realizations of objective specifications
Alan Ruttenberg
question is a voltage setting (or parameter) on some instrument.
There is information that specifies this - the information part is
easy, same as a measurement - a value and a units (label).
That information can be in a file or in the instrument, when the
setting is applied (setting some memory that the onboard computer can
later use).
There is the question of what the relationship between the generically
dependent setting (as found in a file full of other settings or in a
description of a protocol, and the instrument. I think this is a
concretization, but it's also the case that the gdc inheres in the
instrument too.
Q1: Does a GDC information entity inhere in anything that a
concretization of it inheres in?
The tricky part is what the setting is about.
If we were talking about the setting after the instrument had
accomplished what the setting implied (let's say 200 volts between two
parts), the we could say that the setting is_about the relational
quality that is voltage.
However looking forward, that quality might not ever exist. Therefore
I think that the setting is an objective specification. An objective
specification is (currently) information about a realizable entity. We
want to say what that realizable entity is - (almost) equivalently
what the process is.
The process is whatever process it is that instrument attempts (or
does?) achieve the voltage. In the case of the voltage, the
achievement will only be approximate (the usual issue around exact
numbers), in other cases it can be exactly accomplished.
Now we might say this is all to much detail, but the essential thing
to accomplish here is the relation of the setting to the quality. We
need to be able to have some path from a voltage setting to the
voltage quality, while avoiding the "concentration of 0 is not a
concentration" problem.
I've done a draft of what this might look like in
http://purl.obofoundry.org/obo/obi/repository/trunk/src/examples/development/voltage-setting.lisp
and
http://purl.obofoundry.org/obo/obi/repository/trunk/src/examples/development/voltage-setting.owl
I stated the voltage setting as being about a process which has a
relational quality (the voltage) that inheres in the instrument.
Attached is a picture of the class and property hierarchy. Since this
is an example, I've just put in enough of the top level to show it
rather than actually importing bfo/obi/iao.
Comment solicited, as usual.
-Alan
Alan Ruttenberg
> relational quality (the voltage) that inheres in the instrument.
Actually, about a realizable entity that is realized in a process which has a
relational quality (the voltage) that inheres in the instrument.
The OWL says it as I do in this message.
-Alan
Albert Goldfain
In an email to me, Alan had raised the issue of a (potentially) non-referring setting:
"If I specify a setting of a voltage that is beyond the capability of the instrument, then the setting lands up being about nothing?".
I had some thoughts about this. First of all the naive question: what is the harm in having some non-referring settings on a device? Can't we just say (for the purposes of an automated reasoner or someone using the ontology) that the setting only refers to a quality of the instrument when it falls in a certain range and fails to refer to anything if it falls out of this range? Barry would not call it a "setting" in this case, but a specification of a voltage level.
Supposing you have a good answer for that, here is an anecdote from my work on medical devices. Welch Allyn builds a blood pressure device with an onboard setting for ambient temperature range. If a user attempts to take a blood-pressure, the device first checks the ambient temperature against this setting. The result of the reading is captured in an information artifact called a status block. A status block can have optional "extended" device-generated information included in some readings. Here are a few possibilities:
(1) The BP reading succeeds AND the ambient temperature was within range. In this case the status block containing the BP is returned.
(2) The BP reading succeeds AND the ambient temperature was out of range (e.g., the doctor left the device in his car on a very cold day and it hasn't warmed up enough). In this case the status block containing the BP is returned AND an extended status block is included indicating that the ambient temperature was out of range. This is the device's way of tagging the data as potentially inaccurate (at WA they call this data pedigree...I believe Barry calls it data provenance).
(3) The device malfunctions and cannot return a BP because the ambient temperature is severely outside the acceptable limits (e.g., trying to use the BP device on Mount Everest). (Of course, there are also temperatures at which the blood pressure device will cease to be a blood pressure device (i.e., cease to instantiate the universal) and it will just melt...but these three cases are the more interesting ones).
Cases 1,2,3 suggest a series of ranges of values for a reading (i.e., measurement) of a quality. Going back to your example of voltage for some particular instrument consider the following (hypothetical) ranges
R1: 50 - 100 volts. The range in which the instrument will have a disposition to function properly AND accurately (with an error threshold determined by designers).
R2: 100-130 volts. The range in which the instrument will have a disposition to function properly BUT inaccurately
R3: >130 volts. The range in which the instrument will have a disposition to malfunction.
This presupposes that there is a canonical functioning of the instrument and that such a functioning is accompanied by a set of (quantitative) qualities that fall in a certain range. I think the best way to talk about a voltage setting in R2 and R3 (qua information artifact) that ends up being un-attainable by the device is to call out the relationship between the intended voltage setting (which cannot inhere in the instrument because of some other more fundamental qualities such as physical constraints of the design) and the disposition to function inaccurately or to malfunction. I think it is this relational information ABOUT the setting that is most useful (rather than resolving the setting's referent).