then an objective can be something which inheres in a process.
Barry> If you want the objective to inhere in the process before it is
Barry> concluded, and including processes that are interrupted (as well as
Barry> processes that fail to reach their objectives for other reasons) then
Barry> this would mean that the objective would not exist in the future, as
Barry> seems to be required (the objective is something like an aimed for
Barry> end point, isn't it). This is okay; but then the implied analysis is
Barry> not of the objective as such, but rather of something like the
Barry> objective as it exists in the mind of the agent, i.e. again,
Barry> something like the objective specification (or a hope, as you have it
Barry> below).
I don't think that an objective does exist any more once the process has
ended. In normal use, we would refer to it in the past tense.
The objective of the travelling is to get to Newcastle.
The objective of the travelling from last was to get to Newcastle.
It seems clear to me that we need to refer to things that do not exist, but
did exist in the past. Likewise, I would want to be able to refer to process
of walking that I did a while back, even though I am not doing it any longer.
So, that the objective doesn't exist any more, doesn't seem a problem. How is
this handled in BFO? How do I refer to thinks that don't exist?
>> I think that the cornerstone difficulty here, is that RealizableEntity
>> (shouldn't this be RelisableContinuant BTW) inheres only in Continuants.
>> So, considering informally the requirements, it seems to me that:
>>
>> 1) Objective is something related to a process. It's processes that have
>> objectives.
>> 2) Objective is defined by some agent.
>> 3) Objective is a hope for something that might not actually come to
>> pass,
>> even though we want it to. The hope is there whether it actually happens.
Barry> BFO can do hope. No problem.
>> 4) An objective is not intrinsic to the process. The same process may or
>> may
>> not have an objective.
>>
Barry> I think this is just wrong. Where continuants have non-intrinsic
Barry> parts and qualities (you can be suntanned, or not), all parts and
Barry> qualities of processes are intrinsic to them. Your process of walking
Barry> up the hill is a different process according to whether you are doing
Barry> to reach the top or not.
Why?
Take, for instance, travelling. If I am travelling to Worcester on a train, is
there a new process everytime the train changes speed.
Phil
Barry Smith
BFO terms refer only to universals, not to particulars
There is no universal 'objective' in BFO
However, if you can point to an instance of an object that exists, in
the present or in the past, and point to a BFO category which it
instantiates, then you can refer to that, if you wish. There is no
BFO category instantiated by all and only objectives. There may be an
OBI category instantiated by all and only objective specifications.
> >> I think that the cornerstone difficulty here, is that RealizableEntity
> >> (shouldn't this be RelisableContinuant BTW) inheres only in Continuants.
> >> So, considering informally the requirements, it seems to me that:
> >>
> >> 1) Objective is something related to a process. It's processes that have
> >> objectives.
> >> 2) Objective is defined by some agent.
> >> 3) Objective is a hope for something that might not actually come to
> >> pass,
> >> even though we want it to. The hope is there whether it
> actually happens.
>
> Barry> BFO can do hope. No problem.
>
> >> 4) An objective is not intrinsic to the process. The same process may or
> >> may
> >> not have an objective.
> >>
>
> Barry> I think this is just wrong. Where continuants have non-intrinsic
> Barry> parts and qualities (you can be suntanned, or not), all parts and
> Barry> qualities of processes are intrinsic to them. Your process
> of walking
> Barry> up the hill is a different process according to whether
> you are doing
> Barry> to reach the top or not.
>
>Why?
>
>Take, for instance, travelling. If I am travelling to Worcester on a train, is
>there a new process everytime the train changes speed.
You can create as many fiat partitions of any given process as you like.
BS
>Phil
>
>
Phillip Lord
>>
>> I don't think that an objective does exist any more once the process has
>> ended. In normal use, we would refer to it in the past tense.
>>
>> The objective of the travelling is to get to Newcastle. The objective of
>> the travelling from last was to get to Newcastle.
>>
>> It seems clear to me that we need to refer to things that do not exist,
>> but did exist in the past. Likewise, I would want to be able to refer to
>> process of walking that I did a while back, even though I am not doing it
>> any longer.
>>
>> So, that the objective doesn't exist any more, doesn't seem a problem.
>> How is this handled in BFO? How do I refer to thinks that don't exist?
Barry> BFO terms refer only to universals, not to particulars
Yes.
Barry> There is no universal 'objective' in BFO However, if you can point to
Barry> an instance of an object that exists, in the present or in the past,
Barry> and point to a BFO category which it instantiates, then you can refer
Barry> to that, if you wish.
Okay, so that objectives cease to exist once their process finishes is not a
problem.
Barry> There is no BFO category instantiated by all and only objectives.
Barry> There may be an OBI category instantiated by all and only objective
Barry> specifications.
Yes, I know. This has been covered extensively in the last few days. The
question is, where would such a category go under BFO.
Barry> I think this is just wrong. Where continuants have non-intrinsic
Barry> parts and qualities (you can be suntanned, or not), all parts and
Barry> qualities of processes are intrinsic to them. Your process
>> of walking
Barry> up the hill is a different process according to whether
>> you are doing
Barry> to reach the top or not.
>>
>> Why?
>>
>> Take, for instance, travelling. If I am travelling to Worcester on a
>> train, is there a new process everytime the train changes speed.
Barry> You can create as many fiat partitions of any given process as you
Barry> like. BS
Then your argument that a process has only intrinsic parts is wrong.
So, we seem to have an answer. Objective would naturally be placed under
SpecificallyDependentContinuant.
Phil
Barry Smith
objective specification would go under information entity
'objective' is like 'pattern' -- it crosses BFO categorial boundaries
> Barry> I think this is just wrong. Where continuants have non-intrinsic
> Barry> parts and qualities (you can be suntanned, or not), all parts and
> Barry> qualities of processes are intrinsic to them. Your process
> >> of walking
> Barry> up the hill is a different process according to whether
> >> you are doing
> Barry> to reach the top or not.
> >>
> >> Why?
> >>
> >> Take, for instance, travelling. If I am travelling to Worcester on a
> >> train, is there a new process everytime the train changes speed.
>
> Barry> You can create as many fiat partitions of any given process as you
> Barry> like. BS
>
>
>Then your argument that a process has only intrinsic parts is wrong.
Did the first 10 minutes of the 400th century BC exist, 42000(-ish) years ago?
Or did this fiat stretch of time come into existence only when man
created units called 'century' and 'second'?
I tend to thing that it existed already from the start.
Same applies to all the fiat parts of processes which we can, if we
want, delineate.
>So, we seem to have an answer. Objective would naturally be placed under
>SpecificallyDependentContinuant.
we have been round this circle several times already; there are many
objectives (e.g. keep this water boiling) which do not fit under that heading
BS
>
jennifer fostel
On Sun, Jun 1, 2008 at 2:56 PM, Barry Smith <phis...@buffalo.edu> wrote:
>
> At 04:01 PM 5/29/2008, bpeters wrote:
>
>>We are working with BFO as it currently exists in OWL. From CC's
>>proposal I take that we need a StateOfAffairs, Action, and a relation
>>objectiveOf, none of which currently exist in BFO. This is motivated
>>by not wanting to call objectives information, but not really wanting
>>to call them states either and putting them in a relation. I am afraid
>>if plans and their relation to objectives is dealt with in a similar
>>way, this will lead to major BFO revisions. Given the fact that we
>>need to use something for development of OBI now (which can always be
>>revisited later), how horrible is it to stick to
>>
>>'objective specification' is a realizable information entity that
>>specifies a desired endpoint (process boundary) of a process'
>
> Drop 'realizable', I think.
> An objective specification is not a process or plan specification,
> and even the latter are not clearly a realizables, in contrast to: plan.
>
>>'plan specification': 'is a realizable information entity that
>>specifies a series of
>>steps (sub-objectives) to be carried out to accomplish an objective'.
>
> Here, too, drop 'realizable' and drop 'sub-objective'
> Try to define 'step'.
>
>>These
>
> rather: what they specify
>
>>can be concretized and realized by an actor. We would work
>>within what we have and understand.
>>
>>To support this, I would argue that there *are* specifications, and
>>that they are information entities.
>
> I agree
>
>>I understand that it is important
>>to not confuse these with the entities they specify, but that won't
>>make them go away. Their relationships should instead be made
>>explicit:
>
> Exactly
>
>>A specification applies_to a kind of entity each of which
>>can comform_with the specification or not. An objective specification
>>applies_to process boundaries. Some process_boundaries conform_with
>>the objective specification, others don't.
>>
>>
>>
>>
>>
>>
>>
>>
>>On May 29, 4:42 am, Phillip Lord <phillip.l...@newcastle.ac.uk> wrote:
>> > >>>>> "CC" == Cristian Cocos <cri...@ieee.org> writes:
>> >
>> > CC> Otherwise, as I see it, there are two major obstacles to construing
>> > CC> objectives as a subclass of process boundaries simpliciter:
>> >
>> > CC> 1. as I was pointing out earlier, span:ProcessBoundary is
>> *not quite* an
>> > CC> exact synonym for "state of affairs," but closer to "event"--which,
>> > CC> though related to a state, is qualitatively something else;
>> >
>> > I wasn't sure that I got that, but I'll take your word for it.
>> >
>> > CC> 2. as you very well point out, the term "objective" connotes
>> > CC> potentiality to
>> > CC> a much larger extent than other BFO universals (like
>> "Process," say)--to
>> > CC> such an extent that it just cannot be ignored: it is not
>> only *actual*
>> > CC> states of affairs that are objectives, but *potential* (or
>> hypothetical)
>> > CC> states of affairs too.
>> >
>> > Well, both role and function are potentials as well. Most sperm don't
>> > fertilise anything. Doctors don't necessarily see any patients (during the
>> > golf season).
