FIRST CALL FOR PAPERS

                   AAAI-98 Workshop on Software Tools for
                             Developing Agents

                      AAAI-98 Madison, Wisconsin, USA.
                              July 26-27 1998

The study of various kinds of "complete" more or less autonomous intelligent
agents is an important subfield of AI and a large number of software tools
and architectures have been developed to support them.

The workshop will discuss existing and future software tools, especially
integrated toolkits, for developing intelligent agents of various kinds with
a view to providing a conceptual framework for analysing requirements,
assessing strengths and weaknesses of existing tools, and providing
guidelines and suggestions for future developments.  It will include both
bottom-up investigations based on experiences in developing and using
existing toolkits, and top down analyses based on long term research
objectives.

The workshop will bring together researchers who develop or use such tools
to examine the differences between them, discuss common problems and
identify future needs.  Although discussion of specific toolkits is welcome,
the main focus will be on *general* issues affecting the design of toolkits
and their performance rather than on specific applications or their
embodiment of a particular agent theory.  Of particular interest will be
papers which analyse the use of the same tools in several domains or the use
of different tools in the same domain.

Desirable outcomes of the workshop include:

  1. A taxonomy of types of agent development tasks, including both
     practical and research objectives.  This includes identifying task
     features common to a range of tasks, e.g. real-time/anytime,
     uncertainty, multiple goals, etc.
  2. A taxonomy of possible types of toolkits linked to the task taxonomy.
     This includes identifying toolkit features and linking them to task
     features.
  3. An analysis of trade-offs to be considered in designing or selecting a
     toolkit, e.g. generality and flexibility vs efficiency, including
     tradeoffs between directly supporting features in the toolkit vs
     requiring them to be `programmed' by developers or users.
  4. Identifying criteria for evaluating toolkits and technologies for
     instrumentation and logging so that research with agents can become
     more scientific.
  5. Identifying tasks and task environments that are currently mature and
     can be made widely available (e.g., the robocup simulations) and
     technologies for learning, communication, vision, planning and control
     that can be packaged together in an agent-building toolkit, perhaps in
     the form of shared libraries.

More generally, the workshop would help to produce a new more global view of
future requirements which could lead to novel approaches to future toolkits.

Additional information about the workshop is available at:
    http://www.cs.bham.ac.uk/~bsl/aaai-98

Additional information about AAAI-98 is available at:
    http://www.aaai.org/Conferences/National/1998/aaai98.html


Workshop format:

The workshop will include paper presentations and panel-led discussions
together with a poster display.  The paper presentations and panel
discussions will be organised around two main themes:

   * developing individual agents with sophisticated combinations of
     features, e.g. multi-layered architectures, hybrid architectures,
     mechanisms supporting motivation, emotion, intelligent autonomous
     robots, etc.; and
   * developing multi-agent systems composed of small or large collections
     of intelligent agents of varying degrees of sophistication, with
     varying kinds of cooperative and communicative capabilities.

Attendance will be limited to between 25-50 active participants.


Submission requirements:

Potential participants may submit EITHER a full paper addressing the central
themes of the workshop (up to 5000 words), OR a `statements of interest'
which might take the form of an extended abstract or a description or
analysis of a particularly interesting toolkit (up to 1000 words).
Descriptions of toolkits by users or developers may also be submitted as
posters (one page summary).

Postscript files or five hard copies of papers should be sent to:

Brian Logan,
School of Computer Science,
University of Birmingham
Birmingham, B15 2TT, UK

Email: b.s.logan@cs.bham.ac.uk
Tel. +44-121-414-3712
Fax: +44-121-414-4281

 Submission deadline:               March 11, 1998
 Notification date:                 April  1, 1998
 Final date for camera-ready copy:  April 22, 1998


Organising Committee:

Brian Logan
     University of Birmingham, UK (b.s.Logan@cs.bham.ac.uk)
Jeremy Baxter
     DERA Malvern, UK (jbaxter@signal.dra.hmg.gb)
Paul Cohen
     University of Massachusetts, USA (cohen@cs.umass.edu)
Stephen Grand
     Cyberlife Technologies, Cambridge, UK (stepheng@cyberlife.co.uk)

Programme Committee:

Jeffrey Bradshaw, University of West Florida/Boeing, USA
Dolores Canamero, VUB, Belgium
Keith Clark, Imperial College, UK
Michael Coen, MIT, USA
Innes Ferguson, Zuno, UK
Tim Finin, UMBC, USA
Klaus Fischer, DFKI, Germany
Anupam Joshi, University of Missouri, USA
Hyacinth Nwana, BT Labs, UK
Leo Obrst, MITRE Corporation, USA
Enric Plaza, IIIA-CSIC, Catalonia (Spain)
Bart Selman, Cornell University, USA
Aaron Sloman, University of Birmingham, UK
Jan Treur, VU Amsterdam, Netherlands
Mike Wooldridge, QMW, UK