Intelligent Adaptive Agents

Article excerpt

There is a great dispute among researchers about the roles, characteristics, and specifications of what are called agents, intelligent agents, and adaptive agents. Most research in the field focuses on methodologies for solving specific problems (for example, communications, cooperation, architectures), and little work has been accomplished to highlight and distinguish the field of intelligent agents. As a result, more and more research is cataloged as research on intelligent agents. Therefore, it was necessary to bring together researchers working in the field to define initial boundaries, criteria, and acceptable characteristics of the field. The Workshop on Intelligent Adaptive Agents, presented as part of the Thirteenth National Conference on Artificial Intelligence, addressed these issues as well as many others that are presented in this article.

If we were to ask 10 researchers from different organizations or institutions what their personal definition of an intelligent agent is, we would most likely get 8 to 10 different answers. Moreover, if a researcher or any curious person wanted to learn about intelligent agents, he/she might get confused after reading even a few papers of the hundreds that were recently published on agency-related subjects. Reasons for this confusion include the following: First, there is no standard definition of what an intelligent agent is. (Today, almost anything can be called an agent and, typically, an intelligent agent). Second, no clear goals or objectives for the agent (for example, the functions of the agent vary from implicit to explicit, systematic-mechanic to environmental, system requirement to user requirement, simple to complex). Third, the agentuser relationship is either missing or vague.

The Workshop on Intelligent Adaptive Agents, part of the Thirteenth National Conference on Artificial Intelligence (AAAI-96), presented state-of-the-art approaches, ideas, and methodologies for research and development of intelligent adaptive agents. The workshop consisted of two invited talks, presented by Brian Gaines and Barbara Hayes-Roth; four discussion sessions, organized and chaired by John Laird, Sandip Sen, Costas Tsatsoulis, and Kerstin Voigt; two commentary evaluations, presented by Yves Kodratoff and Brad Whitehall; and 10 papers, presented by Keith Decker, Karen Haigh, Ibrahim Imam, John Laird, Ramiro Liscano, Daniela Rus, Sandip Sen, Rahul Sukthankar, Kerstin Voight, and Grace Yee.

Intelligence, adaptation, and agency are three terms with no standard definitions accepted by researchers in the AI community. Defining the scope of the AAAI-96 workshop and understanding these terms are two associated issues. In general, the workshop focused on research involving the three issues together or in different combinations. For example, the scope of the workshop covered research and development on intelligent adaptive methodologies for agents and intelligent agents that behave adaptively. The definition of these terms were discussed in some presentations as well as some discussion sessions. A summary of these discussions is presented in the next section. The research presented at the workshop can be classified according to varying criteria. These criteria and a classification of the papers are presented in the following section. A brief description of the talks presented at the workshop is described in a later section. The last section introduces a classification of the research presented at the workshop according to conceptual and systematic criteria.

What Is an Agent?

Because the workshop presented diverse definitions of what an adaptive agent is, a discussion on the definition of an agent, a society of agents, and an intelligent agent was also an important part of the workshop. Some issues that are typically discussed in defining any agent were found less important than previously believed, especially when clearly differentiating between an agent and a procedure. …