where do canonical things go?
Matthew Pocock
We seem to have reached a point where all (most?) of us agree that color holds
its meaning in relation to a person with a 'normal', tri-chromatic eye.
Similarly, anatomy is defined by 'normal' body parts, their relative sizes
and shapes. Also, the whole genomic sequence of organisms is defined
according to an 'average' over samplings from many phenotypically 'normal'
individuals.
Where do these cannonical versions of things live in BFO? They are not
universals but they are also not particulars. They are something rather
closer to prototypes (or dare I say it, Platonic ideals). In all of these
cases, we are saying what an idealised particular would look like. Since the
prototypical instance has been specifically constructed (it is not a
paragon), according to BFO's realist, existential stance, no such particular
exists. Since universals depend upon particulars for their existence, we
can't model prototypes as universals either.
As a concrete example, there is not, was not and never will be any person who
is the bearer of the cannonical human genome sequence. Even if by some
statistical fluke a person existed who had the same DNA sequence, it is not
*their* genome that was sequenced for the human genome sequencing effort. So,
that particular DNA sequence has no bearer. Hence, by a 'strong' reading of
BFO, that sequence can't be said to exist in reality.
So, where do we place knowledge about cannonical, prototypical things into the
BFO structure? We are finding that some definitions only hold in relation to
these prototypes. According to the principal that you define terms that your
definitions depend upon, we must be able to model them, and yet the current
BFO framework appears to me to rule them out of discussion.
Matthew
> Very interesting! And it strenghtens my belief that today's primordial
> color classification should be a prototypical (trichromatic-canonical)
> classfication; this would be no more odd than the fact that traditional
> anatomy is canonical anatomy.
>
> Ingvar
Ingvar Johansson
> Hi,
>
> We seem to have reached a point where all (most?) of us agree that color holds
> its meaning in relation to a person with a 'normal', tri-chromatic eye.
> Similarly, anatomy is defined by 'normal' body parts, their relative sizes
> and shapes. Also, the whole genomic sequence of organisms is defined
> according to an 'average' over samplings from many phenotypically 'normal'
> individuals.
>
It is a very god point you are making, and it is not limited to the life
sciences (canonical anatomy, canonical perception, and canonical
genome). In the social sciences you have *ideal types* such as "the
economic man", and in physics there are *idealizations* such as
"frictionless planes" and "absolutely elastic collisions".
> Where do these cannonical versions of things live in BFO? They are not
> universals but they are also not particulars.
Here it is necessary to be very careful with respect to terminology. An
Aristotelian universal (immanent realism) exists only in its
*spatiotemporal* instances (particulars), but it is *one and the same*
universal in all these instances, which means that the universal is an
*abstract particular*. The famous "one-in-many" expression can be
expanded in two equivalent ways: as "one-universal-in-many-instances" or
as "one-abstract-particular-in-many-spatiotemporal-particulars".
> They are something rather
> closer to prototypes (or dare I say it, Platonic ideals). In all of these
> cases, we are saying what an idealised particular would look like. Since the
> prototypical instance has been specifically constructed (it is not a
> paragon), according to BFO's realist, existential stance, no such particular
> exists. Since universals depend upon particulars for their existence, we
> can't model prototypes as universals either.
I think the way out is to follow Roman Ingarden's way of looking upon
"fictional objects". Since we can identify and re-identify fictional
objects in science (canonical entities, ideal types, and idealizations)
as well as in novels, cartoons, and theatre plays (Hamlet, etc.), such
objects simply have to be ascribed *some kind* of existence. They are
neither like material nor like Platonic entities. They are for their
existence dependent on "mental acts with directedness", i.e.,
intentional acts, but just as an *ordinary abstract particular* such as
(the universal) sphericity can exist at many points of time and be in
many different places simultaneously, a *fictional abstract particular*
can be in many different intentional acts that are spread out in space
and time. The fact that science contains fictional objects is by no
means an argument in favor of anti-realism and general social
constructivism; the fictional objects have been construed *because*
there are instances in the real spatiotemporal world that are (more or
less) similar to them.
best,
Ingvar
>
>
> As a concrete example, there is not, was not and never will be any person who
> is the bearer of the cannonical human genome sequence. Even if by some
> statistical fluke a person existed who had the same DNA sequence, it is not
> *their* genome that was sequenced for the human genome sequencing effort. So,
> that particular DNA sequence has no bearer. Hence, by a 'strong' reading of
> BFO, that sequence can't be said to exist in reality.
>
> So, where do we place knowledge about cannonical, prototypical things into the
> BFO structure? We are finding that some definitions only hold in relation to
> these prototypes. According to the principal that you define terms that your
> definitions depend upon, we must be able to model them, and yet the current
> BFO framework appears to me to rule them out of discussion.
>
> Matthew
>
>
>> Very interesting! And it strenghtens my belief that today's primordial
>> color classification should be a prototypical (trichromatic-canonical)
>> classfication; this would be no more odd than the fact that traditional
>> anatomy is canonical anatomy.
>>
>> Ingvar
>>
>
> >
--
Ingvar Johansson
IFOMIS, Saarland University
home site: http://ifomis.org/
personal home site:
http://hem.passagen.se/ijohansson/index.html
Matthew Pocock
On Thursday 24 May 2007, Ingvar Johansson wrote:
> Matthew Pocock schrieb:
> > Hi,
> > Where do these cannonical versions of things live in BFO? They are not
> > universals but they are also not particulars.
>
> Here it is necessary to be very careful with respect to terminology. An
> Aristotelian universal (immanent realism) exists only in its
> *spatiotemporal* instances (particulars), but it is *one and the same*
> universal in all these instances, which means that the universal is an
> *abstract particular*. The famous "one-in-many" expression can be
> expanded in two equivalent ways: as "one-universal-in-many-instances" or
> as "one-abstract-particular-in-many-spatiotemporal-particulars".
Hum ... OK, but from my understanding, BFO only
recognises "one-universal-in-many-instances", hence my question.
> > They are something rather
> > closer to prototypes (or dare I say it, Platonic ideals). In all of these
> > cases, we are saying what an idealised particular would look like. Since
> > the prototypical instance has been specifically constructed (it is not a
> > paragon), according to BFO's realist, existential stance, no such
> > particular exists. Since universals depend upon particulars for their
> > existence, we can't model prototypes as universals either.