>> >
>> > CC> Given these difficulties, my suggestion would be to just
>> *not* reflect
>> > CC> "objective" anywhere in the taxonomic tree, but capture it solely via
>> > CC> the objectiveOf/hasObjective relation, along the lines of:
>> >
>> > CC> StateOfAffairs objectiveOf Action
>> >
>> > CC> Action hasObjective StateOfAffairs
>> >
>> > A StateOfAffairs not being an objective? I realise that there is not clean
>> > separation between a StateOfAffairs that is an Objective and one
>> which is not.
>> > The solution is to invent a word like "Fiat" and apply it to occurrents.
>> >
>> > Phil
>>
>
>
> >
>
Phillip Lord
Barry> Did the first 10 minutes of the 400th century BC exist, 42000(-ish)
Barry> years ago? Or did this fiat stretch of time come into existence only
Barry> when man created units called 'century' and 'second'? I tend to thing
Barry> that it existed already from the start. Same applies to all the fiat
Barry> parts of processes which we can, if we want, delineate.
Well, the example you gave is not a process. Let me ask you another question.
When you delineated the fiat part of that time span, did time fundamentally
change? Answer is no.
Likewise, when I choose to place an objective on a process, does that process
change? I don't think so. Hence, the process can have an objective which is
not intrinsic to that process. Hence, the analogy to roles still works, as
this appears to be your only objection.
>> So, we seem to have an answer. Objective would naturally be placed under
>> SpecificallyDependentContinuant.
Barry> we have been round this circle several times already; there are many
Barry> objectives (e.g. keep this water boiling) which do not fit under that
Barry> heading
On "this", I think that BFO only deals with Universals, so it would be "Keep
the water boiling".
If this is an objective to the process "heating water", then I can see no
reason why this should not be a SpecificallyDependentContinuant.
Phil
Matthew Pocock
> Christian's characterization of objectives as process boundaries is
> the first one I've heard that puts objectives outside the head, so I
> like that.
But why do we think that objectives exist outside of people's heads? I'd have
thought just the opposite.
Matthew
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> Barry> You can create as many fiat partitions of any given process as you
> Barry> like. BS
> >>
> >>
> >> Then your argument that a process has only intrinsic parts is wrong.
>
> Barry> Did the first 10 minutes of the 400th century BC exist, 42000(-ish)
> Barry> years ago? Or did this fiat stretch of time come into existence only
> Barry> when man created units called 'century' and 'second'? I
> tend to thing
> Barry> that it existed already from the start. Same applies to all the fiat
> Barry> parts of processes which we can, if we want, delineate.
>
>Well, the example you gave is not a process. Let me ask you another question.
>When you delineated the fiat part of that time span, did time fundamentally
>change? Answer is no.
>
>Likewise, when I choose to place an objective on a process, does that process
>change? I don't think so. Hence, the process can have an objective which is
>not intrinsic to that process. Hence, the analogy to roles still works, as
>this appears to be your only objection.
You went for an aimless stroll yesterday. And happened to reach the
end of a certain street
Today you say: my strolll yesterday had a certain objective, namely
reaching the end of that street
You are, I'm afraid, lying.
(Processes cannot be changed.)
BS
Matthew Pocock
> >Likewise, when I choose to place an objective on a process, does that
> > process change? I don't think so. Hence, the process can have an
> > objective which is not intrinsic to that process. Hence, the analogy to
> > roles still works, as this appears to be your only objection.
>
> You went for an aimless stroll yesterday. And happened to reach the
> end of a certain street
> Today you say: my strolll yesterday had a certain objective, namely
> reaching the end of that street
> You are, I'm afraid, lying.
> (Processes cannot be changed.)
> BS
This sounds compelling.
Now, is it possible for there to be an objective which is impossible to
obtain? I think so. I could have the objective to walk to the moon, or to be
the first person to sail to America, neither of which are achievable. While
these are obviously unachievable, I could, for example, be unaware that
unicorns are mythical, and hold the objective to capture one using a virgin
tied to a tree. Scientific investigations, by their very nature, will often
involve objectives that may with the wisdom of hindsight prove to fall into a
similar category to capturing unicorns, but they are no less objectives for
it.
Matthew
jennifer fostel
understand the workflow and experimental design for users of OBI. We
had toyed with the idea of using a "role" for process as a proxy for
objective, but this was in violation of OBO. Objective specification
is not in violation and serves the purpose we need.
However, objective does in fact only exist in the mind of the
investigator, and can be unrealistic (showing water can "remember"
even when diluted beyond the molecular limit, as published in the
80's). the idea of allowing objective specification into OBO/OBI
seems ideal -- the specification of an objective exists, even if the
objective is unrealistic.
Processes described by folks using OBI are by definition relalizable,
since they have in fact been realized when they are reported in OBI.
They are also not dependent on the objective -- the water was in fact
diluted, and the observations made and recorded, even though the basis
of the observations was the label on the slide rather than the view
under the microscope.
Phillip Lord
>> Likewise, when I choose to place an objective on a process, does that
>> process change? I don't think so. Hence, the process can have an
>> objective which is not intrinsic to that process. Hence, the analogy to
>> roles still works, as this appears to be your only objection.
Barry> You went for an aimless stroll yesterday. And happened to reach the
Barry> end of a certain street Today you say: my strolll yesterday had a
Barry> certain objective, namely reaching the end of that street You are,
Barry> I'm afraid, lying. (Processes cannot be changed.) BS
Well, BFO doesn't deal natively with statements about statements. I could say
"Phil is a philosopher" -- which is not true either, or "Barry hasMass 10kg"
which is not true.
You say that processes cannot be changed; fine, this is what I was arguing.
An objective is not an intrinsic part of a process. So the analogy to role
remains.
Phil
jennifer fostel
however, i understand that OBO does not permit the bearer of a role to
be anything other than an independent continuant.
so, we either change this rule or add "objective specification" for
"roles" of processes and dependent continuants such as a "variable
role"
personally i am more comfortable changing the rule, but i lost that
vote in OBI. possibly OBO can overrule this
we do need a solution however.
Barry Smith
Roles, like every other entity dealt with by BFO, exist. They are not
mere talk.
If something which played a role yesterday, that role existed yesterday.
If you walk yesterday had a certain objective yesterday, and if the
latter is to be viewed as some sort of role, then that role existed yesterday.
BS
>
Barry Smith
However, given our principle of incremental development (where we
test small steps to see if they work before we make bigger ones) I
strongly support the idea of adding 'objective specification' as a
term in OBI as a first step
BS
Phillip Lord
jf> you are right -- there is a strong analogy to role for objective.
jf> however, i understand that OBO does not permit the bearer of a role to
jf> be anything other than an independent continuant.
The documentation in the OWL class suggests that any continuant can bear a
role. Unfortunately, this is not represented in a machine interpretable form.
jf> so, we either change this rule or add "objective specification" for
jf> "roles" of processes and dependent continuants such as a "variable role"
jf> personally i am more comfortable changing the rule, but i lost that vote
jf> in OBI. possibly OBO can overrule this we do need a solution however.
Or you create a new concept called "role" which is a direct child of
SpecificallyDependentContinuant. This can be done within OBI and independent
of any changes to BFO.
Phil
Phillip Lord
Barry> You went for an aimless stroll yesterday. And happened to reach the
Barry> end of a certain street Today you say: my strolll yesterday had a
Barry> certain objective, namely reaching the end of that street You are,
Barry> I'm afraid, lying. (Processes cannot be changed.) BS
>>
>>
>> Well, BFO doesn't deal natively with statements about statements. I could
>> say "Phil is a philosopher" -- which is not true either, or "Barry
>> hasMass 10kg" which is not true.
>>
>> You say that processes cannot be changed; fine, this is what I was
>> arguing. An objective is not an intrinsic part of a process. So the
>> analogy to role remains.
>>
>> Phil
Barry> Roles, like every other entity dealt with by BFO, exist. They are not
Barry> mere talk. If something which played a role yesterday, that role
Barry> existed yesterday. If you walk yesterday had a certain objective
Barry> yesterday, and if the latter is to be viewed as some sort of role,
Barry> then that role existed yesterday. BS
As I am sure you are aware, however, BFO is fallabalist; so it is possible to
express something which is not true.
So, I am missing the point that you make which suggests that it's possible to
express objectives which are not true. Yes, it's possible. It's possible to do
this for any part of BFO.
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
> Barry> You went for an aimless stroll yesterday. And happened to reach the
> Barry> end of a certain street Today you say: my strolll yesterday had a
> Barry> certain objective, namely reaching the end of that street You are,
> Barry> I'm afraid, lying. (Processes cannot be changed.) BS
> >>
> >>
> >> Well, BFO doesn't deal natively with statements about
> statements. I could
> >> say "Phil is a philosopher" -- which is not true either, or "Barry
> >> hasMass 10kg" which is not true.
> >>
> >> You say that processes cannot be changed; fine, this is what I was
> >> arguing. An objective is not an intrinsic part of a process. So the
> >> analogy to role remains.
> >>
> >> Phil
>
> Barry> Roles, like every other entity dealt with by BFO, exist.
> They are not
> Barry> mere talk. If something which played a role yesterday, that role
> Barry> existed yesterday. If you walk yesterday had a certain objective
> Barry> yesterday, and if the latter is to be viewed as some sort of role,
> Barry> then that role existed yesterday. BS
>
>
>As I am sure you are aware, however, BFO is fallabalist; so it is possible to
>express something which is not true.
Yes. But if we know that something is false, then we should not
include a corresponding assertion in the structure of BFO.
