DM: Your response needed asap: Agents99 Workshop Proposal on AgentConver

[Jeffrey M. Bradshaw]

We have decided to propose a workshop for Agents99 on agent  
conversation
policies, and we need to provide a tentative list of potential 
attendees.
Could you drop me an email if you think you might attend such a 
workshop?
In addition, if you would like to present a paper, please let me know
that, too.

Thanks!
Jeff

Jeffrey M. Bradshaw
Associate Technical Fellow
Intelligent Agent Technology
The Boeing Company
P.O. Box 3707, M/S 7L-44
Seattle, WA 98124
jeffrey.m.bradshaw@boeing.com
http://www.coginst.uwf.edu/~jbradsha/
425-865-6086 (Tel)
425-865-2965 (Fax)
-------------------------------------------------------------------------------

      AGENTS-99 WORKSHOP ON SPECIFYING AND IMPLENTING CONVERSATION 
POLICIES

This workshop will target researchers working on on ACL theory as 
well as
those
who are building practical agent systems.  Many people in the agents
community,
including participants in FIPA and in the DARPA CoABS program, have 
observed
that important design decisions in ACLs depend on unstated assumptions
about the
extended conversations in which agents using the ACL will participate.
Because
of this, ACL research is slowly broadening from the specification of
individual
ACL primitives to include the characterization of goal-directed 
conversations
for which agents will use ACLs.  Actual agent conversations typically 
fall
into
several recurrent patterns or types, and the specifications defining 
each of
these types can be encoded as agent conversation policies.  This 
workshop will
focus on a number of basic unanswered questions about agent 
conversations and
their governing policies:

1.  Definitional questions:
        * What exactly is (and is not) a conversation policy (CP)?  
What
          important properties of agent interaction is the CP 
abstraction
          intended to capture?
        * How are CPs individuated?  When are two conversations 
instances of
          the same policy?  Are there interesting equivalence classes 
of CPs?
          Is there a hierarchy of types of CPs?
        * How can CPs legally compose with other CPs?
        * Do CPs have a semantics or pragmatics that is distinct from 
that of
          the individual message types which compose the CP?

2.  Formalism questions:
        * What is the best (or even an adequate) specification 
language for
CPs?
          Finite state machines?  Logic specifications?  Goal trees?
        * What formal properties do we want to prove for CPs?
        * Are the same analytical tools we use for network protocols
appropriate
          for conversations using agent protocols?

3.  Practical questions:
        * How are CPs implemented in agent systems?  Are they 
downloaded,
          prebuilt into the agent logic, or constructed from axioms 
on-the-fly?
        * How do agents negotiate the use of a particular CP?
        * How could an agent "learn" an unfamiliar CP?
        * How might an agent legally deviate from a CP?

We plan to select papers by a program committee, with the 
understanding
that we
will be accepting most papers.  We do not wish this workshop to be
exclusionary;
rather, we want to provide an organized opportunity for different 
researchers
who are grappling with these questions to come together.  Because 
there is a
clear gap between theory and practice in this field, we expect that 
most of
the
papers will be steps towards a general account of conversation 
policies,
rather
than more polished developments of an agreed-upon framework.


*******************************

Jeffrey M. Bradshaw
Associate Technical Fellow
Intelligent Agent Technology
The Boeing Company
(425) 865-6086 (msg)
email: jeffrey.m.bradshaw@boeing.com
http://www.coginst.uwf.edu/~jbradsha/