>
> I think the way out is to follow Roman Ingarden's way of looking upon
> "fictional objects". Since we can identify and re-identify fictional
> objects in science (canonical entities, ideal types, and idealizations)
> as well as in novels, cartoons, and theatre plays (Hamlet, etc.), such
> objects simply have to be ascribed *some kind* of existence. They are
> neither like material nor like Platonic entities. They are for their
> existence dependent on "mental acts with directedness",
This sounds remarkably like "fictional objects" are cognitive constructs. I
personally have no problem with dividing things into what is there, and what
we think about what is there, but that is not the approach taken by BFO.
To come back to my original question though, is it possible to model
prototypes in BFO? If so, how? If not, is it something that BFO feels it
should be able to represent, and if so, what needs to be added to the current
model?
> best,
> Ingvar
Matthew
Ingvar Johansson
> Hi Ingvar,
>
> On Thursday 24 May 2007, Ingvar Johansson wrote:
>
>> Matthew Pocock schrieb:
>>
>>> Hi,
>>>
>
>
>>> Where do these cannonical versions of things live in BFO? They are not
>>> universals but they are also not particulars.
>>>
>> Here it is necessary to be very careful with respect to terminology. An
>> Aristotelian universal (immanent realism) exists only in its
>> *spatiotemporal* instances (particulars), but it is *one and the same*
>> universal in all these instances, which means that the universal is an
>> *abstract particular*. The famous "one-in-many" expression can be
>> expanded in two equivalent ways: as "one-universal-in-many-instances" or
>> as "one-abstract-particular-in-many-spatiotemporal-particulars".
>>
>
> Hum ... OK, but from my understanding, BFO only
> recognises "one-universal-in-many-instances", hence my question.
>
To me, being an immanent realist, this expression is synomymous to
"one-abstract-particular-in-many-spatiotemporal-particulars".
>
>>> They are something rather
>>> closer to prototypes (or dare I say it, Platonic ideals). In all of these
>>> cases, we are saying what an idealised particular would look like. Since
>>> the prototypical instance has been specifically constructed (it is not a
>>> paragon), according to BFO's realist, existential stance, no such
>>> particular exists. Since universals depend upon particulars for their
>>> existence, we can't model prototypes as universals either.
>>>
>> I think the way out is to follow Roman Ingarden's way of looking upon
>> "fictional objects". Since we can identify and re-identify fictional
>> objects in science (canonical entities, ideal types, and idealizations)
>> as well as in novels, cartoons, and theatre plays (Hamlet, etc.), such
>> objects simply have to be ascribed *some kind* of existence. They are
>> neither like material nor like Platonic entities. They are for their
>> existence dependent on "mental acts with directedness",
>>
>
> This sounds remarkably like "fictional objects" are cognitive constructs. I
> personally have no problem with dividing things into what is there, and what
> we think about what is there, but that is not the approach taken by BFO.
>
Here comes what I think is a general truth: top-top-level ontologies
will have to face some hard problems when they try to connect themselves
with low-level ontologies.
> To come back to my original question though, is it possible to model
> prototypes in BFO? If so, how? If not, is it something that BFO feels it
> should be able to represent, and if so, what needs to be added to the current
> model?
>
Since I take it for granted that Barry wants in some way or other to
connect BFO and the FMA, something has to be said about prototypes or
canoncial taxonomies/partonomies - either in a meta-BFO or directly in BFO.
Ingvar
>
>> best,
>> Ingvar
>>
>
> Matthew
Alan March
I am presently working on an ontology regarding medical procedures and a
paper by Boris Henning came to my attention ("Complex procedures
instantiated by their proper parts" in
http://www.iccl.tu-dresden.de/announce/CommonSense-2005/hennig2.pdf.
Henning proposes a simple process hierarchy which, at least as far as I can
see, would up to a point be compatible with BFO. Nevertheless, given the
fact that I am still unfamiliar with the practical use of BFO, I would
appreciate your opinions on the following:
- How would Henning's "telic" and "non-telic" processes fit into BFO? Or are
they implied somewhere in BFO and thus redundant? Or are they just plain
incompatible with the BFO model as a whole?
- Is Henning's "procedure" (defined as a complex telic process) necessary or
would it suffice to assume that a process could have other processes as its
parts and such a class is unneccesary?
I admit that I may probably getting the whole picture wrong out of ignorance
of BFO's "big picture" (or weltanschaung, if I may) but I thought this
group's the best place to seek for an answer and would be most grateful for
your advice.
Regards
Alan
Alan Ruttenberg
On May 31, 2007, at 12:23 PM, Alan March wrote:
> - How would Henning's "telic" and "non-telic" processes fit into
> BFO? Or are
> they implied somewhere in BFO and thus redundant? Or are they just
> plain
> incompatible with the BFO model as a whole?
A telic process would be the realization of a realizable entity.
> - Is Henning's "procedure" (defined as a complex telic process)
> necessary or
> would it suffice to assume that a process could have other
> processes as its
> parts and such a class is unneccesary?
I think basically, that procedures are the same as his telic
processes, and that he
lumps realizable entities, (e.g. function) in with their realizations
(e.g. processes).
Can elaborate if this is too telegraphic.
-Alan
Smith, Barry
BFO is divided into two types of ontologies (sometimes called SNAP,
for the continuants, and SPAN, for the occurrents). SNAP ontologies
are always indexed to a time (like snapshots); SPAN ontologies
embrace processes taking place within an interval of time (for
example the entire lifetime of the universe). From the SPAN
perspective we view processes timelessly (or, equivalently,
fourdimensionally). From the SPAN perspective, if a process is telic,
then it is telic from the start (or, preferably, atemporally); thus
the whole process type is already instantiated by the very first
phase. This seems to me not problematic from the BFO perspective. It
begins to appear problematic only if one applies to occurrents
expectations appropriate to continuants.