Fallibilism does not mean that you treat truth and falsehood as
equals. It means that you strive to state the truth, knowing full
well that you will make mistakes which you will have to fix down the road.
>So, I am missing the point that you make which suggests that it's possible to
>express objectives which are not true. Yes, it's possible. It's possible to do
>this for any part of BFO.
Objectives are not in the business of being either true or false. The
issue concerns the fact that some objectives are never REALIZED.
Hence we cannot identify objective with: terminal boundary of a
process. Since a process can be associated with an objective, even
with an objective which, if realized would constitute a terminal
boundary, without ever having such a terminal boundary. Each: your
objective is to reach the top of a mountain but you get bored and go
back home before you are even half way up.
BS
Phillip Lord
>>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>> As I am sure you are aware, however, BFO is fallabalist; so it is
>> possible to express something which is not true.
Barry> Yes. But if we know that something is false, then we should not
Barry> include a corresponding assertion in the structure of BFO.
Barry> Fallibilism does not mean that you treat truth and falsehood as
Barry> equals. It means that you strive to state the truth, knowing full
Barry> well that you will make mistakes which you will have to fix down the
Barry> road.
>> So, I am missing the point that you make which suggests that it's
>> possible to express objectives which are not true. Yes, it's possible.
>> It's possible to do this for any part of BFO.
Barry> Objectives are not in the business of being either true or false. The
Barry> issue concerns the fact that some objectives are never REALIZED.
Barry> Hence we cannot identify objective with: terminal boundary of a
Barry> process. Since a process can be associated with an objective, even
Barry> with an objective which, if realized would constitute a terminal
Barry> boundary, without ever having such a terminal boundary. Each: your
Barry> objective is to reach the top of a mountain but you get bored and go
Barry> back home before you are even half way up. BS
The objective is in my head and maybe in other peoples heads. Likewise, your
role as a philosopher is in your head and perhaps in other peoples.
If you stop doing anything philosophical, then you will cease to be one. You
will loose the role. Likewise, if a process is not used for a particular
purpose any longer, it will lose the objective. The objective is realised at
the point that I, or you, imbue the process with this objective by using the
process for a particular purpose; likewise, you gain the role of being a
philosopher when you start using long words, or a computer scientist when you
start counting from zero.
As far as I can see, either there should be an objective and a role, or their
should be an objective specification and a role specification.
Phil
Barry Smith
Not so -- I could believe that I have (say) the professor role, and
others could believe that this is so (e.g. when I go on holiday), but
these believes would be false. They are true only when I have the
professor role; when an instance of this role exists and inheres in
me. Believing something doth not make it so; not even in the realm of roles.
>>If you stop doing anything philosophical, then you will cease to be one. You
>>will loose the role. Likewise, if a process is not used for a particular
>>purpose any longer, it will lose the objective.
We do not use processes. Uses ARE processes. We use independent continuants.
>>The objective is realised at
>>the point that I, or you, imbue the process with this objective by using the
>>process for a particular purpose;
imbuing is only the first step; realizing (if you're lucky) comes later
>>likewise, you gain the role of being a
>>philosopher when you start using long words, or a computer scientist when you
>>start counting from zero.
>>
>>As far as I can see, either there should be an objective and a role, or their
>>should be an objective specification and a role specification.
Maybe. But at the moment we have roles and we have objective
specifications; when we see the need for the other halves of these
pairs in the work of OBI we can readdress including them too.
Incremental, well-tested steps is our motto.
BS
>>Phil
>>
>>
>>
>>
>>
>>
lampucerka
>
>>> The objective is in my head and maybe in other peoples heads. Likewise, your
>>> role as a philosopher is in your head and perhaps in other peoples.
>>>
>
> Not so -- I could believe that I have (say) the professor role, and
> others could believe that this is so (e.g. when I go on holiday), but
> these believes would be false. They are true only when I have the
> professor role; when an instance of this role exists and inheres in
> me. Believing something doth not make it so; not even in the realm of roles.
>
>
>
them (or 'them', might you prefer) so.
depending on the shared view on what 'professor' means, local
regulations in the particular country, or whatever else is relevant, one
is a professor if one fulfills some particular criteria (e.g., has a
contract of a particular type). whether one believes one is a professor
is somewhat independent (though perhaps not in the statistical sense).
but that still does not provide any evidence for that roles are real (=
not mental constructs) entities. there is the one who's recognized as
a professor, there is his/her contract (and what a contract is might
certainly be discussed in equally inconclusive way as many other issues
here), and he/she is *considered* a professor (*attributed* the 'role')
by the virtue of having the contract -- but there may be no role object
inhering in a magical way in the person.
(it appears to me that whether there exist or not roles does not play
any (pardon me) role; we have no way to actually confirm or disconfirm
the putative existence of such weird entities (aka dependent
continuants), and it all rests on wishful speculations, why not focus on
proving the utility, in practical situations, of building ontologies as
if there were roles, rather than repeatedly appeal to void, even if
incidentally correct, claims.)
>>> If you stop doing anything philosophical, then you will cease to be one. You
>>> will loose the role. Likewise, if a process is not used for a particular
>>> purpose any longer, it will lose the objective.
>>>
>
> We do not use processes. Uses ARE processes. We use independent continuants.
>
>
hm, why could we not use processes? at least, it does not appear to me
that *speaking* of using a process is an obvious nonsense (yes, i know
we are doing ONTOLOGY, not linguistics). it could be argued the other
way round (we do not use continuants, we use processes), again with no
obvious way to actually check with the reality. if 'using processes'
makes the life of an ontological engineer, or his/her customers', more
or less difficult, then there is a ground for making a choice.
is it only independent continuants that can be used? can one use a role
(e.g., can a policeman use his role as a policeman to achieve some benefit)?
>>> The objective is realised at
>>> the point that I, or you, imbue the process with this objective by using the
>>> process for a particular purpose;
>>>
>
> imbuing is only the first step; realizing (if you're lucky) comes later
>
>
>>> likewise, you gain the role of being a
>>> philosopher when you start using long words, or a computer scientist when you
>>> start counting from zero.
>>>
>>> As far as I can see, either there should be an objective and a role, or their
>>> should be an objective specification and a role specification.
>>>
>
> Maybe. But at the moment we have roles and we have objective
> specifications; when we see the need for the other halves of these
> pairs in the work of OBI we can readdress including them too.
>
we have = we say we have, and carve the ontologies accordingly. we do
not really know what we have.
> Incremental, well-tested steps is our motto.
>
(...)
vQ
Barry Smith
>Barry Smith wrote:
> >
> >>> The objective is in my head and maybe in other peoples heads.
> Likewise, your
> >>> role as a philosopher is in your head and perhaps in other peoples.
> >>>
> >
> > Not so -- I could believe that I have (say) the professor role, and
> > others could believe that this is so (e.g. when I go on holiday), but
> > these believes would be false. They are true only when I have the
> > professor role; when an instance of this role exists and inheres in
> > me. Believing something doth not make it so; not even in the
> realm of roles.
> >
> >
> >
>believing that roles are anything beyond social constructs does not make
>them (or 'them', might you prefer) so.
>depending on the shared view on what 'professor' means, local
>regulations in the particular country, or whatever else is relevant, one
>is a professor if one fulfills some particular criteria (e.g., has a
>contract of a particular type). whether one believes one is a professor
>is somewhat independent (though perhaps not in the statistical sense).
>but that still does not provide any evidence for that roles are real (=
>not mental constructs) entities.
Clearly many real entities are mental constructs. Very many of the
entities represented by OBI are mental constructs -- they exist
because organisms endowed with mental powers do very complex things together.
Neither BFO nor OBI reduces 'real existence' to 'physical existence'
BS
> there is the one who's recognized as
>a professor, there is his/her contract (and what a contract is might
>certainly be discussed in equally inconclusive way as many other issues
>here), and he/she is *considered* a professor (*attributed* the 'role')
>by the virtue of having the contract -- but there may be no role object
>inhering in a magical way in the person.
>(it appears to me that whether there exist or not roles does not play
>any (pardon me) role; we have no way to actually confirm or disconfirm
>the putative existence of such weird entities (aka dependent
>continuants), and it all rests on wishful speculations, why not focus on
>proving the utility, in practical situations, of building ontologies as
>if there were roles, rather than repeatedly appeal to void, even if
>incidentally correct, claims.)
You are welcome to follow this strategy. But we have already embarked
on a quite complex project to prove the practical utility of another
(I believe simpler) strategy.
> >>> If you stop doing anything philosophical, then you will cease
> to be one. You
> >>> will loose the role. Likewise, if a process is not used for a particular
> >>> purpose any longer, it will lose the objective.
> >>>
> >
> > We do not use processes. Uses ARE processes. We use independent
> continuants.
> >
> >
>
>hm, why could we not use processes? at least, it does not appear to me
>that *speaking* of using a process is an obvious nonsense (yes, i know
>we are doing ONTOLOGY, not linguistics). it could be argued the other
>way round (we do not use continuants, we use processes), again with no
>obvious way to actually check with the reality. if 'using processes'
>makes the life of an ontological engineer, or his/her customers', more
>or less difficult, then there is a ground for making a choice.
'We can use X' makes sense only if 'we need not do so' also makes sense.
I can use my hammer, but I need not do so.
Can you give me an example of some process (some instance, occurring
at some place and time) where this holds?
>is it only independent continuants that can be used? can one use a role
>(e.g., can a policeman use his role as a policeman to achieve some benefit)?
>
Here the criterion I outline above does not give such a clear answer.
Notice that we can use something only if it exists (otherwise it is
just as if we are using something). Thus if we can use roles, then
this would seem to be an argument against your view that there are no
roles (but it is only as if there are roles) above.
