Re: Evidence

Dear Matthias,

I think your summary is very sensible. There is one practical issue 
arising from your final point.

My involvement in this discussion was prompted by two points - the use 
of reification to represent evidence, and the commitment to a 
probabilistic formalism.

IMHO, I would suggest that both of these represent premature commitment 
to a representation which is not really justified. Once you start 
exploring evidential reasoning, you rapidly see that reification would 
have to be repeated multiple times in order to capture even fairly basic 
ideas. Similarly, probabilistic formalisms are popular, but require some 
commitments that may not be sustainable in some areas (I'm thinking of 
conditional independence here).

At the same time, I'm keen for practical advances, and so would suggest 
some principled short-cuts, preferably in a fairly extensible form.

HTH,

Matt

> 
> * Evidence is indeed a very fuzzy term, and it might be applied to several things.
> 
> * There is a lot of theoretical and practical work behind 'evidence' (even a discipline called "evidence science"), and we should invest significant work into finding out what has already been done, and how it can be applied to our Semantic Web infrastructure.
> 
> * We want to 'keep the momentum going', and cannot wait until we have understood and agreed on everything about evidence. Hence, I think the best solution would be creating a generic OWL annotation property for representing evidence. It should be defined as something like 'the observation of the existence of A gives a rational observer a good reason to believe in the existence of B'. 'A' could be anything, e.g. an experiment, a XML file that describes the results of an experiment, or the paper a scientist wrote after interpreting the XML file. However, we should try to focus our descriptions on biological and physical reality where possible (so the default choice would be referring to the experiment, and not the XML file or the opinion of the scientist).
> 
> * In the view of BFO-friendly ontologies, there exists no thing that IS evidence. Instead, evidence is a ROLE that can be plaid by things in a certain context.
> 
> cheers,
> Matthias Samwald
> 
> ----------
> 
> Yale Center for Medical Informatics, New Haven /
> Section on Medical Expert and Knowledge-Based Systems, Vienna /
> http://neuroscientific.net
> 
> 
> 
> 
> .

-- 
http://acl.icnet.uk/~mw
http://adhominem.blogsome.com/
+44 (0)7834 899570

Received on Tuesday, 12 June 2007 18:33:18 UTC