BS
William Bug
Here some random thoughts.BFO is divided into two types of ontologies (sometimes called SNAP,for the continuants, and SPAN, for the occurrents). SNAP ontologiesare always indexed to a time (like snapshots); SPAN ontologiesembrace processes taking place within an interval of time (forexample the entire lifetime of the universe). From the SPANperspective we view processes timelessly (or, equivalently,fourdimensionally). From the SPAN perspective, if a process is telic,then it is telic from the start (or, preferably, atemporally); thusthe whole process type is already instantiated by the very firstphase. This seems to me not problematic from the BFO perspective. Itbegins to appear problematic only if one applies to occurrentsexpectations appropriate to continuants.BS
Alan March
From: Kashyap, Vipul [mailto:VKAS...@PARTNERS.ORG]
Sent: Monday, June 04, 2007 12:07 PM
To: William Bug; bfo-d...@googlegroups.com
Cc: Alan March; Boris Hennig; Pierre Grenon; Kashyap, Vipul; michael....@boeing.com
Subject: RE: [bfo-discuss] Re: Questions regarding processesBill and Barry,
Thanks for the various clarifications around different types of processes.
http://www.aiai.ed.ac.uk/project/enterprise/enterprise/ontology.html
http://en.wikipedia.org/wiki/Process%28computing%29
IMHO, I would consider the definitions from above as important use cases that
could help clarify and validate the BFO. Clearly the kinds of processes
enumerated above are different from the kinds of processes identified (as of yet)
in biology.
At the same time, this may be one way of validating the BFO structure and maybe
refining the various notions of process etc.
Was wondering if there was some interest in articulating the definitions of a process
from the viewpoints of distributed computation, biology and AI (planning) and coming up
with a comprehensive framework for the same.
One interesting outcome of this could be a clear identification of points of difference
between a biological process and other types of processes (distributed computation, AI).
If there is interest in this, I could create a wiki page and get this going? What do you think Bill?
Michael: I have included you in this e-mail because you worked with Austin Tate on his Planning
Ontologies and think you could clarify for us the notion of a “Process” in the context of Barry’s work
for BFO.
Cheers,
---Vipul
Here some random thoughts.
BFO is divided into two types of ontologies (sometimes called SNAP,
for the continuants, and SPAN, for the occurrents). SNAP ontologies
are always indexed to a time (like snapshots); SPAN ontologies
embrace processes taking place within an interval of time (for
example the entire lifetime of the universe). From the SPAN
perspective we view processes timelessly (or, equivalently,
fourdimensionally). From the SPAN perspective, if a process is telic,
then it is telic from the start (or, preferably, atemporally); thus
the whole process type is already instantiated by the very first
phase. This seems to me not problematic from the BFO perspective. It
begins to appear problematic only if one applies to occurrents
expectations appropriate to continuants.
BS
I believe this was exactly the misunderstanding at the root of the discussion on the W3C SemWeb HCLS list last week re: process in the computational domain - i.e., and OS-level process. Such a "process" is really an independent continuant (not even a relalizable entity), contains the pointers to a piece of binary code and roles related to its history as a running process (e.g., "parent", "child", "orphan", "zombie", etc.). In fact, it really would be a non-sequitur to describe a bfo:process as "orphan" or "zombie".
At 12:33 AM 6/1/2007, Alan Ruttenberg wrote:
The information transmitted in this electronic communication is intended only for the person or entity to whom it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of or taking of any action in reliance upon this information by persons or entities other than the intended recipient is prohibited. If you received this information in error, please contact the Compliance HelpLine at 800-856-1983 and properly dispose of this information.
William Bug
Bill and Barry,
Thanks for the various clarifications around different types of processes.
http://www.aiai.ed.ac.uk/project/enterprise/enterprise/ontology.html
http://en.wikipedia.org/wiki/Process%28computing%29
IMHO, I would consider the definitions from above as important use cases that
could help clarify and validate the BFO. Clearly the kinds of processes
enumerated above are different from the kinds of processes identified (as of yet)
in biology.
At the same time, this may be one way of validating the BFO structure and maybe
refining the various notions of process etc.
Was wondering if there was some interest in articulating the definitions of a process
from the viewpoints of distributed computation, biology and AI (planning) and coming up
with a comprehensive framework for the same.
One interesting outcome of this could be a clear identification of points of difference
between a biological process and other types of processes (distributed computation, AI).
If there is interest in this, I could create a wiki page and get this going? What do you think Bill?
Michael: I have included you in this e-mail because you worked with Austin Tate on his Planning
Ontologies and think you could clarify for us the notion of a "Process" in the context of Barry's work
for BFO.
Cheers,
---Vipul
Here some random thoughts.
BFO is divided into two types of ontologies (sometimes called SNAP,
for the continuants, and SPAN, for the occurrents). SNAP ontologies
are always indexed to a time (like snapshots); SPAN ontologies
embrace processes taking place within an interval of time (for
example the entire lifetime of the universe). From the SPAN
perspective we view processes timelessly (or, equivalently,
fourdimensionally). From the SPAN perspective, if a process is telic,
then it is telic from the start (or, preferably, atemporally); thus
the whole process type is already instantiated by the very first
phase. This seems to me not problematic from the BFO perspective. It
begins to appear problematic only if one applies to occurrents
expectations appropriate to continuants.
BS
I believe this was exactly the misunderstanding at the root of the discussion on the W3C SemWeb HCLS list last week re: process in the computational domain - i.e., and OS-level process. Such a "process" is really an independent continuant (not even a relalizable entity), contains the pointers to a piece of binary code and roles related to its history as a running process (e.g., "parent", "child", "orphan", "zombie", etc.). In fact, it really would be a non-sequitur to describe a bfo:process as "orphan" or "zombie".
At 12:33 AM 6/1/2007, Alan Ruttenberg wrote:
The information transmitted in this electronic communication is intended only for the person or entity to whom it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of or taking of any action in reliance upon this information by persons or entities other than the intended recipient is prohibited. If you received this information in error, please contact the Compliance HelpLine at 800-856-1983 and properly dispose of this information.
William Bug
Matthew Pocock
> However, it is really as we extend beyond the definition to work
> through clear examples/use cases - IN OWL - that the true
> computational value of a foundational ontology such as BFO becomes
> more clear. For instance, one would (or I would) really want to have
> a graphical depiction of how the various independent and dependent
> continuants (e.g., CPU registers, program counters, execution &
> branch units, instructions, instruction sets, execution state, I/O
> lines, etc.) inter-relate to participate in a computational_process.