> >>> The objective is realised at
> >>> the point that I, or you, imbue the process with this objective
> by using the
> >>> process for a particular purpose;
> >>>
> >
> > imbuing is only the first step; realizing (if you're lucky) comes later
> >
> >
> >>> likewise, you gain the role of being a
> >>> philosopher when you start using long words, or a computer
> scientist when you
> >>> start counting from zero.
> >>>
> >>> As far as I can see, either there should be an objective and a
> role, or their
> >>> should be an objective specification and a role specification.
> >>>
> >
> > Maybe. But at the moment we have roles and we have objective
> > specifications; when we see the need for the other halves of these
> > pairs in the work of OBI we can readdress including them too.
> >
>we have = we say we have, and carve the ontologies accordingly. we do
>not really know what we have.
In science we never have anything more than fallible hypotheses --
this is so whatever we do.
BS
lampucerka
>
>> believing that roles are anything beyond social constructs does not make
>> them (or 'them', might you prefer) so.
>> depending on the shared view on what 'professor' means, local
>> regulations in the particular country, or whatever else is relevant, one
>> is a professor if one fulfills some particular criteria (e.g., has a
>> contract of a particular type). whether one believes one is a professor
>> is somewhat independent (though perhaps not in the statistical sense).
>> but that still does not provide any evidence for that roles are real (=
>> not mental constructs) entities.
>>
>
> Clearly many real entities are mental constructs. Very many of the
> entities represented by OBI are mental constructs -- they exist
> because organisms endowed with mental powers do very complex things together.
> Neither BFO nor OBI reduces 'real existence' to 'physical existence'
>
ok, i meant non-mental. you seem to have objected to phil's view,
afaiu, that roles are just mental constructs, and that was what i was
challenging. your response here is correct but not to the point.
>
>> (it appears to me that whether there exist or not roles does not play
>> any (pardon me) role; we have no way to actually confirm or disconfirm
>> the putative existence of such weird entities (aka dependent
>> continuants), and it all rests on wishful speculations, why not focus on
>> proving the utility, in practical situations, of building ontologies as
>> if there were roles, rather than repeatedly appeal to void, even if
>> incidentally correct, claims.)
>>
>
> You are welcome to follow this strategy. But we have already embarked
> on a quite complex project to prove the practical utility of another
> (I believe simpler) strategy.
>
well, what's the experimental design intended to provide the proof? how
will you test your hypotheses? what are the controls? (are they not
considered proved a priori, as it seems?)
>
>> hm, why could we not use processes? at least, it does not appear to me
>> that *speaking* of using a process is an obvious nonsense (yes, i know
>> we are doing ONTOLOGY, not linguistics). it could be argued the other
>> way round (we do not use continuants, we use processes), again with no
>> obvious way to actually check with the reality. if 'using processes'
>> makes the life of an ontological engineer, or his/her customers', more
>> or less difficult, then there is a ground for making a choice.
>>
>
> 'We can use X' makes sense only if 'we need not do so' also makes sense.
> I can use my hammer, but I need not do so.
> Can you give me an example of some process (some instance, occurring
> at some place and time) where this holds?
>
>
i don't quite get the grounds for your criterion, but let it be. (you'd
better justify it, though, if you want your quite complex project have
symptoms of scientific work.)
i can surely use a chemical process to obtain some products, but if i do
not need the products, i need not do so.
>> is it only independent continuants that can be used? can one use a role
>> (e.g., can a policeman use his role as a policeman to achieve some benefit)?
>>
>>
>
> Here the criterion I outline above does not give such a clear answer.
> Notice that we can use something only if it exists (otherwise it is
> just as if we are using something). Thus if we can use roles, then
> this would seem to be an argument against your view that there are no
> roles (but it is only as if there are roles) above.
>
>
very smart. read my question as 'given that there are roles, ...?'
*you* do insist there are roles, and i did not seem to need to
overqualify the question. again, you've just made an attempt to avoid
the challenge.
>
>>> Maybe. But at the moment we have roles and we have objective
>>> specifications; when we see the need for the other halves of these
>>> pairs in the work of OBI we can readdress including them too.
>>>
>>>
>> we have = we say we have, and carve the ontologies accordingly. we do
>> not really know what we have.
>>
> In science we never have anything more than fallible hypotheses --
> this is so whatever we do.
>
sigh. in science, at least, we have methods that allow us to
cross-check with the reality in a convincing, repeatable way, applying
various techniques to test the statistical significance of the results.
bfo rests on no such grounds.
vQ
Barry Smith
***ideally, the experimental design would involve the advocates of an
alternative strategy committing to building a suite of reference
ontologies based on this strategy and applying them to the same kinds
of problems to which, e.g., the GO is applied; we could then compare
the results*** (this process of comparison is becoming easier
1: we will soon reach the point where we can compare the results of
applying different (early and late) versions of the GO for purposes
of gene set enrichment analysis and compare whether improvements made
in part under the inspiration of our preferred realist strategy do
indeed bring practical benefits.
2: HL7, caBIG, etc. are in a sense already realizing it in advance of the fact
> >
> >> hm, why could we not use processes? at least, it does not appear to me
> >> that *speaking* of using a process is an obvious nonsense (yes, i know
> >> we are doing ONTOLOGY, not linguistics). it could be argued the other
> >> way round (we do not use continuants, we use processes), again with no
> >> obvious way to actually check with the reality. if 'using processes'
> >> makes the life of an ontological engineer, or his/her customers', more
> >> or less difficult, then there is a ground for making a choice.
> >>
> >
> > 'We can use X' makes sense only if 'we need not do so' also makes sense.
> > I can use my hammer, but I need not do so.
> > Can you give me an example of some process (some instance, occurring
> > at some place and time) where this holds?
> >
> >
>
>i don't quite get the grounds for your criterion, but let it be. (you'd
>better justify it, though, if you want your quite complex project have
>symptoms of scientific work.)
>i can surely use a chemical process to obtain some products, but if i do
>not need the products, i need not do so.
no -- if you do not use that process (as you put it) then that
process would never have existed;
contrast: if you do not use that hammer, that hammer would still have
existed; 'using something' implies pre-existence of that something
and typically also post-existence.
> >> is it only independent continuants that can be used? can one use a role
> >> (e.g., can a policeman use his role as a policeman to achieve
> some benefit)?
> >>
> >>
> >
> > Here the criterion I outline above does not give such a clear answer.
> > Notice that we can use something only if it exists (otherwise it is
> > just as if we are using something). Thus if we can use roles, then
> > this would seem to be an argument against your view that there are no
> > roles (but it is only as if there are roles) above.
> >
> >
>
>very smart. read my question as 'given that there are roles, ...?'
>*you* do insist there are roles, and i did not seem to need to
>overqualify the question. again, you've just made an attempt to avoid
>the challenge.
would you like to restate the challenge?
> >
> >>> Maybe. But at the moment we have roles and we have objective
> >>> specifications; when we see the need for the other halves of these
> >>> pairs in the work of OBI we can readdress including them too.
> >>>
> >>>
> >> we have = we say we have, and carve the ontologies accordingly. we do
> >> not really know what we have.
> >>
> > In science we never have anything more than fallible hypotheses --
> > this is so whatever we do.
> >
>
>sigh. in science, at least, we have methods that allow us to
>cross-check with the reality in a convincing, repeatable way, applying
>various techniques to test the statistical significance of the results.
>bfo rests on no such grounds.
see *** above
BS
>vQ
>
>
lampucerka
>
>> well, what's the experimental design intended to provide the proof? how
>> will you test your hypotheses? what are the controls? (are they not
>> considered proved a priori, as it seems?)
>>
>
> ***ideally, the experimental design would involve the advocates of an
> alternative strategy committing to building a suite of reference
> ontologies based on this strategy and applying them to the same kinds
> of problems to which, e.g., the GO is applied; we could then compare
> the results*** (this process of comparison is becoming easier
> 1: we will soon reach the point where we can compare the results of
> applying different (early and late) versions of the GO for purposes
> of gene set enrichment analysis and compare whether improvements made
> in part under the inspiration of our preferred realist strategy do
> indeed bring practical benefits.
> 2: HL7, caBIG, etc. are in a sense already realizing it in advance of the fact
>
>
ideally, when i embark on a quite complex project, to make it scientific
i hope someone else will embark on an alternative quite complex project
which i a priori regard as silly.
>
>> i don't quite get the grounds for your criterion, but let it be. (you'd
>> better justify it, though, if you want your quite complex project have
>> symptoms of scientific work.)
>> i can surely use a chemical process to obtain some products, but if i do
>> not need the products, i need not do so.
>>
>
> no -- if you do not use that process (as you put it) then that
> process would never have existed;
> contrast: if you do not use that hammer, that hammer would still have
> existed; 'using something' implies pre-existence of that something
> and typically also post-existence.
>
>
>
i can start a fire, and then use it to warm up some food or not, and my
choice will not (i guess) change the fire's existence. oh my, i've got
a feeling that a fire is not a process.
>
>>> Here the criterion I outline above does not give such a clear answer.
>>> Notice that we can use something only if it exists (otherwise it is
>>> just as if we are using something). Thus if we can use roles, then
>>> this would seem to be an argument against your view that there are no
>>> roles (but it is only as if there are roles) above.
>>>
>>>
>>>
>> very smart. read my question as 'given that there are roles, ...?'
>> *you* do insist there are roles, and i did not seem to need to
>> overqualify the question. again, you've just made an attempt to avoid
>> the challenge.
>>
>
> would you like to restate the challenge?
>
>
>
no, thank you. i have poor experience in restating challenges on this list.
>
>>> In science we never have anything more than fallible hypotheses --
>>> this is so whatever we do.
