DigitalFriend Blog

14 July 2003

AAMAS03 Workshop on Humans & Multi-Agent Systems, organised by NASA Researchers

By: gosh'at' (Steve Goschnick)

Often, research papers are judged heavily in the review process by the terminology used - a particular meaning is associated with a term that a community or group has more-or-less settled upon. For this reason I was particularly interested to see which of the following terms came to the fore in this cutting-edge workshop: human-in-the-loop; human-agent interaction; human-robot-agent interaction; mixed human-agent communication; human interaction with multi-agent systems.

AAMAS-2003 is the second running of this international conference that is a merger of three earlier forerunners in the intelligent software agents space. So, the new name is the International Joint Conference on Autonomous Agents and Multi-Agent Systems (AAMAS) - thats a mouthful!

However, my interests (literally) are in two of the affiliated workshops, one of them organised by NASA researchers no less, that are spot-on my research interests, where I've previously have found few researchers in this space, particularly in the local conferences and workshops. The NASA organised one is the Workshop on Humans and Multi-Agent Systems organised and jointly chaired by Debra Schreckenghost and Cheryl Martin, both of the NASA Johnson Space Center. Papers at the workshops at this conference (and others) are often more cutting-edge than many in the main conference stream, as they are less likely to knock-out papers from innovative newcomers via academic-paper technicalities and other subtleties in the Review process (which sometimes cloaks an old-guard protecting the dominant approach in their domain).

The format of this workshop was very instructive in getting engagement and cross-fertilisation of ideas: authors were paired up with one Presenting their paper, while the other did a Commentary regarding how their own paper compared with it - not a critique as such but a commentary that explored the possible relationships with the other paper Presented.

That session was then followed by a lively Panel titled: Communities of Humans and Agents, that I was very happy to participate in, which included: David Morley, Steve Goschnick, Debra Schreckenghost, Elizabeth Sklar, Katia Sycara and Milind Tambe.

I was particularly happy to have our paper accepted in such esteemed company. The NASA people and several of the other authors were right on my research vibe. Our paper is titled: Enacting and Interacting with an Agent-based Digital Self in a 24x7 Web Services World.

The paper is linked here as a .pdf file if you are interested:

Some of the general noteworthy points that came via the Panel session include:

Getting back to my initial question about which term might be most used/come-to-the-fore: many of the people involved were either from NASA or involved in Defence research, and as such their interests were heavily aligned with 'teams' involving both humans and agents (where the agent might also include robots). Whereas my primary interest and that of several other participants, were in agents enhancing the individual (who may well be involved in teams with other people but that was a secondary issue). I.e. I am most interested in the synergy of human + supporting intelligent software agents, which may well create something new, a very empowered individual. So a clear term didn't emerge from this group during discussions. However, HAI (Human-Agent Interaction), as does its predecessor HCI, has that 3-letter pattern so often called upon in Computing.

The other workshop I got a joint paper into was titled 'Web Services and Agent-Based Engineering'. This paper was with my very first Research Masters student Mr Tony Jin, and it is titled: 'Utilizing Web Services in an Agent Based Transaction Model (ABT)'**. I'm happy with this paper and of course with Tony's great research work. I'd put up a list of possible research topics, no more; and he was quite taken by the idea of an 'Envelope-of-Capability' in my ShadowBoard multi-agent architecture, and in putting it to work in using distributed independent web services (WS), to solve the problem of their reliability, durability and quality. It turned out that Tony was strong in transactional databases and he has become intrigued in how ACID properties in tradition transaction models may be up-held in a world full of independent distributed web services - the main envisaged sub-agents in my distributed Shadowboard Agent architecture. I.e. While one couldn't guarantee a result from a single WS, by having a series of alternative but similarly purposed web services, each rated by various quality factors including availability and speed, wrapped together at a meta-data level as an Envelope-of-Capability in an agent system, it becomes possible to deliver a highly reliable complex system, based on many simple and distributed micro/atomic services.

** A later, expanded version of this workshop paper was published as a Book Chapter as follows: Jin, T. & Goschnick, S.B. (2005) Utilizing Web Services in an Agent Based Transaction Model, Ch.13, Extending Web Services Technologies: The Use of Multi-Agent Approaches, edited by Lawrence Cavedon, Zakaria Maamar, David Martin and Boualem Benatallah; Springer, pp.273-291, doi10.1007/0-387-23344-X_13, ISBN: 0-387-23343-1.

An early Draft copy of that Chapter is linked here as a .pdf file for your interest:



About Us | Site Map | Privacy Policy | Contact Us | ©2003 Solid Software Pty Ltd