> Cheers,
> Bill
Hi,
I broadly agree with what say. I just wanted to make a point about processes
arrising from the execution of code vs some other types of process, and it
has to do with abstraction. We define computer processes as being executions
of computer programs, but this tells us vanishingly little about what is
being physically done, even if it provides almost full knowledge of what it
will calculate. Two executions of the exact same code may involve
unrecognisably different physical interactions between disparate
computational components but produce utterly valid results.
We can get the same result using...
big vs little endian arithmetic (not to mention all those in the middle)
arbitrary vs adequate precision
emulator vs running directly
computation then execution vs execution in interpreter
any chip with an appropriate instruction set
compilation to x86 or RISC
evaluation in hardware or on paper
Computation is defined purely on the information, and never on how it is
represented. CPUs, memory and so on are all about information carriers where
as functions, algorithms, objects are all about information representations.
This is closely related to some stuff I wrote a little while ago about
ontological interpretations of algorithms, available from this googledoc.
http://docs.google.com/Doc?id=dhkzc9hp_0tx33wc
Matthew
Smith, Barry
BFO tries to do justice to this distinction (between the execution
qua physical process, and the execution qua execution of an
algorithm) by means of its distinction between algorithm as
realizable entity and those processes in physical reality which are
the realizations thereof. A full account of executions requires
reference also to the relevant realizable entity. One advantage of
this is that it gives us a clear place to put algorithms within the
ontology. The one weak point in Matthew's document is its strange
definition of 'algorith':
algorithm: how to compute the outputs of a function given the
inputs. Usually expressed as a series of instructions, or as a series
of relations that hold at all stages of the computation.
which does not look like a definition to me at all.
BS
William Bug
Matthias Samwald
> This distinction Matthew is making is an important one, obviously -
> being able to characterize the "virtualized" computation process and
> map that to the specific hardware/OS/virtual machine implementation.
> As Barry mentions, BFO can make this link between realizable entity
> and physical substrate possible, but one does need to be very clear
> in the definitions one creates.
I think we have entered an important and quite fundamental discussion
that is not only confined to processes: information resources /
entities. With 'information resource' I mean an entity that is
commonly seen as independent ('abstracted') from any physical
manifestations. A simple example would be a text, or a database entry.
It is not clear to me at the moment how this would best be represented
in BFO. Can we even make statements about information resources in
BFO, or should we resort to solely describing their physical
manifestations (e.g. the books we can hold in our hands, the patterns
of magnetic code on our hard drives, the arrangements of pixels on our
computer screens)? I would actually like that idea, but I can see that
many users of the ontology would find that approach highly
unintuitive.
I have seen some ontologies where database entries or other
information entities were classified as bfo:objects or
bfo:continuants, and I wonder if that is correct from an ontological
viewpoint. An example would be the draft of the Clinical Trial
Ontology:
http://www.bioontology.org/wiki/index.php/Workshop_on_Clinical_Trial_Ontology
(at bottom of screen).
I guess this issue is also related to the recent discussion about
canonical entities.
cheers,
Matthias Samwald
Ingvar Johansson
>
> I think we have entered an important and quite fundamental discussion
> that is not only confined to processes: information resources /
> entities. With 'information resource' I mean an entity that is
> commonly seen as independent ('abstracted') from any physical
> manifestations. A simple example would be a text, or a database entry.
>
> It is not clear to me at the moment how this would best be represented
> in BFO. Can we even make statements about information resources in
> BFO, or should we resort to solely describing their physical
> manifestations (e.g. the books we can hold in our hands, the patterns
> of magnetic code on our hard drives, the arrangements of pixels on our
> computer screens)?
Here you will stumble on the problem of "emergent wholes". I have
written a paper discussing this with the relation between pixel patterns
and computer images as my main example. See:
http://triplec.uti.at/files/tripleC4(2)_Johansson.pdf
best,
Ingvar
> I would actually like that idea, but I can see that
> many users of the ontology would find that approach highly
> unintuitive.
> I have seen some ontologies where database entries or other
> information entities were classified as bfo:objects or
> bfo:continuants, and I wonder if that is correct from an ontological
> viewpoint. An example would be the draft of the Clinical Trial
> Ontology:
> http://www.bioontology.org/wiki/index.php/Workshop_on_Clinical_Trial_Ontology
> (at bottom of screen).
>
> I guess this issue is also related to the recent discussion about
> canonical entities.
>
>
> cheers,
> Matthias Samwald
>
>
> >
Smith, Barry
I think Barry has added me to this thread because of the comments
about my "Complex Procedures" paper. I hope you don't mind when I
post some thoughts about the definitions in the wiki page on processes.
1. Different stages of biological and clinical process can also be
executed by different participants, for instance a nurse and a
doctor, or a male and a female animal. So this should not be made a
defining feature of "computational process" in contrast to the
others two. It should be treated as a general option for all kinds
of processes.
2. It is important that "computational process" is defined as the
execution of a program, and something like this is missing in the
definition of "clinical care process." Not everything that is done
by clinical staff in the context of health care is also a clinical
care process (think of talking, breathing, humming, etc.). What
distinguishes clinical care processes from other kinds of thing that
nurses, doctors etc. do is that they have a point in the context of
health care, and are in some stricter sense part of health care.
I think it should be the general form of all those processes
which are realizations of realizable entities that they can involve
one or more participants.
The differences among such process will then lie in:
(a) The kind of realizable. Some realizables are specifiable in
detail by programs or algorithms, others are less strictly
determined by scripts, rules, or norms (such as human actions or
clinical procedures), and for some there may be no set of rules in
any strict sense, but only a pattern they typically conform to. This
distinction thus depends on the ontology of realizables.
(b) What the point is. This is how one can distinguish clinical
tasks from other tasks, e.g. processes like medical treatment from
other processes that may happen in the same context such as talking,
breathing, humming. The purpose of talking is (usually) not to cure anyone.
I have no idea how a good definition of "biological" process should
look like. The following is only an attempt:
"Biological Process: a realization of a realizable that is part of
the life of some living being."
This is far from perfect, since (1) life might also be a biological
process, which will make it circular - we would need an independent
account of what life is; and (2) not every process in the life of a
living being need be a biological process - we need to say more
about what it is to be "part of the life" of a living being.
