Re: PROV-ISSUE-337 (agent-and-entity): agent should not be a subclass of entity [prov-dm]

Hi Luc,

I'm happy with this definition:

"An agent is something that bears some form of responsibility for an
activity taking place or for the existence of an entity."

I think the question is whether we have the statement "An agent may be
a particular type of entity." This is true but we are discussing
whether it could be an activity possibly.

Thanks
Paul

On Mon, Apr 16, 2012 at 7:50 AM, Luc Moreau <L.Moreau@ecs.soton.ac.uk> wrote:
>
> Paul, Simon, Curt,
>
> This is the revised definition that is  now in the DM.
>
> An agent is something that bears some form of responsibility for an activity
> taking place. An agent may be a particular type of entity.
>
> This means that the model can be used to express provenance of the agents
> themselves.
>
>
> In addition, I would like also to bring the following amendment
>
> An agent is something that bears some form of responsibility for an activity
> taking place or for the existence of an entity. An agent may be a particular
> type of entity.
>
>
> This would make it clear why attribution links entity to agent.
>
> Thoughts, comments?
>
> Regards,
> Luc
>
>
> On 15/04/12 16:45, Paul Groth wrote:
>
> Hi Simon & Curt:
>
> Another way to say this is that we *do not* say anything about whether
> an agent is an entity or an activity but if you want to use the
> properties that have to do with entity or activity well you need to
> say that an agent is one or the other.
>
> In RDF-speak:
>
> - Agent is not disjoint from entity or activity
> - Agent is not in the domain or range of any of the properties of that
> associate activity and entity.
>
> I think this permissiveness is nice.  For the purposes of
> interoperability I currently don't see a case for constraining an
> agent to be an entity.
>
> regards
> Paul
>
> On Sun, Apr 15, 2012 at 5:13 PM, Miles, Simon <simon.miles@kcl.ac.uk> wrote:
>
>
> Hello Curt,
>
> I think it may be more than that (though I realise I didn't convey this
> well). Yes, saying that an action is responsible for something does imply
> that the action was performed by a hidden entity that was 'really'
> responsible. But it also says more: that it is that particular action that
> was responsible, and not anything else the entity did.
>
> With all agents being entities, I can say:
>
> touchFileX wasAssociatedWith UnknownPerson
> backupFileX wasInformedBy touchFileX
> and maybe also
> backupFileX wasAssociatedWith UnknownPerson
>
> But what this does not express is that UnknownPerson did touchFileX so that
> backupFileX would happen. UnknownPerson might have done many things and,
> while backupFileX was caused by touchFileX, that might be coincindental.
>
> If I can say:
>
> backupFileX wasAssociatedWith touchFileX
>
> then it's clear that touchFileX was responsible for backupFileX occurring,
> i.e. the intent was behind touchFileX.
>
> I share your wariness about activities being agents and I'm not completely
> convinced myself, even given my examples. However, I also think that by
> allowing agents to be activities we are not, as with some past debates,
> getting influenced by contrived corner use cases, but rather just being less
> restrictive on how people model things when there's no benefit to doing so.
>
> thanks,
> Simon
>
> Dr Simon Miles
> Senior Lecturer, Department of Informatics
> Kings College London, WC2R 2LS, UK
> +44 (0)20 7848 1166
>
> Mapping Dublin Core to the Open Provenance Model:
> http://eprints.dcs.kcl.ac.uk/1386/
> ________________________________________
> From: Curt Tilmes [Curt.Tilmes@nasa.gov]
> Sent: 15 April 2012 15:34
> To: public-prov-wg@w3.org
> Subject: Re: PROV-ISSUE-337 (agent-and-entity): agent should not be a
> subclass   of entity [prov-dm]
>
> On 04/15/2012 10:17 AM, Miles, Simon wrote:
>
>
> Curt:
>
>
> I'm suggesting that agents should be just entities and not activities.
> I'd like to see a good case where an activity is an agent.
>
>
> Maybe when the activity is an action with an intention behind it and
> we don't wish to model who held the intention and performed the
> action, just the action itself. Doing "touch file.x" was responsible
> for "file.x being backed up", not just a cause of it. "Saying 'shut
> the window'" was responsible for the activity of the window being
> shut. regardless of who said it. Modelling the activities as agents
> and using wasAssociatedWith allows the responsibility to be
> expressed and so blame to later be ascribed.
>
>
> So the activity performed or directed by a 'hidden agent' is a
> modelled as a proxy for that agent, in place of actually expressing
> that agent.
>
> You still aren't suggesting that the activity *is* the agent, just
> that we use it as an agent in place of one we don't know about yet.
>
> I guess that is ok, but I really don't see the problem with just
> making up a largely undescribed agent as a placeholder to describing
> that agent more fully in the future.
>
> Curt
>
>
>



-- 
--
Dr. Paul Groth (p.t.groth@vu.nl)
http://www.few.vu.nl/~pgroth/
Assistant Professor
Knowledge Representation & Reasoning Group
Artificial Intelligence Section
Department of Computer Science
VU University Amsterdam

Received on Monday, 16 April 2012 08:06:31 UTC