[an error occurred while processing this directive]

CMSC 791a
Intelligent Agents

Schedule

Reading

Week one

Hyacinth S. Nwana (1996), Software Agents: An Overview, The Knowledge Engineering Review Vol 11 (3). (postscript 2.6M). Agent software is a rapidly developing area of research. However, the 'overuse' of the word agent has tended to mask the fact that, in reality, there is a truly heterogeneous body of research being carried out under this banner. This overview paper presents a typology of agents. It places them in context, defines them and then goes on, inter alia, to overview critically the rationales, hypotheses, goals, challenges and state-of-the-art demonstrators of the various agent types in the typology. Hence, it attempts to make explicit much of what is usually implicit in the agents literature. It also proceeds to overview some other general issues which pertain to all the types of agents in the typology. This paper largely reviews software agents, and it also contains some strong opinions that are not necessarily widely accepted by the agent community. [42 pages]

Week two

M. J. Wooldridge and N. R. Jennings, (1995), Intelligent Agents: Theory and Practice, The Knowledge Engineering Review 10 (2) 115-152. ( postscript (273K), compressed postscript (114K))

Week three

KQML as an agent communication language Tim Finin, Yannis Labrou, and James Mayfield, in Jeff Bradshaw (Ed.), ``Software Agents'', MIT Press, Cambridge, to appear, (1997).

Cohen, P.R. and H. J. Levesque [1990]. Performatives in a rationally based speech act theory, Proceedings of the 28th Annual Meeting of the Association for Computational Linguistics, 1990, pp. 79-88. Abstract: A crucially important adequacy test of any theory of speech acts is its ability to handle performatives. This paper provides a theory of performatives as a test case for our rationally based theory of illocutionary acts. We show why ``I request you...'' is a request, and ``I lie to you that p'' is self-defeating. The analysis supports and extends earlier work of theorists such as Bach and Harnish [1] and takes issue with recent claims by Searle [10] that such performative-as-declarative analyses are doomed to failure.

Week four

Cohen, P.R. and H.J. Levesque [1995]. Communicative Actions for Artificial Agents, Proceedings of the International Conference on Multi-Agent Systems, AAAI Press, San Francisco, June, 1995. Abstract: This paper considers the semantics of the agent communication language KQML. By using this language for communication, agents will be able to request and provide services. Indeed, numerous projects have shown how the language can profitably support interoperation among distributed agents. However, before becoming a widely-accepted standard, it would be worthwhile to examine the language in detail, especially the semantical issues it raises. This paper identifies numerous difficulties with the language, and an attempt is made to point to their resolution. The paper illustrates the kind of semantics we believe to be necessary to characterize agent communication languages, identifies an important adequacy condition (compositionality) and shows how to compose a question from a request and an inform. Finally, the paper discusses possible impacts to be felt on various KQML decisions from the semantical issues raised here.

Cohen, P.R. and H.J. Levesque [1991]. Teamwork, Nous 25(4), Special Issue on Cognitive Science and Artifical Intelligence, pp. 487-512. Abstract: What is involved when a group of agents decide to do something together? Joint action by a team appears to involve more than just the union of simultaneous individual actions, even when those actions are coordinated. We would not say that there is any teamwork involved in ordinary automobile traffic, even though the drivers act simultaneously and are coordinated (one hopes) by the traffic signs and rules of the road. But when a group of drivers decide to do something together, such as driving somewhere as a convoy, it appears that the group acts more like a single agent with beliefs, goals, and intentions of its own, over and above the individual ones. But given that actions are performed by individuals, and that it is individuals who ultimately have the beliefs and goals that engender action, what motivates agents to form teams and act together? In some cases, the answer is obviously the inherent value in doing something together, such as playing tennis, performing a duet, or dancing. These are examples of activities that simply cannot be performed alone. But in many cases, team activity is only one way among many of achieving the goals of the individuals. What benefits do agents expect to derive from their participation in a group effort? In this paper, we attempt to provide an answer to these questions. In particular, we argue that a joint activity is one that is performed by individuals sharing certain specific mental properties. We show how these properties affect and are affected by properties of the participants. Regarding the benefits of teamwork, we show that in return for the overhead involved in participating in a joint activity, an agent expects to be able to share the load in achieving a goal in a way that is robust against certain possible failures and misunderstandings.