Boris
Smith, Barry
>On Tuesday 05 June 2007, Smith, Barry wrote:
> > The one weak point in Matthew's document is its strange
> > definition of 'algorith':
> >
> > algorithm: how to compute the outputs of a function given the
> > inputs. Usually expressed as a series of instructions, or as a series
> > of relations that hold at all stages of the computation.
> >
> > which does not look like a definition to me at all.
> > BS
>
>definition? The content of this definition is exactly what I intend it to be
>but perhaps the presentation could be improved.
>
a definition should at the very least supply necessary and sufficient
conditions, so that everything which satisfies the right hand side
also satisfies the left hand side and vice versa. Your right hand
side is not formulated in such a way that one can check whether
anything satisfies it;
a good definition in addition takes the form:
an A is a B which Cs
where 'B' is a parent term for 'A' in an is-a hierarchy.
Compare: a man is a rational animal.
BS
>Matthew
Fabian Neuhaus
Hi Matthew,
Barry asked me to respond to your question. And since he is the 'decider' when it comes to BFO, I guess that I am now writing ex cathedra.
The short answer is: there are no canonical entities.
Strictly speaking "canonical" as adjective is nonsense, the right way to use "canonical" is as adverb that affects the modality of a whole statement. "Canonically" is from a logical point of view within the same syntactic category like "necessarily", "probably" or "it ought to be the case".
Consider, for example, the statements:
(1) There are no homeless people in Germany.
(2) It ought to be the case, that there are no homeless people in Germany.
Statement (1) is false, statement (2) is true. Even more the operator "it ought to be the case" changed the empirical statement (1) into a statement that cannot be evaluated empirically. (Well, if you believe that ethics is an empirical science you might object, but in any case the way to evaluate (2) is very different from the way to evaluate (1).)
Canonically behaves very similar. Statement (3) is false, statement (4) is true, and the way to evaluate (4) is very different from the way to evaluate (3).
(3) Every human has an appendix.
(4) Canonically, every human has an appendix.
It is rather hard to explain what the semantics of "canonically" is. It certainly can't explained statistically. At the Dagstuhl meeting in March there was a presentation where a default semantics approach was discussed, and this is closer to the truth, but there are counterexamples. As far as I can tell "canonically" means just "in accordance to the standard that has been developed by anatomists".
If you are interested you will find a discussion of the subject in the section 4 of the following paper.
http://www.acsu.buffalo.edu/~fneuhaus/relations_in_anatomical_ontologies_sa.pdf
Best
Fabian
> -----Ursprüngliche Nachricht-----
> Von: bfo-d...@googlegroups.com
> Gesendet: 24.05.07 10:30:59
> An: bfo-d...@googlegroups.com
> Betreff: [bfo-discuss] where do canonical things go?
>
>
> Hi,
>
> We seem to have reached a point where all (most?) of us agree that color holds
> its meaning in relation to a person with a 'normal', tri-chromatic eye.
> Similarly, anatomy is defined by 'normal' body parts, their relative sizes
> and shapes. Also, the whole genomic sequence of organisms is defined
> according to an 'average' over samplings from many phenotypically 'normal'
> individuals.
>
> Where do these cannonical versions of things live in BFO? They are not
> universals but they are also not particulars. They are something rather
> closer to prototypes (or dare I say it, Platonic ideals). In all of these
> cases, we are saying what an idealised particular would look like. Since the
> prototypical instance has been specifically constructed (it is not a
> paragon), according to BFO's realist, existential stance, no such particular
> exists. Since universals depend upon particulars for their existence, we
> can't model prototypes as universals either.
Matthew Pocock
> Hi Matthew,
>
> Barry asked me to respond to your question. And since he is the 'decider'
> when it comes to BFO, I guess that I am now writing ex cathedra.
>
> The short answer is: there are no canonical entities.
This is prety much the point I was making - canonicals do not exist in the
same sense as particulars do. You are folding a lot into the 'there are'
predicate here. You're making a very strong assumption that the only
existance relation that matters is that between universals and particulars.
If you claim (as you seem to) that universals exist, then it seems odd to
claim canonicals do not. The relation between unversals and particulars is
clearly different to that between canonicals and particulars, but I don't see
how the one can be elevated to the heights of the admissable while excluding
the other. Indeed, I can conceive of entirely usefull and expressive
ontological frameworks entirely built about canonicals and particulars to the
exclusion of universals. After all, this is prety close to what bayesian
belief networks do.
To take a concreate example from the paper that you have linked, the
anotomical concept of 'female pelvice' is distinct from the concept
of 'pelvices of females'. The first is a canonical, the second is a
universal. There is no particular female that contains in their body
the 'female pelvice' - it doesn't even make sense to say this. This is in
contrast to 'the pelvices of femails' which is instantiated in every case
that there is a particular femail who has a particular pelvice. The 'female
pelvice' and 'pelvices of femails' are clearly not the same type of thing -
one is intensionally defined, and the other extensionally, one is
instantiated by example the other by existance.
Obviously, we can botch this in the OWL representation - just create an OWL
class for 'female pelvice' and for 'pelvices of females', but while this has
utility, it voilates the mapping of the BFO to the OWL.
There is clearly a pressing need to be able to express this kind of
relationship between canonicals and particulars - it's difficult to see how
within the BFO framework, either anatomies or the human genome sequence can
be represented with out it. What I was trying to ask is, how do we represent
it? Is it a case of adding in a new universal-level entity, or is there some
other way to model this? If you have decided that we should not be
representing this, then how do we deal constructively with things that
clearly are canonical, other than ignore them and hoping that they go away?
Matthew
Fabian Neuhaus
On 7 Jun 2007, at 18:50, Matthew Pocock wrote:
> On Thursday 07 June 2007, Fabian Neuhaus wrote:
>> Hi Matthew,
>>
>> Barry asked me to respond to your question. And since he is the
>> 'decider'
>> when it comes to BFO, I guess that I am now writing ex cathedra.
>>
>> The short answer is: there are no canonical entities.
>
> This is prety much the point I was making - canonicals do not exist in
> the
> same sense as particulars do. You are folding a lot into the 'there
> are'
> predicate here.
Not really. It is the same existential quantifier as in "there are no
unicorns".
> You're making a very strong assumption that the only
> existance relation that matters is that between universals and
> particulars.
