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

Re: PROV-ISSUE-4: Defining Agent using FOAF's definition

From: martin <martin@ics.forth.gr>
Date: Thu, 23 Jun 2011 18:59:26 +0300
Message-ID: <4E0362DE.5080001@ics.forth.gr>
To: public-prov-wg@w3.org
Jim,

I agree with research groups, organizations, corporations, etc.,

but in my opinion, software agents working on someone's behalf are a
nebulous concept. Since more than 15 years I see ontologies that include
machine under "Agent", but I have not seen any objective definition from
which degree on a machine would be regarded to "act" or "work" on someones behalf,
rather than mechanically "react". The scenario is very intense in science
fiction, but todays robots are far from having intensions of their own.

Legally, I can not sue my e-commerce robot for disobeying me. The Actor
who has set up the machine is the one responsible. In provenance, I would
not search a S/W agent to give me more information about his project.

Machines and S/W, to my understanding, are sufficiently described by being
used, run or executed as tools, whereas Actors would have initiative, responsibility,
rights, and are not "started", "executed" or "run".

Besides the intriguing philisophical questions about what "agency" may mean,
I see no relevant question for provenance that would be simplified by
not distinguishing people from their tools.

To my opinion "simple" model is not the one with the minimal number of concepts,
but the one, for which it is simple to say if an item is instance of a concept of the
model yes or no, and which have clear and distinct properties from each other.
Therefore I vote against machines or S/W being "Agents" in the
same sense as people.