>>>
>>>
>> sigh. in science, at least, we have methods that allow us to
>> cross-check with the reality in a convincing, repeatable way, applying
>> various techniques to test the statistical significance of the results.
>> bfo rests on no such grounds.
>>
>
> see *** above
>
yes, do it.
vQ
Barry Smith
>Barry Smith wrote:
> >
> >> well, what's the experimental design intended to provide the proof? how
> >> will you test your hypotheses? what are the controls? (are they not
> >> considered proved a priori, as it seems?)
> >>
> >
> > ***ideally, the experimental design would involve the advocates of an
> > alternative strategy committing to building a suite of reference
> > ontologies based on this strategy and applying them to the same kinds
> > of problems to which, e.g., the GO is applied; we could then compare
> > the results*** (this process of comparison is becoming easier
> > 1: we will soon reach the point where we can compare the results of
> > applying different (early and late) versions of the GO for purposes
> > of gene set enrichment analysis and compare whether improvements made
> > in part under the inspiration of our preferred realist strategy do
> > indeed bring practical benefits.
> > 2: HL7, caBIG, etc. are in a sense already realizing it in
> advance of the fact
> >
> >
>
>ideally, when i embark on a quite complex project, to make it scientific
>i hope someone else will embark on an alternative quite complex project
>which i a priori regard as silly.
When scientists embark on quite complex projects, and when other
scientists say: your project will not work because it rests on
presupposition p, then one surely quite reasonable response is: can
you make an analogous project work which does not rest on p. In the
field of ontologies HL7, for example, is already sort of doing this,
where p is the proposition to the effect that there is a distinction
between an activity and its documentation.
> >
> >> i don't quite get the grounds for your criterion, but let it be. (you'd
> >> better justify it, though, if you want your quite complex project have
> >> symptoms of scientific work.)
> >> i can surely use a chemical process to obtain some products, but if i do
> >> not need the products, i need not do so.
> >>
> >
> > no -- if you do not use that process (as you put it) then that
> > process would never have existed;
> > contrast: if you do not use that hammer, that hammer would still have
> > existed; 'using something' implies pre-existence of that something
> > and typically also post-existence.
> >
> >
> >
>
>i can start a fire, and then use it to warm up some food or not, and my
>choice will not (i guess) change the fire's existence. oh my, i've got
>a feeling that a fire is not a process.
a fire is indeed not a process
think of the Olympic flame ...
it continues to exist, self-identically, from quadrennium to
quadrennium ... like any good continuant
> >
> >>> Here the criterion I outline above does not give such a clear answer.
> >>> Notice that we can use something only if it exists (otherwise it is
> >>> just as if we are using something). Thus if we can use roles, then
> >>> this would seem to be an argument against your view that there are no
> >>> roles (but it is only as if there are roles) above.
> >>>
> >>>
> >>>
> >> very smart. read my question as 'given that there are roles, ...?'
> >> *you* do insist there are roles, and i did not seem to need to
> >> overqualify the question. again, you've just made an attempt to avoid
> >> the challenge.
> >>
> >
> > would you like to restate the challenge?
> >
> >
> >
>no, thank you. i have poor experience in restating challenges on this list.
good
BS
Phillip Lord
>> ideally, when i embark on a quite complex project, to make it scientific
>> i hope someone else will embark on an alternative quite complex project
>> which i a priori regard as silly.
Barry> When scientists embark on quite complex projects, and when other
Barry> scientists say: your project will not work because it rests on
Barry> presupposition p, then one surely quite reasonable response is: can
Barry> you make an analogous project work which does not rest on p.
This is not a scientific issue; it's an engineering one. Building an analogous
project with does not rest of p is no test. You need a controlled experiment.
There will be too many differences to draw conclusions.
My issue with realism is that there seems to be no clear criteria for
determining what "reality" means, other than "science is hard". Now, science
is about producing models that are (or could be) descriptive and/or predictive
of experimental data.
So, realism seems to reduce to the idea that our ontologies should represent
these models and not metamodel at any point. This would be fine, except that
to make a computable representation you need some kind of metamodel -- the
Isabelle representation of BFO, for example, contains such a metamodel
explicitly. The OWL and OBO representations take theirs from OWL and OBO
respectively.
Secondly, by not allowing metamodels it gets impossible to represent two
contradictory models; now as BFO is aimed as biomedicine, it's inability to
represent quantuum or relativistic physics is probably not an issue. But
colour might be. Likewise, it makes it hard to represent other modelling
techniques, where they do the job better. We need to be able to refer, for
example, to the mathematical models of physics.
This is doable with in BFO, but in this case it makes life more complicated
rather than simple -- mathematical models are not "reality" in this sense, so
that become generically dependent continuants. Making a statement that a body
has a velocity where velocity = $dr/dt$ becomes a type error.
All modelling is about abstraction and picking an appropriate level for this
abstraction; BFO has done this (correctly I think) by ignoring reality and
representing a Newtonian world. As a rule of thumb, "don't do metamodels" is a
good idea, but as a dogma it's wrong. I agree, Science is Hard; one size will
not fit all situations.
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> ideally, when i embark on a quite complex project, to make it scientific
> >> i hope someone else will embark on an alternative quite complex project
> >> which i a priori regard as silly.
>
> Barry> When scientists embark on quite complex projects, and when other
> Barry> scientists say: your project will not work because it rests on
> Barry> presupposition p, then one surely quite reasonable response is: can
> Barry> you make an analogous project work which does not rest on p.
>
>This is not a scientific issue; it's an engineering one. Building an analogous
>project with does not rest of p is no test. You need a controlled experiment.
>There will be too many differences to draw conclusions.
You are welcome to organize such a controlled experiment. In the mean
time we, who believe that p is crucial, have work to do.
>My issue with realism is that there seems to be no clear criteria for
>determining what "reality" means, other than "science is hard". Now, science
>is about producing models that are (or could be) descriptive and/or predictive
>of experimental data.
"reality" is like "truth" -- the only way you find out what reality
is like is to do science; that is what we are trying to do.
>So, realism seems to reduce to the idea that our ontologies should represent
>these models and not metamodel at any point.
Rather -- we hold that metamodels are rather secondary; indeed I
prefer not to use the term 'model' at all (because it suggests that
we are in the business of studying, not cells in reality, but little
red plastic models of cells in our labs -- I believe that we have
gone round this circle before, so please do not feel provoked).
Ontologies are not models of reality; they are (if they are any good)
representations of reality.
>This would be fine, except that
>to make a computable representation you need some kind of metamodel -- the
>Isabelle representation of BFO, for example, contains such a metamodel
>explicitly. The OWL and OBO representations take theirs from OWL and OBO
>respectively.
Before we can create a computer implementation of an ontology (which
is in turn a representation of some domain of reality) we should work
hard on building the ontology in such a way that it is good. What
could that mean? Perhaps: does justice to what biologists think? And
what could that mean? Perhaps: does justice to the way (as far as our
best current science goes) cells, etc., are in reality.
>Secondly, by not allowing metamodels it gets impossible to represent two
>contradictory models; now as BFO is aimed as biomedicine, it's inability to
>represent quantuum or relativistic physics is probably not an issue. But
>colour might be. Likewise, it makes it hard to represent other modelling
>techniques, where they do the job better. We need to be able to refer, for
>example, to the mathematical models of physics.
The idea is that we have a silo problem in biomedicine which arises
as a result of the fact that (in your terminology) computer
scientists like to create new metamodels, or models, or whatever you
call them. BFO and OBO Foundry are designed to put some constraint on
such proliferation. Already we have had several profitable
deliberations, some of them stretching over many months, which have
led to adjustments to BFO which have, I believe, made it stronger.
Reading between the lines, you are in the above proposing a major
adjustment to enable BFO to do justice to mathematics/physics.
Physical entities too belong to reality. Hence BFO should be able to
deal with them. I agree, in the long run. But we are still trying to
learn how to work. Let us make this thing work in biomedicine first.
>This is doable with in BFO, but in this case it makes life more complicated
>rather than simple -- mathematical models are not "reality" in this sense, so
>that become generically dependent continuants. Making a statement that a body
>has a velocity where velocity = $dr/dt$ becomes a type error.
>
>All modelling is about abstraction and picking an appropriate level for this
>abstraction; BFO has done this (correctly I think) by ignoring reality and
>representing a Newtonian world.
More specifically, we represent the quasi-Newtonian world as described e.g. in:
http://quantum.phys.cmu.edu/CQT/
This representation is designed to be compatible with the equations
of quantum mechanics. But that is a long long story.
>As a rule of thumb, "don't do metamodels" is a
>good idea, but as a dogma it's wrong. I agree, Science is Hard; one size will
>not fit all situations.
Of course not.
BS
>Phil
>
>
lampucerka
>
>> My issue with realism is that there seems to be no clear criteria for
>> determining what "reality" means, other than "science is hard". Now, science
>> is about producing models that are (or could be) descriptive and/or predictive
>> of experimental data.
>>
>
> "reality" is like "truth" -- the only way you find out what reality
> is like is to do science; that is what we are trying to do.
>
>
Barry,
on the pain of being boring (and bored), i feel an urge to repeat the
blasphemy that what you do is *not* science. with all respect, the obo
movement is hardly scientific; it does as much science as a kitchen
photographer does cooking. you wouldn't call constructing iso standards
'science', would you? (i know you disregard iso standards anyway.)
indeed, you do refer to scientific papers and use the results of
scientific investigations while building the ontologies, but that's not
enough for the claim. i of course realize that there is no such thing
as a commonly agreed definition of science, and that philosophers have
long had arguments about what constitutes (or should constitute) the
ultimate set of criteria. still, why not admit once and for all that
your current enterprise is engineering, if not just development, rather
than science. i thought a philosopher would be much more careful about
using the terms appropriately.
as Phil points out above, your often (mis)used argument in response to
various criticisms is that science is hard. i dare suggest that you put
an end to this.