> If you claim (as you seem to) that universals exist, then it seems odd
> to
> claim canonicals do not. The relation between unversals and
> particulars is
> clearly different to that between canonicals and particulars, but I
> don't see
> how the one can be elevated to the heights of the admissable while
> excluding
> the other.
I don't see any the intrinsic link between universals and 'canonicals'
-- I am not even sure what 'canonicals' are supposed to be and what
theoretical role they are supposed to fulfill.
> Indeed, I can conceive of entirely usefull and expressive
> ontological frameworks entirely built about canonicals and particulars
> to the
> exclusion of universals. After all, this is prety close to what
> bayesian
> belief networks do.
Well, I can not only conceive, but I know very well-developed
ontological positions that get along fine without universals. To my
knowledge nobody ever claimed that BFO is the only conceivable
ontological theory.
> To take a concreate example from the paper that you have linked, the
> anotomical concept of 'female pelvice' is distinct from the concept
> of 'pelvices of females'.
> The first is a canonical, the second is a
> universal.
[just to clarify to people who did not read the paper: In the paper we
discuss the example 'female pelvis', but the distinction Matthew is
making here is not from the paper. According to the position that Barry
and I defend in the paper there are no canonical entities ]
> There is no particular female that contains in their body
> the 'female pelvice' - it doesn't even make sense to say this. This is
> in
> contrast to 'the pelvices of femails' which is instantiated in every
> case
> that there is a particular femail who has a particular pelvice. The
> 'female
> pelvice' and 'pelvices of femails' are clearly not the same type of
> thing -
> one is intensionally defined, and the other extensionally, one is
> instantiated by example the other by existance.
>
> Obviously, we can botch this in the OWL representation - just create
> an OWL
> class for 'female pelvice' and for 'pelvices of females', but while
> this has
> utility, it voilates the mapping of the BFO to the OWL.
You seem to assume that there are entities of the kind you call
"canonicals" in BFO. There aren't any. Thus I don't see how the
problem that you describe can occur. (By the way, if there was the
'canonical female pelvis' it would not be a universal, but a
particular. Thus it would not be represented as an OWL class.)
> There is clearly a pressing need to be able to express this kind of
> relationship between canonicals and particulars -
No, there isn't. There are no 'canonicals', thus there are no relations
between 'canonicals' and other particulars.
> it's difficult to see how
> within the BFO framework, either anatomies or the human genome
> sequence can
> be represented with out it.
> What I was trying to ask is, how do we represent
> it? Is it a case of adding in a new universal-level entity, or is
> there some
> other way to model this? If you have decided that we should not be
> representing this, then how do we deal constructively with things that
> clearly are canonical, other than ignore them and hoping that they go
> away?
As I sketched in my last email and as we discussed in the paper that I
referred you to: the right way to deal with canonical anatomy is to
distinguish canonical anatomy from what Cornelius Rosse calls "applied
anatomy". Applied anatomies contain empirical statements about all
hearts, livers, etc. Canonical anatomies don't. Every statement x
within a canonical anatomy has to be read "canonically, x holds" --
which means roughly: "according to the consensus between anatomists, x
is the norm." This norm is the result of scientific observations by
anatomists, considerations about what is "healthy", and last not but
least a result of the historical development of anatomy as medical
field. To have a norm like that is extremely useful because it gives
clinicians a reference point when they want to discuss abnormalities.
In any case I don't see a reason to invoke the help of mysterious
entities like 'canonicals'.
best
Fabian
Ingvar Johansson
>
> a definition should at the very least supply necessary and sufficient
> conditions, so that everything which satisfies the right hand side
> also satisfies the left hand side and vice versa. Your right hand
> side is not formulated in such a way that one can check whether
> anything satisfies it;
>
> a good definition in addition takes the form:
>
In my opinion, there is not just one single form that all good real
definitions fit into. All the definitions in the International System of
Units (the SI-system) are good definitions, but they do not fit the form
below. Here is one example: 'velocity =def. length divided by time'.
Ingvar
> an A is a B which Cs
>
> where 'B' is a parent term for 'A' in an is-a hierarchy.
> Compare: a man is a rational animal.
> BS
>
>
>
>> Matthew
>>
>
>
>
> >
Waclaw Kusnierczyk
> You seem to assume that there are entities of the kind you call
"canonicals" in BFO. There aren't any. Thus I don't see how the problem
that you describe can occur. (By the way, if there was the 'canonical
female pelvis' it would not be a universal, but a particular. Thus it
would not be represented as an OWL class.)
>
>
>> There is clearly a pressing need to be able to express this kind of
>> relationship between canonicals and particulars -
>
> No, there isn't. There are no 'canonicals', thus there are no
relations between 'canonicals' and other particulars.
1)
I haven't read the paper to the single word, but it does not appear to
me that you make a statement there to the effect that there are no
canonical entities. Yes, you say that canonical anatomies, as opposed
to applied anatomies, do not make empirical claims; but that's that.
Furthermore, you provide definitions that demand one to quantify over
anatomical structures (in the particular sense -- particular bodies
etc.) which are structured so as to fulfil the demands of a canonical
anatomy. Obviously, as you say, there are no canonical particulars, and
thus your quantifications are trivially void.
Let me give an example. You discuss statements such as 'A part of B'.
In Sec. 3, you provide a semantics for such statements which makes them
mean that every A is a part of a B (modulo time, I simplify for
readability).
You then discuss such statements as made in the context of canonical
anatomies; i.e., you reformulate 'A part of B' as 'canonically, A part
of B'. The semantics given in Sec. 4 makes such statements mean that
every A is a part of some B, where 'every' ranges over such As and Bs
that are parts of a canonical body -- a body such that is structured
according to the canonical anatomy in question.
Now, the meaning of statements of the form 'A relation B'. Under the
first interpretation, such statements are for the most part false --
because there usually are exceptions. 'Appendix is part of Body' is
false, because not all bodies have appendices, etc.
But under the other interpretation, they are only vacuously true, since
there are no canonical bodies. 'Canonically, Carpal bone is part of
Hand' is vacuously true, because (as you claim) there are no bodies
which comply with the canonical anatomy, so the range of 'for all' is empty.
2)
And why the claim that there are *no* canonical bodies? For what reason
there *must not* be a body that incidentally is exactly as described by
a canonical anatomy?
