W3C home > Mailing lists > Public > public-prov-wg@w3.org > October 2011

Re: PROV-ISSUE-134: Non-Human Agent vs. Human Agent [Data Model]

From: Graham Klyne <GK@ninebynine.org>
Date: Tue, 25 Oct 2011 09:51:57 +0100
Message-ID: <4EA678AD.3070306@ninebynine.org>
To: Provenance Working Group WG <public-prov-wg@w3.org>
CC: Provenance Working Group Issue Tracker <sysbot+tracker@w3.org>
This viewpoint is possibly supported by some work mentioned in a recent paper 
from the #derive2011 workshop(?): 
http://www.eurecom.fr/~troncy/Publications/Troncy_Shaw-aswc09.pdf  (I don't have 
a formal citation for this yet, as the link came to me via a recent twitter 
conversation.)

A number of the event ontologies surveyed seem to make a similar distinction.

I think this whole area of event modelling is very relevant for provenance, as 
the more I look at the provenance model, the more it looks like an 
event-mediated structure for talking about the production of Entities.  I think 
the survey in the above paper is worth reading.

#g
--

On 24/10/2011 00:04, Provenance Working Group Issue Tracker wrote:
>
> PROV-ISSUE-134: Non-Human Agent vs. Human Agent [Data Model]
>
> http://www.w3.org/2011/prov/track/issues/134
>
> Raised by: Reza B'Far
> On product: Data Model
>
> I propose to revisit the previously discussed, but not concluded, topic of "Types" of Agents.  I had brought up this topic and the following was suggested as a reference -
>
> http://sourceforge.net/apps/mediawiki/trdf/nfs/project/t/tr/trdf/7/7a/ProvenanceVocabularyOverview.png
>
> There are a large set of use-cases (not just in my particular interest of Governance) where, whether the actions of an agent are directly controlled by a human being versus an automated mechanism makes a very significant difference in inferencing over the available instance data.  Examples:
>
> 1.  Human agent modifying a legal document versus the legal document being modified by a system agent that converts data formats.
> 2.  Human agent modifying a setting in a system whose provenance model is important for governing that system versus a system agent doing the same:  Example - Provenance of a "License" where Human agent expiring a license by changing/enforcing a date is quite a different event than a system agent changing/enforcing a date (say as a part of a mass/cascade update to a series of records) that causes expiration of a license.
>
> Other use-cases are available if need-be.  I actually claim that the number of such use-cases are increasing given the proliferation of pipe-and-filter architectures being deployed within Big Data infrastructures (where either pipes or filters can be Non-Human Agent/Actors).  Furthermore, as another evidence, there are other references to UML Use-Case and Sequence Diagrams where the distinction is becoming prevalent.
>
> As a solution, I suggest we take the same approach that the aforementioned URL above has taken.
>
>
>
>
Received on Tuesday, 25 October 2011 08:53:34 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:46 GMT