"the only way you find out what reality is like is to do science; that
is what we are trying to do." -- are you suggesting that by building
ontologies you actually discover what reality is?
vQ
Barry Smith
in a scientific perspective. But I think it is unproblematic that the
work of creating and maintaining the SI system of units is a part of
science. Ontological work such as is being performed within the OBO
Foundry (mainly by biologists) is, I think, similarly a part of
science (and this is why we are working, e.g., to establish a
framework of expert peer review for ontologies).
It is perhaps understandable that engineers should fight so strongly
in favor of the view that ontologies are not part of science at all,
but rather part of engineering. The Gene Ontology is, certainly, an
unglamorous part of science. But it is a part of science nonetheless.
And while it is unglamorous, this does not mean that it is trivial.
Also, regarding
>"the only way you find out what reality is like is to do science; that
>is what we are trying to do." -- are you suggesting that by building
>ontologies you actually discover what reality is?
one fundamental point is that you should not build ontologies for
their own sake, but always only as a part of some serious scientific
attempt to understand reality. To this degree, building ontologies
can help you to discover what reality is.
BS
lampucerka
> I certainly have my worries about much of the work of ISO when viewed
> in a scientific perspective. But I think it is unproblematic that the
> work of creating and maintaining the SI system of units is a part of
> science. Ontological work such as is being performed within the OBO
> Foundry (mainly by biologists) is, I think, similarly a part of
> science (and this is why we are working, e.g., to establish a
> framework of expert peer review for ontologies).
>
i appreciate the work done by both you and the biologists you mention;
they certainly do science, but that science is not in building the
ontologies, but providing, in a scientific manner, the knowledge that
then becomes encoded in the ontologies.
> It is perhaps understandable that engineers should fight so strongly
> in favor of the view that ontologies are not part of science at all,
> but rather part of engineering.
don't care about engineers; the problem is, again and again, in that
the term 'science' is used here ad nauseam to explain away any doubts as
to your approach. unfortunately.
> The Gene Ontology is, certainly, an
> unglamorous part of science. But it is a part of science nonetheless.
> And while it is unglamorous, this does not mean that it is trivial.
>
it is not trivial. whether it is a part of science, is another issue.
it has been constructed for use by scientists, but its construction was
not scientific. read the paper by Ashburner et al., Nature Genetics
2000. it's not a research paper, it's just a commentary. it does not
even include the term 'science' in any form other than this one comment:
"Computer scientists have made significant contributions to linguistic
formalisms and computational tools for developing complex vocabulary
systems using reason-based structures, and we hope that our ontologies
will be useful in providing a well-developed data set for this community
to test their systems". likewise with 'research'; the only use of the
term is in "Ultimately, an ontology can be a vital tool enabling
researchers to turn data into knowledge." explicitly, go is here
regarded a service to science. in this sense it perhaps is 'part of
science', but this is a vague term, and not quite to the point.
your own (and colleagues) recent paper in nature biotech is again not a
research paper, it is a perspective, and there references to science and
research are of the same sort as above. you do say "Criticism, too, is
welcomed: the Foundry is an attempt to apply the scientific method to
the task of ontology development, and thus it accepts that no resource
will ever exist in a form that cannot be further improved", but very
recently, on this thread, you were asked to state what constitutes the
scientific approach of obo foundry, and your response was not
particularly convincing.
it is undeniable that the go has played a substantial role is research,
both in biology and in computer science, where it has been the source of
knowledge and a target of investigations, respectively. it does not
make the go any more scientific in itself, even though it has been
conceived by leading scientists.
> Also, regarding
>
>
>> "the only way you find out what reality is like is to do science; that
>> is what we are trying to do." -- are you suggesting that by building
>> ontologies you actually discover what reality is?
>>
>
> one fundamental point is that you should not build ontologies for
> their own sake, but always only as a part of some serious scientific
> attempt to understand reality. To this degree, building ontologies
> can help you to discover what reality is.
>
well, you can build ontologies for all sorts of purposes: for
understanding how people do that, for studying the properties of such
artifacts (which all can be considered 'serious attempts to understand
reality'), or just to enable communication between agents, which may be
very serious but have nothing to do with science. of course building
ontologies may help me to discover what reality is, but to make that
discovery i need to do substantially more than build ontologies.
there is no agreement on where the borders between the scientific and
the non-scientific lie, and i have no particular motivation for
insisting on my views here; let it be this or the other way. the point
is, even if your activities are indeed science, 'science is hard' is
hardly a scientific argument. people knew that already in the pre-obo era.
vQ
lampucerka
>
>> Barry> When scientists embark on quite complex projects, and when other
>> Barry> scientists say: your project will not work because it rests on
>> Barry> presupposition p, then one surely quite reasonable response is: can
>> Barry> you make an analogous project work which does not rest on p.
>>
>> This is not a scientific issue; it's an engineering one. Building an analogous
>> project with does not rest of p is no test. You need a controlled experiment.
>> There will be too many differences to draw conclusions.
>>
>
> You are welcome to organize such a controlled experiment. In the mean
> time we, who believe that p is crucial, have work to do.
>
>
there is undeniably a lot of work to do. note two issues, however.
$1: science makes all attempts to provide reliable, repeatable,
convincing proofs. you seem to prefer beliefs.
$2: a scientific activity is not a legal argument. it is not the
opponent's responsibility to prove you're guilty; it is your
responsibility to prove yourself. but you have just suggested, when
asked about the scientific foundation of your work, something somewhat
to the contrary.
vQ
Phillip Lord
>> This is not a scientific issue; it's an engineering one. Building an
>> analogous project with does not rest of p is no test. You need a
>> controlled experiment. There will be too many differences to draw
>> conclusions.
Barry> You are welcome to organize such a controlled experiment.
My point -- I don't think we can.
Barry> In the mean time we, who believe that p is crucial, have work to do.
There is no reason to partronise me, Barry. I am here because I would like OBI
to be useful for the purposes to which we are trying to put it.
>> My issue with realism is that there seems to be no clear criteria for
>> determining what "reality" means, other than "science is hard". Now,
>> science is about producing models that are (or could be) descriptive
>> and/or predictive of experimental data.
Barry> "reality" is like "truth" -- the only way you find out what reality
Barry> is like is to do science; that is what we are trying to do.
Yes, as I say, "reality" is just a less specific way of saying "descriptive or
predictive model of experimental data".
>> So, realism seems to reduce to the idea that our ontologies should
>> represent these models and not metamodel at any point.
Barry> Rather -- we hold that metamodels are rather secondary; indeed I
Barry> prefer not to use the term 'model' at all (because it suggests that
Barry> we are in the business of studying, not cells in reality, but little
Barry> red plastic models of cells in our labs -- I believe that we have
Barry> gone round this circle before, so please do not feel provoked).
Barry> Ontologies are not models of reality; they are (if they are any good)
Barry> representations of reality.
In some cases, yes, ontologies are representations of predictive models.
>> This would be fine, except that to make a computable representation you
>> need some kind of metamodel -- the Isabelle representation of BFO, for
>> example, contains such a metamodel explicitly. The OWL and OBO
>> representations take theirs from OWL and OBO respectively.
Barry> Before we can create a computer implementation of an ontology
Barry> (which is in turn a representation of some domain of reality) we
Barry> should work hard on building the ontology in such a way that it is
Barry> good.
I didn't say that. I said computable representation. A computable
representation is one where the semantics is defined enough to enable clear
inferences to be made in a reproducible way.
If an ontology is nothing more than a representation of reality -- that is a
predictive or descriptive model of experimental data, then any scientific
theory is an ontology.
Barry> What could that mean? Perhaps: does justice to what biologists think?
Barry> And what could that mean? Perhaps: does justice to the way (as far as
Barry> our best current science goes) cells, etc., are in reality.
It depends on the circumstances, as I have said. If we wish to describe the
physical properties of entities, in most cases physics has already done this
better, and using better representational techniques that the logical
representations of an ontology. Here it makes more sense to describe the
theories, rather than represent them in a semantically poorer form.
In general, though, we should represent the theory directly rather than
metamodelling. Even here, we have to chose our level of abstraction.
I think that "Function" is a useful abstraction for example. I also hope that
it's one we can get away from, as we gain the tools to build predictive models
that quantitatively represent the different physical properties and activities
of proteins, for example.
Where I find problems with "the way things are in reality" is that, apparently
the abstraction "function" is an entity that really exists, while the
abstraction "objective" is an entity which really doesn't. Hence, we describe
one directly in relation to an entity and the other as being something a
biologist thinks about an entity.
I find this rather confusing.
>> Secondly, by not allowing metamodels it gets impossible to represent two
>> contradictory models; now as BFO is aimed as biomedicine, it's inability
>> to represent quantuum or relativistic physics is probably not an issue.
>> But colour might be. Likewise, it makes it hard to represent other
>> modelling techniques, where they do the job better. We need to be able to
>> refer, for example, to the mathematical models of physics.
Barry> The idea is that we have a silo problem in biomedicine which arises
Barry> as a result of the fact that (in your terminology) computer
Barry> scientists like to create new metamodels, or models,
The two (model and metamodel) are different.
The silo problem arises because different people apply different abstractions.
Some of this can be solved by discussion. Some of it stems from the different
requirements for the models that are built.
Barry> or whatever you call them. BFO and OBO Foundry are designed to put
Barry> some constraint on such proliferation.
Reducing complexity can be good where is arises from unnecessary differences
in representation. Where is arises from differing requirements then it can be
more complex.