> As I sketched in my last email and as we discussed in the paper that
I referred you to: the right way to deal with canonical anatomy is to
distinguish canonical anatomy from what Cornelius Rosse calls "applied
anatomy". Applied anatomies contain empirical statements about all
hearts, livers, etc. Canonical anatomies don't.
3)
Would you please explain why you use the term 'applied anatomy', which,
as it appears, has not been used by Cornelius in any of the four papers
coauthored by him that you refer to in your article.
Furthermore, instantiated anatomies (as Cornelius calls them) contain
systematically collected data about those particulars that have happened
to be observed and evidenced -- not *all* hearts, livers, etc.
Instantiated anatomies -- contrarily to what you say -- do not contain
empirical statements about *all* anatomical structure particulars.
It could be argued that canonical anatomies -- contrarily to what you
say -- do: they are based (induced, not as you say, after Cornelius,
deduced) on evidence, and they say, as you note in your paper, how every
anatomical structure is supposed to be composed; and that most or even
all actual structures do not comply, does not make the scope of the
canonical anatomy any narrower). (I won't insist on this last argument.)
vQ
Fabian Neuhaus
I am sorry that I did not respond to you earlier.
(ad 1+2) There is a misunderstanding. When I wrote that "there are no canonicals" I was answering Mathew who believes that there are particulars, universals and some third kind of entity -- "canonicals". I rejected the idea that "canonicals" in this sense exist. However, I have no problem at all with the idea that some given anatomical structure has the property of being canonical (which just means that its structure meets the standard that anatomists created). I guess that it is an empirical fact that there is no canonical human, because it would be just very unlikely that there is one, but it would be certainly possible. But these canonical entities are just plain particulars, they don't belong into a different mysterious ontological category. (So the quantifiers are not vacuous, they range over anatomical structures.)
(3) I am sorry about confusing the terminology, you are correct it is not "applied anatomy" but "instantiated". I agree that most of the information represented in an instantiated anatomy ontology won't be universal in the sense that it applies to *all* entities of a given kind. That's because there are almost no rules without exception in anatomy. But I don't see why there should not be universals statements in an instantiated anatomy, e.g. "All hearts are organs" is true.
I am not sure whether I follow you, but I would agree that the scope of a canonical anatomy is not narrower than the one in instantiated anatomy. The quantifiers range over the same kind of entities. The major difference is that the truth conditions of statements within a canonical anatomy are evaluated radically different from the truth conditions within instantiated anatomy. In the latter case the truth of a statement depends purely on the relations between anatomical structures; in
the former case the consensus of the anatomical community needs to be taken into account.
Best
Fabian
> -----Ursprüngliche Nachricht-----
> Von: bfo-d...@googlegroups.com
> Gesendet: 09.06.07 02:59:45
> An: bfo-d...@googlegroups.com
> Betreff: [bfo-discuss] Re: where do canonical things go?
Waclaw Kusnierczyk
> Waclaw
>
> I am sorry that I did not respond to you earlier.
>
> (ad 1+2) There is a misunderstanding. When I wrote that "there are no canonicals" I was answering Mathew who believes that there are particulars, universals and some third kind of entity -- "canonicals". I rejected the idea that "canonicals" in this sense exist. However, I have no problem at all with the idea that some given anatomical structure has the property of being canonical (which just means that its structure meets the standard that anatomists created). I guess that it is an empirical fact that there is no canonical human, because it would be just very unlikely that there is one, but it would be certainly possible. But these canonical entities are just plain particulars, they don't belong into a different mysterious ontological category. (So the quantifiers are not vacuous, they range over anatomical structures.)
Sorry about the misunderstanding, I should not have read and answered
the post at 3 am. My fault.
> (3) I am sorry about confusing the terminology, you are correct it is not "applied anatomy" but "instantiated". I agree that most of the information represented in an instantiated anatomy ontology won't be universal in the sense that it applies to *all* entities of a given kind. That's because there are almost no rules without exception in anatomy. But I don't see why there should not be universals statements in an instantiated anatomy, e.g. "All hearts are organs" is true.
>
> I am not sure whether I follow you, but I would agree that the scope of a canonical anatomy is not narrower than the one in instantiated anatomy. The quantifiers range over the same kind of entities. The major difference is that the truth conditions of statements within a canonical anatomy are evaluated radically different from the truth conditions within instantiated anatomy. In the latter case the truth of a statement depends purely on the relations between anatomical structures; in
> the former case the consensus of the anatomical community needs to be taken into account.
I'll think about the rest and possibly come with further nagging
questions ;)
Thanks for the answer.
vQ
Phillip Lord
>>>>> "FN" == Fabian Neuhaus <fneu...@web.de> writes:
FN> (ad 1+2) There is a misunderstanding. When I wrote that "there
FN> are no canonicals" I was answering Mathew who believes that
FN> there are particulars, universals and some third kind of entity
FN> -- "canonicals". I rejected the idea that "canonicals" in this
FN> sense exist. However, I have no problem at all with the idea
FN> that some given anatomical structure has the property of being
FN> canonical (which just means that its structure meets the
FN> standard that anatomists created). I guess that it is an
FN> empirical fact that there is no canonical human, because it
FN> would be just very unlikely that there is one, but it would be
FN> certainly possible. But these canonical entities are just plain
FN> particulars, they don't belong into a different mysterious
FN> ontological category. (So the quantifiers are not vacuous, they
FN> range over anatomical structures.)
You seem not to understand the purpose of a canonical entity.
A canonical heart, for example, is an abstraction over all the hearts
that we consider to be not pathological and which represents the
features of these hearts.
It's is not a particular; the notion of a canonical heart is useful
regardless of whether you find a human who has a heart which is
exactly the same as your canonicalisation. Indeed, if you talk about a
canonical mammalian heart, then you will be unlikely to ever find
one.
The point is that a canonical heart is not one which meets the
standards that anatomists created; the canonical heart IS the standard
that anatomists create.
It's very similar to the notion of the normal human body temperature
which is about 37C. This is the normal human body temperature and does
not inhere in any of the particular human bodies with their particular
human temperatures.
