RE: [OWL-S] Negative effects/delete lists

Sheila and Bijan,
 What Sheila suggested for the release is along the lines of what I had in
mind.
 --Katia
 P.S. The meaning behind my stating, "since we are aiming for wide
applicability of OWL-S" is as follows: the wider OWL-S user community would
benefit from such warnings and advice since users may not have PhD's in
logic which may enable them to divine dangers of negative effects without
being warned about them explicitly.

-----Original Message-----
From: Sheila McIlraith [mailto:sheila@cs.toronto.edu] 
Sent: Sunday, March 28, 2004 10:23 PM
To: Bijan Parsia
Cc: Katia Sycara; public-sws-ig@w3.org
Subject: Re: [OWL-S] Negative effects/delete lists


Hi Bijan,

I wonder if we can write a tool to ensure that negative effects coding
rules are not violated?  I suspect it's doable, though computationally
intensive.

Regarding what to write in the release. I think a short warning
statement followed by details (problem statement, example, general
solution, and ontology writing tips) would be suitable.
Here's the skeleton.  Sorry it isn't filled in.  It's just
a straw proposal.


WARNING:  Negative effects of NEGATIVE EFFECTS
OWL-S enables the user to describe both the positive and
negative effects of a process.  The provision of negative
effects must be treated with care or it will lead to
counterintuitive and unintended entailments of your ontology.
For this reason, extreme care must be taken when encoding negative
effects. Below we provide an example to illustrate the problem, and
provide some coding guidelines.


Problem:
--------
<insert>


Example:
--------
<insert example>


General Solution:
-------------------
Do not have effects that are complementary to the entailments of
your KB. (Similarly, don't have entailments of your KB that are
complementsOf some effects.)


Ontology Writing Tips:
-----------------------
- if "X" is the effect of a process, ensure "X" is not use with
  complement-of, elsewhere in the ontology
- the transitive version of the above...




Sheila


On Sun, 28 Mar 2004, Bijan Parsia wrote:

>
> On Mar 28, 2004, at 7:25 PM, Katia Sycara wrote:
>
> > Bijan,
> >  Given that we aim for wide applicability of OWL-S, we should be
> > prescriptive and advisive.
> >  --Katia
>
> Hmm. That wasn't the line of reasoning I was expecting. And I tend to
> think of 'prescriptive' as some sort of contrary to 'advisive' (which
> is a word I made up :)).
>
> So, what sort of prescriptions and/or advice do you think we should
> give?
>
> Sample prescription:
> 	Do not have effects that are complementary to the entailments of
your
> KB. (Similarly, don't have entailments of your KB that are
> complementsOf some effects.)
>
> (Advising version:
> 	Effects which are the complement of some entailment of the KB may
> result in counterintuitive situations. [Insert example where an axiom
> entails, say, that you always have such and such a credit card, but
> some service has an effect that the card is canceled. This isn't
> *incoherent*, but it may be a bit surprising.])
>
> I guess it wasn't clear, but if we're not just letting chips fall as
> they may, I would like some feedback on what we should say.
>
> Cheers,
> Bijan Parsia.
>
>

Received on Monday, 29 March 2004 10:11:19 UTC