For example, BFO could just have used the metamodel of OBO or for OWL, rather
than full first order logic. Should we throw away the Isabelle representation,
to constrain this proliferation? 0
Barry> Reading between the lines, you are in the above proposing a major
Barry> adjustment to enable BFO to do justice to mathematics/physics.
Barry> Physical entities too belong to reality. Hence BFO should be able to
Barry> deal with them. I agree, in the long run. But we are still trying to
Barry> learn how to work. Let us make this thing work in biomedicine first.
Well, lots of biomedicine is about physical measurements -- you may remember
the discussion about capacitance.
In the end, I am suggesting is that we are not too doctrinaire; physical
measurements are one circumstance where modelling reality in our ontology is,
I think, not the best way forward. Perhaps this is even just a documentation
issue -- "by velocity we mean $dr/dt$". There are other circumstances also.
These circumstances should be considered on their merits, rather than the
merits of the philosophy of ontology building.
Phil
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> This is not a scientific issue; it's an engineering one. Building an
> >> analogous project with does not rest of p is no test. You need a
> >> controlled experiment. There will be too many differences to draw
> >> conclusions.
>
> Barry> You are welcome to organize such a controlled experiment.
>
>My point -- I don't think we can.
>
> Barry> In the mean time we, who believe that p is crucial, have work to do.
>
>There is no reason to partronise me, Barry. I am here because I would like OBI
>to be useful for the purposes to which we are trying to put it.
>
Sorry. I some times get carried away.
> >> My issue with realism is that there seems to be no clear criteria for
> >> determining what "reality" means, other than "science is hard". Now,
> >> science is about producing models that are (or could be) descriptive
> >> and/or predictive of experimental data.
>
> Barry> "reality" is like "truth" -- the only way you find out what reality
> Barry> is like is to do science; that is what we are trying to do.
>
>Yes, as I say, "reality" is just a less specific way of saying "descriptive or
>predictive model of experimental data".
I hope that next time you fly across the Atlantic the descriptive or
predictive model of experimental data you are flying in does not fall
into the water.
> >> So, realism seems to reduce to the idea that our ontologies should
> >> represent these models and not metamodel at any point.
>
> Barry> Rather -- we hold that metamodels are rather secondary; indeed I
> Barry> prefer not to use the term 'model' at all (because it suggests that
> Barry> we are in the business of studying, not cells in reality, but little
> Barry> red plastic models of cells in our labs -- I believe that we have
> Barry> gone round this circle before, so please do not feel provoked).
> Barry> Ontologies are not models of reality; they are (if they
> are any good)
> Barry> representations of reality.
>
>In some cases, yes, ontologies are representations of predictive models.
Sorry: I am really only concerned with ontologies built in the course
of scientific research.
Several points:
1. I think you find this rather confusing because your odd view of
reality (as a descriptive or predictive model of experimental data)
leads you astray -- thus it has the odd implication that there was no
reality before people started to do experiments
2. nowhere did I say that objectives lack reality; I did not take a
stand on this question; rather, I pointed out, trivially, that on
many occasions what people think of as their objectives are not
realized; and (slightly less trivially) that so many things can count
as (what people think of as their) objectives that the ontology will
be hard to do.
3. I do not like your talk of 'abstraction', since it implies that
there is something concrete from which functions, for example, are
abstracted. Hence that there were no functions before people came
along to take this particular abstraction step. I think dynosaur
hearts had functions, just as does your heart.
> >> Secondly, by not allowing metamodels it gets impossible to represent two
> >> contradictory models; now as BFO is aimed as biomedicine, it's inability
> >> to represent quantuum or relativistic physics is probably not an issue.
> >> But colour might be. Likewise, it makes it hard to represent other
> >> modelling techniques, where they do the job better. We need to
> be able to
> >> refer, for example, to the mathematical models of physics.
>
> Barry> The idea is that we have a silo problem in biomedicine which arises
> Barry> as a result of the fact that (in your terminology) computer
> Barry> scientists like to create new metamodels, or models,
>
>The two (model and metamodel) are different.
>
>The silo problem arises because different people apply different abstractions.
>Some of this can be solved by discussion. Some of it stems from the different
>requirements for the models that are built.
>
> Barry> or whatever you call them. BFO and OBO Foundry are designed to put
> Barry> some constraint on such proliferation.
>
>Reducing complexity can be good where is arises from unnecessary differences
>in representation. Where is arises from differing requirements then it can be
>more complex.
>
>For example, BFO could just have used the metamodel of OBO or for OWL, rather
>than full first order logic. Should we throw away the Isabelle representation,
>to constrain this proliferation? 0
I guess that, if the day is ever reached when we establish just what
is the correct logical idiom for doing good ontology work in support
of scientific research, then if I am still alive I will be duty bound
to attempt to persuade people to use only that logical idiom in
everything they do.
> Barry> Reading between the lines, you are in the above proposing a major
> Barry> adjustment to enable BFO to do justice to mathematics/physics.
> Barry> Physical entities too belong to reality. Hence BFO should be able to
> Barry> deal with them. I agree, in the long run. But we are still trying to
> Barry> learn how to work. Let us make this thing work in biomedicine first.
>
>Well, lots of biomedicine is about physical measurements -- you may remember
>the discussion about capacitance.
>
>In the end, I am suggesting is that we are not too doctrinaire; physical
>measurements are one circumstance where modelling reality in our ontology is,
>I think, not the best way forward. Perhaps this is even just a documentation
>issue -- "by velocity we mean $dr/dt$". There are other circumstances also.
>These circumstances should be considered on their merits, rather than the
>merits of the philosophy of ontology building.
I agree with this wholeheartedly.
BS
Phillip Lord
>> Yes, as I say, "reality" is just a less specific way of saying
>> "descriptive or predictive model of experimental data".
Barry> I hope that next time you fly across the Atlantic the descriptive or
Barry> predictive model of experimental data you are flying in does not fall
Barry> into the water.
I have flown across the Atlantic a number of times and many others than I know
have also. The prediction would be that this is unlikely to happen.
Barry> representations of reality.
>>
>> In some cases, yes, ontologies are representations of predictive models.
Barry> Sorry: I am really only concerned with ontologies built in the course
Barry> of scientific research.
I think that all science is building predictive (or descriptive) models. The
inverse isn't so true (all predictive models are science). As I said,
experimental data is required. In some cases, the subject of the experimental
data is important.
Barry> Several points:
Barry> 1. I think you find this rather confusing because your odd view of
Barry> reality (as a descriptive or predictive model of experimental data)
Barry> leads you astray -- thus it has the odd implication that there was no
Barry> reality before people started to do experiments
This is one of the reasons, I don't use the word "reality". If we consider the
implications of your statement:
Barry> "reality" is like "truth" -- the only way you find out what reality
Barry> is like is to do science
Then while reality may have existed before science, no one could possibly
understand what it was like. I suspect, however, that cavemen also understood
that water is wet.
Barry> 2. nowhere did I say that objectives lack reality; I did not take a
Barry> stand on this question; rather, I pointed out, trivially, that on
Barry> many occasions what people think of as their objectives are not
Barry> realized; and (slightly less trivially) that so many things can count
Barry> as (what people think of as their) objectives that the ontology will
Barry> be hard to do.
Agreed. But sitting on the fence has many of the same implications as saying
that objectives are not real.
Barry> 3. I do not like your talk of 'abstraction', since it implies that
Barry> there is something concrete from which functions, for example, are
Barry> abstracted. Hence that there were no functions before people came
Barry> along to take this particular abstraction step. I think dynosaur
Barry> hearts had functions, just as does your heart.
We've bound around this many times before. A heart has many physical
properties, many things that it does. It's "function" is the simplification
that we make, so that dealing with these many properties becomes tractable.
Likewise, we make the abstraction that I have a height, while experimentally
we know that my height depends on the observers motion and the local gravity.
Likewise, we make the abstraction of an organism and a species because they
work in general, although both have strong edge cases.
The point of science is not only to produce a predictive or descriptive model
of experimental data, but to produce one which is smaller than the original
experimental data. Otherwise, you haven't really achieved that much.
>> Reducing complexity can be good where is arises from unnecessary
>> differences in representation. Where is arises from differing
>> requirements then it can be more complex.
>>
>> For example, BFO could just have used the metamodel of OBO or for OWL,
>> rather than full first order logic. Should we throw away the Isabelle
>> representation, to constrain this proliferation? 0
Barry> I guess that, if the day is ever reached when we establish just what
Barry> is the correct logical idiom for doing good ontology work in support
Barry> of scientific research, then if I am still alive I will be duty bound
Barry> to attempt to persuade people to use only that logical idiom in
Barry> everything they do.
And here's the rub. We will never reach this point. First Order Logic is more
expressive. But it's computationally intractable and that's not going to
change. Selection of a formalism is always going to be a compromise. Different
people with different requirements will make different decisions. I'm afraid
it's one of those areas where "silos" are inevitable.
>> In the end, I am suggesting is that we are not too doctrinaire; physical
>> measurements are one circumstance where modelling reality in our ontology
>> is, I think, not the best way forward. Perhaps this is even just a
>> documentation issue -- "by velocity we mean $dr/dt$". There are other
>> circumstances also. These circumstances should be considered on their
>> merits, rather than the merits of the philosophy of ontology building.
Barry> I agree with this wholeheartedly.
Good. It's nice when we agree.
Phil
Phillip Lord
>> Yes, as I say, "reality" is just a less specific way of saying
>> "descriptive or predictive model of experimental data".
Barry> I hope that next time you fly across the Atlantic the descriptive or
Barry> predictive model of experimental data you are flying in does not fall
Barry> into the water.
I have flown across the Atlantic a number of times and many others than I know
have also. The prediction would be that this is unlikely to happen.