Phil
Fabian Neuhaus
> -----Ursprüngliche Nachricht-----
> Von: bfo-d...@googlegroups.com
> Gesendet: 13.06.07 16:27:47
>
>
>
>
>
Phil,
I do not know whether you read my previous emails, but you seem to have a wrong impression of my position. I am completely in agreement that statements about the heart within canonical anatomy are CREATING a standard and are not an empirical description of all (or some) existing hearts. And I am aware that probably no existing human heart meets the standard that anatomists created which you refer to as 'the canonical heart' -- and nothing I wrote contradicts with this.
As a logician I try to make sense (= understand the truth conditions) of the statements within a canonical anatomy. I understand that it is easy to believe that the truth of
(1) "The canonical heart has four chambers"
seem to presuppose the existence of an entity called "the canonical heart". After all, the statement (1) is struturally similar to (2):
(2) "The Queen lives in England"
Statement (2) obviously presupposes the existence of the Queen. Thus -- one might think -- for the same reason (1) presupposes the existence of 'the canonical heart'.
If you adhere to this kind of reasoning, you are in good company. For example, Meinong's ontology includes square circles, since "The square circle is square" is true (according to him).
I am not neither a friend of "square circles" nor of "canonicals" or other mysterious entities. Thus I pointed out that the best way to understand statements like (1) is as a convenient shorthand for the statement (3):
(3) Canonically, the heart has four chambers.
(4) The heart has four chambers.
The truth-conditions of (3) don't involve a canonical heart, because in (3) "canonically" is a logical operator similar to "necessarily" or "possibly". The semantics of this operator are not entirely clear to me, but, roughly, (3) is true if and only if (4) is a part of the accepted standard for hearts among anatomists. And whether this is the case can be easily checked by looking at textbooks on anatomy.
If you really believe in some Meinongian heaven that includes the "canonical heart" -- that's fine, I guess. My position is more economical in the sense that I need less entities in my ontology. And it seems to me that the 'canonicals' do not have any explanatory value. But this leads to purely philosophical debates about methods of comparing competing theories. Thus, I happy to concede that a theory that involves 'canonicals' would work. But I do prefer my approach, and -- to come back to the original questions which spawned this discussion -- BFO also does not embrace the existence of "canonicals".
Best
Fabian
Phillip Lord
>> The point is that a canonical heart is not one which meets the
>> standards that anatomists created; the canonical heart IS the
>> standard that anatomists create.
>>
>> It's very similar to the notion of the normal human body
>> temperature which is about 37C. This is the normal human body
>> temperature and does not inhere in any of the particular human
>> bodies with their particular human temperatures.
>>
>>
FN> I do not know whether you read my previous emails, but you seem
FN> to have a wrong impression of my position.
Apologies if I missed this. I read most, but not all of the
"canonical" emails.
FN> I am not neither a friend of "square circles" nor of
FN> "canonicals" or other mysterious entities.
You have said that you understand what canonical is and your
explanation seems to agree with mine. I don't see why you describe it
as a "mysterious entity". It's seems that as straightforward as
"universal" for instance.
FN> If you really believe in some Meinongian heaven that includes
FN> the "canonical heart" -- that's fine, I guess.
Do I believe in this? Who knows. Do I think it would be useful for
doing science, then yes.
FN> My position is more economical in the sense that I need less
FN> entities in my ontology. And it seems to me that the
FN> 'canonicals' do not have any explanatory value.
And, yet, they have been widely used.
FN> But this leads to purely philosophical debates about methods of
FN> comparing competing theories.
Actually, trying to work out which things are going to most useful
seems very practical and pragmatic to me, rather than purely
philosophical.
FN> Thus, I happy to concede that a theory that involves
FN> 'canonicals' would work. But I do prefer my approach, and -- to
FN> come back to the original questions which spawned this
FN> discussion -- BFO also does not embrace the existence of
FN> "canonicals".
Okay, well this is a clear answer.
Phil
Waclaw Kusnierczyk
> FN> If you really believe in some Meinongian heaven that includes
> FN> the "canonical heart" -- that's fine, I guess.
>
> Do I believe in this? Who knows. Do I think it would be useful for
> doing science, then yes.
>
> FN> My position is more economical in the sense that I need less
> FN> entities in my ontology. And it seems to me that the
> FN> 'canonicals' do not have any explanatory value.
>
> And, yet, they have been widely used.
In taxonomy -- the classification of (living) organisms -- there is the
notion of type, which very roughly corresponds to canonicals. There, a
type is usually a specimen -- an individual that did exist. Usually --
depictions of specimens seem to be allowed as well. 'type' is not a
well-defined term, though. The following is a definition from the
glossary of ICZN:
"type, n.
A term used alone, or forming part of a compound term, to denote a
particular kind of specimen or taxon."
ICBN provides the following definition:
"7.2. A nomenclatural type (typus) is that element to which the name of
a taxon is permanently attached, whether as a correct name or as a
synonym. The nomenclatural type is not necessarily the most typical or
representative element of a taxon."
In general, both ICBN and ICZN are a good source of fun. Furthermore,
though taxa have real-world representatives, it is not clear how to
assess similarity so that a given individual comes out as more similar
to this or that type.
vQ
Alan Ruttenberg
...Thus I pointed out that the best way to understand statements like (1) is as a convenient shorthand for the statement (3):
(3) Canonically, the heart has four chambers.
(4) The heart has four chambers.
The truth-conditions of (3) don't involve a canonical heart, because in (3) "canonically" is a logical operator similar to "necessarily" or "possibly". The semantics of this operator are not entirely clear to me, but, roughly, (3) is true if and only if (4) is a part of the accepted standard for hearts among anatomists. And whether this is the case can be easily checked by looking at textbooks on anatomy.
Fabian Neuhaus
Alan,
A model of a heart is not an instance of a heart. Of course we can
learn a lot about the behavior of real hearts by studying -- for
example -- computer simulations of the heart, and this is obviously
only possibly because the behavior of the computer program corresponds
in some interesting way to the behavior of a heart. (By the way, the
simulation does not need to be of a 'canonical' heart, we can learn
something by studying simulations of diseased hearts, too.) But this
correspondence does not entail that the simulation is an instance of
Heart. Analogously, we can learn something about a house by studying
its blueprint, but this does not mean that the blueprint is an instance
of the type House.
I have not thought about the relation between a model and the universal
that it is representing. That's an interesting topic.
Fabian