Week six

A. Farquhar, R. Fikes, & J. Rice. The Ontolingua Server: A Tool for Collaborative Ontology Construction. Knowledge Systems Laboratory, KSL-96-26, September 1996. Reusable ontologies are becoming increasingly important for tasks such as information integration, knowledge-level interoperation, and knowledge-base development. We have developed a set of tools and services to support the process of achieving consensus on common shared ontologies by geographically distributed groups. These tools make use of the world-wide web to enable wide access and provide users with the ability to pub-lish, browse, create, and edit ontologies stored on an ontology server. Users can quickly assemble a new ontology from a library of modules. We discuss how our system was constructed, how it exploits existing protocols and browsing tools, and our experience supporting hundreds of users. We describe applications using our tools to achieve con-sensus on ontologies and to integrate information. The Ontolingua Server may be accessed through the URL http://ontolingua.stanford.edu/.

R. Fikes, A. Farquhar, & W. Pratt. Information Brokers: Gathering Information from Heterogeneous Information Sources. In John H. Stewman, Ed., Ninth Florida Artificial Intelligence Research Symposium (FLAIRS-96), Key West, Florida, 192-197. Eckerd College, 1996. The Internet provides dramatic new opportunities for gathering information from multiple, distributed, heterogeneous information sources. However, this distributed environment poses difficult technical problems for the information-seeking client, including finding the information sources relevant to an interest, formulating questions in the terms that the sources understand, interpreting the retrieved information, and assembling the information retrieved from several sources into a coherent answer. In this paper, we describe techniques that will enable vendors and buyers to build and maintain network-based information brokers capable of retrieving information about services and products via the Internet from multiple vendor catalogs and data bases for both human and computer-based clients.

Read for April 14

James Mayfield, Tim Finin, Rajkumar Narayanaswamy, Chetan Shah William MacCartney & Keith Goolsbey, The Cycic Friends Network: Getting Cyc agents to reason together. Proceedings of the CIKM '95 Workshop on Intelligent Information Agents. (PostScript version, 150Kb), Presentation Slides (PostScript version, 350Kb).

Read for April 16

Agent Development Support for Tcl, R. Scott Cost, Tim Finin, Jeegar Lakhani, Ethan Miller, Charles Nicholas and Ian Soboroff, submitted to 5th Annual Tcl/Tk Workshop '97, July 14-17, 1997 Boston. Tcl/Tk is an attractive language for the design of intelligent agents because it allows the quick construction of prototypes and user interfaces; new scripts can easily be bound at runtime to respond to events; and execution state is encapsulated by the interpreter, which helps in agent migration. However, a system of intelligent agents must share a common language for communicating requests and knowledge. We have integrated KQML (Knowledge Query Manipulation Language), one such standard language, into Tcl/Tk. The resulting system, called TKQML, provides several benefits to those building intelligent agent systems. First, TKQML allows easy integration of existing tools which have Tcl/Tk interfaces with an agent system by using Tcl to move information between KQML and the application. Second, TKQML is an excellent language with which to build agents, allowing on-the-fly specification of message handlers and construction of graphical interfaces. This paper describes the implementation of TKQML, and discusses its use in our intelligent agent system for information retrieval.

Read for Monday, April 21, 1997

Grace Crowder and Charles Nicholas, Resource Selection in CAFE: an Architecture for Network Information Retrieval Networked Information Retrieval workshop at SIGIR'96, August 1996. Final versions of all the workshop papers are available in electronic form here.

Claudia Pearce and Ethan Miller, "The TELLTALE Dynamic Hypertext Environment: Approaches to Scalability," in Advances in Intelligent Hypertext, J. Mayfield and C. Nicholas, eds. Lecture Notes in Computer Science, Springer-Verlag, in preparation.

Read for Wednesday, April 22, 1997

This class will look at the Foundation for Intelligent Physical Agents.

Promotion of the development and specification of agent technologies is the task of the Foundation for Intelligent Physical Agents (FIPA). FIPA is an international non-profit association of companies and organisations which agree to share efforts to produce in a timely fashion internationally agreed specifications of generic agent technologies that are usable across a large number of applications providing a high level of interoperability across applications. The target of FIPA-specified agent technologies are Intelligent Physical Agents (IPA). IPAs are devices intended for the mass market, capable of executing actions to accomplish goals imparted by or in collaboration with human beings or other IPAs, with a high degree of intelligence.

Visit the FIPA site and read the current draft FIPA proposal for a standard Agent Communication Language along with some comments (hardcopy).

Read for Wednesday, April 30, 1997

Mobile Agents White Paper, Jim White, General Magic, 1996.

Mobile Agents: Are they a good idea?, Colin Harrison, David Chess and Aaron Kershenbaum. Research report, IBM T.J. Watson Research Center, March 1995.

Additional readings:

Robert S. Gray. Agent Tcl: A flexible and secure mobile-agent system. In Mark Diekhans and Mark Roseman, editors, Proceedings of the Fourth Annual Tcl/Tk Workshop (TCL 96), Monterey, California, July 1996. This paper describes the architecture of Agent Tcl and its security mechanisms. The paper is available in compressed postscript (71,833 bytes).