Best,

Martin



On 6/21/2011 11:05 PM, Myers, Jim wrote:
> Martin,
> I would agree that trying to define intensity of agency is a black hole,
> beyond perhaps something that mirrors the dc:creator/dc:contributor type
> of split, i.e. if the group feels that controls vs. participates is a
> useful if still vague distinction, then OK.
>
> I do think that research groups, organizations, corporations, etc. are
> valid things though, as would be workflow engines/schedulers and
> software agents working on someone's behalf.
>
>   Jim
>
>> -----Original Message-----
>> From: public-prov-wg-request@w3.org [mailto:public-prov-wg-
>> request@w3.org] On Behalf Of martin
>> Sent: Tuesday, June 21, 2011 3:53 PM
>> To: public-prov-wg@w3.org
>> Subject: Re: PROV-ISSUE-4: Defining Agent using FOAF's definition
>>
>> Dear All,
>>
>> I fear the definition of "agent" as a role, in particular tied to
> causality, will lead us
>> to philosophical pitfalls.
>> There is no objective notion of causality, beyond the controlled
> environment of
>> logical machinery, and even in this environment it is not
> straightforward. Social
>> causality is known to be subjective and view-dependent. In physics,
> there is no
>> more a notion of strict causality.
>>
>> If we want to make the model simple, we should not speculate about the
>> intensity of "agency", but state if a human being or a social human
> group is
>> directly present, involved in or legally responsible for a process.
> Then the modes
>> of presence, involvement or "agency" can be defined by subproperties.
>> Therefore, in ISO21127, we preferred the term "Actor" over "Agent".
>>
>> A possible construct would be to define an "Agent" as an n-ary
> relationship (an
>> RDF class) between the actual persons and the process in question.
> Such models
>> however introduce an ugly indirection, which satisfies more
> philosophical views
>> than practical needs. Ultimately, this is however the need if "Agent"
> is seen as a
>> process-dependent role, because otherwise the real persons behind
> cannot be
>> identified, and this is what we need for provenance, much more than
> details of
>> their ways of acting. This problem was discussed between the working
> groups of
>> CIDOC CRM and ABC Harmony
>>
> (http://tc.eserver.org/authors/Doerr,_Martin,_Jane_Hunter_and_Carl_Lagoz
> e)
>>
>> In the course of finding very generic classes, one is easily lost in
> philosophical
>> problems. It is not only important to create as few concepts as
> possible, but also
>> concepts of things and facts that can easily be veryfied and
> identified. Presence
>> of people or legal responsibility is normally well-defined in the
> processes we are
>> interested in.
>>
>> The European library practice registers "authors" in the sense of
> "Actor"
>> presented above, whereas the American tradition used to register
> "authors"
>> relative to the creation of a book. Recently, the American practice
> was found
>> highly problematic for information integration, and slowly the
> European is
>> internationally adopted. We should not make the same mistakes.
>>
>> We should not be misled by the physicality of a human body that a
> social group,
>> such as a research team, may constitute one Actor or Agent, but not
> easily be
>> identified as a "Thing".
>>
>>
>> Best,
>>
>> Martin
>>
>> On 6/21/2011 9:34 PM, Myers, Jim wrote:
>>> Stephan,
>>>
>>> I can see the argument that an agent could be seen as a view, but it
>>> doesn't sound as compelling to me as student or employee as views -
>>> those view would have additional properties (student ID, major,
>>> salary, etc.) and such views have some longevity. In contrast, I'm
> not
>>> sure what new properties me as an agent would have and it seems more
> like
>> I'd be creating the agent-view-of-me just for PIL (single use to have
> something
>> to control a process). If we have me, and a student view of me, do we
> also want
>> a student-agent-view of me to take a test? I raise this example in
> part to show
>> that if agency is a role, one could still have views such as
> student-view-of-me
>> that could play that role.
>>>
>>> I guess I should also ask if agents are IVPs of other things - would
>>> they still be a special type or are they just things too, as other
>>> IVPs currently are? If they are not special types, I think this
> would
>>> become a model where things can control processes and a
> suggested/best
>> practice might be to create an IVP in cases where there's a need to
> distinguish
>> some characteristics of a thing-as-agent that are different than the
> thing itself. Is
>> that the intent?
>>
>>>
