Re: FOAF/DC Terms discussion summary and PROPOSED RESOLUTION

Hi, Phil.

> [snip]
> 
> Almost... it's the accountability thing that we need to get across which 
> leads me to suggest accountableAgent (bit long I know but it has the 
> desired feel). The essence of POWDER being that you can go back to 
> whoever is the identified author of the document and say "did you really 
> say that example.org is mobileOK? 'cos if you did I'll include it in my 
> search results, otherwise I'll assume it's not."
> 
> So the notion of SocialAgent indicating that the creative act was 
> performed by people cf. software is useful but, well, I might trust a 
> content analyser just as much/more than I trust a bunch of people - and 
> I'd still want to check that, if repeated, the analysis would come out 
> the same.

I wonder whether we need to define less ambiguously the notion of 
"accountability", at least as it should be used in POWDER. From my point 
of view, it should be separated from the one(s) of trustworthiness / 
reputation. A software agent generating POWDER documents can be 
considered as trustworthy by me and/or can have a good reputation in 
that the generated documents usually provide a "correct" description of 
the corresponding set of resources. But can it be considered as 
accountable / responsible for the claims in the generated POWDER docs? I 
don't think so.

IMHO, the relationship between a POWDER doc and the software agent which 
generated it might be modelled by a wdr:generator property, but the 
POWDER doc should always have a dcterms:creator / foaf:maker (meaning 
that the dcterms:creator / foaf:maker has created the POWDER doc using 
wdr:generator).

As far as AccountableAgent is concerned, I see it as a subclass of 
SocialAgent.

Andrea

Received on Wednesday, 9 July 2008 13:30:50 UTC