Barry> representations of reality.
>>
>> In some cases, yes, ontologies are representations of predictive models.
Barry> Sorry: I am really only concerned with ontologies built in the course
Barry> of scientific research.
I think that all science is building predictive (or descriptive) models. The
inverse isn't so true (all predictive models are science). As I said,
experimental data is required. In some cases, the subject of the experimental
data is important.
Barry> Several points:
Barry> 1. I think you find this rather confusing because your odd view of
Barry> reality (as a descriptive or predictive model of experimental data)
Barry> leads you astray -- thus it has the odd implication that there was no
Barry> reality before people started to do experiments
This is one of the reasons, I don't use the word "reality". If we consider the
implications of your statement:
Barry> "reality" is like "truth" -- the only way you find out what reality
Barry> is like is to do science
Then while reality may have existed before science, no one could possibly
>> Reducing complexity can be good where is arises from unnecessary
>> differences in representation. Where is arises from differing
>> requirements then it can be more complex.
>>
>> For example, BFO could just have used the metamodel of OBO or for OWL,
>> rather than full first order logic. Should we throw away the Isabelle
>> representation, to constrain this proliferation? 0
Barry> I guess that, if the day is ever reached when we establish just what
Barry> is the correct logical idiom for doing good ontology work in support
Barry> of scientific research, then if I am still alive I will be duty bound
Barry> to attempt to persuade people to use only that logical idiom in
Barry> everything they do.
And here's the rub. We will never reach this point. First Order Logic is more
expressive. But it's computationally intractable and that's not going to
change. Selection of a formalism is always going to be a compromise. Different
people with different requirements will make different decisions. I'm afraid
it's one of those areas where "silos" are inevitable.
>> In the end, I am suggesting is that we are not too doctrinaire; physical
>> measurements are one circumstance where modelling reality in our ontology
>> is, I think, not the best way forward. Perhaps this is even just a
>> documentation issue -- "by velocity we mean $dr/dt$". There are other
>> circumstances also. These circumstances should be considered on their
>> merits, rather than the merits of the philosophy of ontology building.
Barry> I agree with this wholeheartedly.
Barry Smith
> >>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:
>
> >> Yes, as I say, "reality" is just a less specific way of saying
> >> "descriptive or predictive model of experimental data".
>
> Barry> I hope that next time you fly across the Atlantic the descriptive or
> Barry> predictive model of experimental data you are flying in
> does not fall
> Barry> into the water.
>
>I have flown across the Atlantic a number of times and many others than I know
>have also. The prediction would be that this is unlikely to happen.
flew in a model, did you?
> Barry> representations of reality.
> >>
> >> In some cases, yes, ontologies are representations of predictive models.
>
> Barry> Sorry: I am really only concerned with ontologies built in
> the course
> Barry> of scientific research.
>
>I think that all science is building predictive (or descriptive) models. The
>inverse isn't so true (all predictive models are science). As I said,
>experimental data is required. In some cases, the subject of the experimental
>data is important.
so reality, then? or is it, for you, models all the way down
> Barry> Several points:
> Barry> 1. I think you find this rather confusing because your odd view of
> Barry> reality (as a descriptive or predictive model of experimental data)
> Barry> leads you astray -- thus it has the odd implication that
> there was no
> Barry> reality before people started to do experiments
>
>
>This is one of the reasons, I don't use the word "reality". If we consider the
>implications of your statement:
You provided me with a definition of "reality". This definition
caused, I think, certain problems. Now you want to claim you don't
use the word. Sounds like you're changing the rules in the middle of the game.
> Barry> "reality" is like "truth" -- the only way you find out what reality
> Barry> is like is to do science
>
>Then while reality may have existed before science, no one could possibly
>understand what it was like. I suspect, however, that cavemen also understood
>that water is wet.
Good. And that cavewomen are different from cavemen. That certain
mushrooms are edible. And many, many other things, many of predictive
value; some of life or death importance. I guess you will say that
they were doing science. That the mushrooms and cavewomen were mere models.
> Barry> 2. nowhere did I say that objectives lack reality; I did not take a
> Barry> stand on this question; rather, I pointed out, trivially, that on
> Barry> many occasions what people think of as their objectives are not
> Barry> realized; and (slightly less trivially) that so many
> things can count
> Barry> as (what people think of as their) objectives that the ontology will
> Barry> be hard to do.
>
>Agreed. But sitting on the fence has many of the same implications as saying
>that objectives are not real.
I hope not.
> Barry> 3. I do not like your talk of 'abstraction', since it implies that
> Barry> there is something concrete from which functions, for example, are
> Barry> abstracted. Hence that there were no functions before people came
> Barry> along to take this particular abstraction step. I think dynosaur
> Barry> hearts had functions, just as does your heart.
>
>We've bound around this many times before. A heart has many physical
>properties, many things that it does. It's "function" is the simplification
>that we make, so that dealing with these many properties becomes tractable.
I am beginning to suspect that for you everything is a simplification
(model) -- for me, functions are part of reality; they are not
simplifications; I am not interested in simplifications.
>Likewise, we make the abstraction that I have a height, while experimentally
>we know that my height depends on the observers motion and the local gravity.
Sigh.
>Likewise, we make the abstraction of an organism and a species because they
>work in general, although both have strong edge cases.
Your great great grandmother was an abstraction?
BS
Phillip Lord
Barry> At 05:17 PM 6/17/2008, you wrote:
>> I have flown across the Atlantic a number of times and many others than I
>> know have also. The prediction would be that this is unlikely to happen.
Barry> flew in a model, did you?
If you wish. It's easy to reduce the process of science down to the apparently
ridiculous.
Is the cat alive or dead?
Was your mother or father descended from a monkey?
Proof by disproof?
Ah, but it's only a theory, right?
Barry> representations of reality.
>> >>
>> >> In some cases, yes, ontologies are representations of predictive
>> >> models.
>>
Barry> Sorry: I am really only concerned with ontologies built in
>> the course
Barry> of scientific research.
>>
>> I think that all science is building predictive (or descriptive) models.
>> The inverse isn't so true (all predictive models are science). As I said,
>> experimental data is required. In some cases, the subject of the
>> experimental data is important.
Barry> so reality, then? or is it, for you, models all the way down
No. It stops at experimental data. We can argue whether experimental data, and
what I see and experience is real, but this just make us disappear up our
backsides.
Barry> Several points:
Barry> 1. I think you find this rather confusing because your odd view of
Barry> reality (as a descriptive or predictive model of experimental data)
Barry> leads you astray -- thus it has the odd implication that
>> there was no
Barry> reality before people started to do experiments
>>
>>
>> This is one of the reasons, I don't use the word "reality". If we
>> consider the implications of your statement:
Barry> You provided me with a definition of "reality". This definition
Barry> caused, I think, certain problems. Now you want to claim you don't
Barry> use the word. Sounds like you're changing the rules in the middle of
Barry> the game.
No, I asked you whether what you meant by "reality" was what I understood the
process of science to be. I have said repeatedly that I don't find "reality"
very useful because its fuzzy and lacks clearly defined criteria.
As below, your definition also produced problems.
Barry> "reality" is like "truth" -- the only way you find out what reality
Barry> is like is to do science
>>
>> Then while reality may have existed before science, no one could possibly
>> understand what it was like. I suspect, however, that cavemen also
>> understood that water is wet.
Barry> Good. And that cavewomen are different from cavemen. That certain
Barry> mushrooms are edible. And many, many other things, many of predictive
Barry> value; some of life or death importance. I guess you will say that
Barry> they were doing science. That the mushrooms and cavewomen were mere
Barry> models.
No. The extension of your statement that "the only way you find out what
reality is like is to do science" is: a) the caveman was doing science or b)
the caveman had no understanding of reality.
For everyday living, I have a rule of thumb, an idea about "reality". It's a
useful concept, like "common sense".
For science, however, I know that "common sense" and "reality" are much less
helpful. Is there really a species? Is there really a function? Does the
photon really go through two holes at once? The notion of "reality" gets
stretched very rapidly and moves away from common experience. I think that you
recognise this -- you have used the phrase "biological reality" yourself.
But we have experiments, we have experimental data. We try to build a model
which predict future experiments, or one which describes existing knowledge.
These models are scientific theories. That's it. Nothing else.
Barry> 3. I do not like your talk of 'abstraction', since it implies that
Barry> there is something concrete from which functions, for example, are
Barry> abstracted. Hence that there were no functions before people came
Barry> along to take this particular abstraction step. I think dynosaur
Barry> hearts had functions, just as does your heart.
>>
>> We've bound around this many times before. A heart has many physical
>> properties, many things that it does. It's "function" is the
>> simplification that we make, so that dealing with these many properties
>> becomes tractable.
Barry> I am beginning to suspect that for you everything is a simplification
Barry> (model) -- for me, functions are part of reality; they are not
Barry> simplifications; I am not interested in simplifications.
Most of science is. If we do not simplify, then we just have to represent the
primary data of all our experiments and we have achieved very little. I can
tell you the results of all my yeast crosses, or I can describe mendelian
genetics. The latter is a simplification of the primary results.
With functions, my hope is that eventually we can get rid of them, when we
have more precise, mathematically predictive models of the behaviour of
biological entities; at the moment, we have neither the maths, the compute nor
the data for this. Until that happens, function is a useful tool to describe
our experimental data.
Phil
waku
>
> >i can start a fire, and then use it to warm up some food or not, and my
>
> a fire is indeed not a process
> think of the Olympic flame ...
> it continues to exist, self-identically, from quadrennium to
> quadrennium ... like any good continuant
>
decade to decade ... like any good continuant. it was there some
thirty years ago, and it is here now, still the same life.
(siiiigh)
vQ