>>> Jim
>>>
>>> *From:*Stephan Zednik [mailto:zednis@rpi.edu]
>>> *Sent:* Tuesday, June 21, 2011 1:59 PM
>>> *To:* Satya Sahoo
>>> *Cc:* Myers, Jim; Graham Klyne; Luc Moreau; public-prov-wg@w3.org
>>> *Subject:* Re: PROV-ISSUE-4: Defining Agent using FOAF's definition
>>>
>>> Hi all,
>>>
>>> I add my agreement to the statement that it is important to have a
> way to
>> describe the provenance of an agent.
>>>
>>> Now for some random thoughts:
>>>
>>> I would like to amend Satya's last definition of agent to:
>>>
>>> "Athing that is actively causally involved in a process execution is
> an Agent"
>>>
>>> What is interesting about this definition is that it ties the
>>> agent-ness of a thing to the duration of the process execution. I
>>> think you could say the same about the prior definition that omitted
> 'actively'.
>> My interest in adding 'actively' is to differentiate things that
> participate by their
>> own power to inert factors that influence the process results.
>>>
>>> Back to the definition of agent.
>>>
>>> Perhaps agent status is an IVP of a thing?
>>>
>>> For a quick analogy; I think a thing is an agent like a student is a
>>> person. I would not argue that an agent is a role or non-thing
> concept
>>> anymore than I would argue that a student is a role or a non-person
> concept.
>> Or to put it a better way, I think it would be just as odd to argue
> that an agent is
>> a role a thing takes as it would be to argue that a student is a role
> a person takes.
>>>
>>> --Stephan
>>>
>>> On Jun 21, 2011, at 9:21 AM, Satya Sahoo wrote:
>>>
>>>
>>>
>>> Hi Jim,
>>>
>>>   >  However agents are modeled, I think it is important to have a way
>>> to describe their provenance
>>>
>>> I agree. For example, the manufacturer or place and date of
>>> manufacture of a sensor (acting as an agent in a sensor network) are
> relevant
>> provenance information.
>>>
>>> Best,
>>>
>>> Satya
>>>
>>> On Tue, Jun 21, 2011 at 11:14 AM, Myers, Jim<MYERSJ4@rpi.edu
>> <mailto:MYERSJ4@rpi.edu>>  wrote:
>>>
>>> We debated quite a bit for OPM and ended up making agent a separate
>>> 'class' because agents appeared to blend being a thing and acting
> like
>>> a process, along with the challenge that artifacts were immutable
> and
>>> agents were not. Given mutable things, and potential interest/use
> cases where
>> the provenance of agents is of interest, making people and
> organizations
>> PIL:things that have an agent role in a process seems like a
> possible/useful
>> approach.
>>>
>>> However agents are modeled, I think it is important to have a way to
> describe
>> their provenance...
>>>
>>> Jim
>>>
>>>   >  -----Original Message-----
>>>   >  From: public-prov-wg-request@w3.org
>>> <mailto:public-prov-wg-request@w3.org>  [mailto:public-prov-wg-
>>> <mailto:public-prov-wg->   >  request@w3.org<mailto:request@w3.org>]
> On
>>> Behalf Of Graham Klyne>  Sent: Tuesday, June 21, 2011 2:56 AM>
> To:
>> Zednik, Stephan T.
>>>   >  Cc: Luc Moreau; public-prov-wg@w3.org
>>> <mailto:public-prov-wg@w3.org>   >  Subject: Re: PROV-ISSUE-4:
> Defining
>>> Agent using FOAF's definition
>>>
>>>   >
>>>   >  Stephan Zednik wrote:
>>>   >  >  A thing assumes the role of agent when actively participating
> in
>>> a>  process>  execution?
>>>   >
>>>   >  *If* the concept of an agent is needed, then I think this
> approach is useful.
>>>   >  I.e. it's similar to the view discussion, more easily captured in
> relations.
>>>   >
>>>   >  But I think someone (Jim?) made a comment that the whole notion
> of
>>> an agent>  may not be needed if some of the other concepts can be
> loosened
>> up a little.
>>>   >
>>>   >  #g
>>>   >  --
>>>   >
>>>   >  Stephan Zednik wrote:
>>>   >  >  Hi all,
>>>   >  >
>>>   >  >  To answer Luc's question I originally intended to say that I
>>> thought>  >  an agent can be defined independently of process
>>> execution and I>  >  agreed that an agent should be a node whose
>>> relationship to a process>  >  execution should be defined by a
>> control/participation/influence(?) edge.
>>>   >  >
>>>   >  >  As I thought about it a bit more I began to wonder if agent was
>>> better>  >  described as a role (active participant) a thing takes
> in
>>> the context>  >  of some specific action (in this case a process
>>> execution). An agent>  >  is definitely a thing, but is that thing
>>> always an agent? Or is it an>  >  agent within the context/scope of
> the act it
>> has participated in?
>>>   >  >
>>>   >  >  A thing assumes the role of agent when actively participating
> in
>>> a>  >  process execution?
>>>   >  >
>>>   >  >  I think I am leaning towards making 'agent' status of a thing
>>
>>>> dependent upon active participation in a process execution.
>>>   >  >
>>>   >  >  --Stephan
>>>   >  >
>>>   >  >  On Jun 20, 2011, at 11:28 PM, Luc Moreau
>>> <L.Moreau@ecs.soton.ac.uk<mailto:L.Moreau@ecs.soton.ac.uk>   >  >
>> <mailto:L.Moreau@ecs.soton.ac.uk<mailto:L.Moreau@ecs.soton.ac.uk>>>
>> wrote:
>>>   >  >
>>>   >  >>  Hi,
>>>   >  >>
>>>   >  >>  Reiterating a previous comment I made, can an Agent be defined
>>>>>> independently of process execution?
>>>   >  >>
>>>   >  >>  We can use the definitions of Control/Participation to define
> an
>>>>>> agent's involvement in process execution.
>>>   >  >>
>>>   >  >>  If we see agents/things/process executions as nodes and>  >>
>>> Control/Generation/... as edges of a graph, it would be nice if
> nodes
>>>>>> could be defined independently of edges.
>>>   >  >>
>>>   >  >>  Luc
>>>   >  >>
>>>   >  >>
>>>   >  >>  On 21/06/11 02:33, Satya Sahoo wrote:
>>>   >  >>>  Hi Paul and Stephan,
>>>   >  >>>  In both your definitions, what criteria distinguishes an
> "agent"
>>>   >  >>>  from a "process" - in terms of "do stuff"/"active role or
>>> produces a>  >>>  specified effect"?
>>>   >  >>>
>>>   >  >>>  Reviewing the candidate definitions of Agents, I see that
>>> Jun's,>  >>>  Khalid's and my definitions use an explicit reference
> to
>>> a process>  >>>  (execution).
>>>   >  >>>
>>>   >  >>>  What do you think?
>>>   >  >>>
>>>   >  >>>  Thanks.
>>>   >  >>>
>>>   >  >>>  Best,
>>>   >  >>>  Satya
>>>   >  >>>
>>>   >  >>>  On Mon, Jun 20, 2011 at 4:48 PM, Stephan Zednik<   >  >>>
>>> <mailto:zednis@rpi.edu<mailto:zednis@rpi.edu>>zednis@rpi.edu
>>> <mailto:zednis@rpi.edu>  <mailto:zednis@rpi.edu
> <mailto:zednis@rpi.edu>>>
>> wrote:
>>>   >  >>>
>>>   >  >>>  I like this definition from the New Oxford American
> Dictionary
>>>>>>> because it ties in nicely with provenance>  >>>   >  >>>  "A
>>> person on thing that takes an active role or produces a>  >>>
>>> specified effect."
>>>   >  >>>
>>>   >  >>>  --Stephan
>>>   >  >>>
>>>   >  >>>  On Jun 20, 2011, at 2:08 PM, Paul Groth wrote:
>>>   >  >>>
>>>   >  >>>  >  Hi All,
>>>   >  >>>  >
>>>   >  >>>  >  What would people think of just adopting FOAF's definition
> of
>>>>>>> Agent for now:
>>>   >  >>>  >
>>>   >  >>>  >  The Agent class is the class of agents; things that do
> stuff.
>>> A>  >>>  well known sub-class is Person, representing people. Other
>>> kinds>  >>>  of agents include Organization and Group.
>>>   >  >>>  >
>>>   >  >>>  >
>>>   >  >>>  >  thanks,
>>>   >  >>>  >  Paul
>>>   >  >>>  >
>>>   >  >>>  >
>>>   >  >>>
>>>   >  >>>
>>>   >  >>>
>>>   >
>>>
>>
>>
>> --
>>
>> --------------------------------------------------------------
>>    Dr. Martin Doerr              |  Vox:+30(2810)391625        |
>>    Research Director             |  Fax:+30(2810)391638        |
>>                                  |  Email: martin@ics.forth.gr |
>>                                                                |
>>                  Center for Cultural Informatics               |
>>                  Information Systems Laboratory                |
>>                   Institute of Computer Science                |
>>      Foundation for Research and Technology - Hellas (FORTH)   |
>>                                                                |
>>    Vassilika Vouton,P.O.Box1385,GR71110 Heraklion,Crete,Greece |
>>                                                                |
>>            Web-site: http://www.ics.forth.gr/isl               |
>> --------------------------------------------------------------
>>
>
>
>
>


-- 

--------------------------------------------------------------
  Dr. Martin Doerr              |  Vox:+30(2810)391625        |
  Research Director             |  Fax:+30(2810)391638        |
                                |  Email: martin@ics.forth.gr |
                                                              |
                Center for Cultural Informatics               |
                Information Systems Laboratory                |
                 Institute of Computer Science                |
    Foundation for Research and Technology - Hellas (FORTH)   |
                                                              |
  Vassilika Vouton,P.O.Box1385,GR71110 Heraklion,Crete,Greece |
                                                              |
          Web-site: http://www.ics.forth.gr/isl               |
--------------------------------------------------------------
Received on Thursday, 23 June 2011 16:00:05 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:32 GMT