The Association for the Advancement of Artificial Intelligence held its 2001 Fall Symposium Series November 2-4, 2001 at the Sea Crest Conference Center in North Falmouth, Massachusetts. The topics of the five symposia in the 2001 Fall Symposia Series were (1) Anchoring Symbols to Sensor Data in Single and Multiple Robot Systems, (2) Emotional and Intelligent II: The Tangled Knot of Social Cognition, (3) Intent Inference for Collaborative Tasks, (4) Negotiation Methods for Autonomous Cooperative Systems, and (5) Using Uncertainty within Computation. This article contains brief reports of those five symposia.

The innovative reapplication of a multiagent system for human-in-the-loop (HIL) simulation was a consequence of appropriate agent-oriented design. The use of intelligent agents for simulating human decision making offers the potential for analysis and design methodologies that do not distinguish between agent and human until implementation. With this as a driver in the design process, the construction of systems in which humans and agents can be interchanged is simplified. Two systems have been constructed and deployed to provide defense analysts with the tools required to advise and assist the Australian Defense Force in the conduct of maritime surveillance and patrol. The experiences gained from this process indicate that it is simpler, both in design and implementation, to add humans to a system designed for intelligent agents than it is to add intelligent agents to a system designed for humans.

The operation of a human organization requires dozens of everyday tasks to ensure coherence in organizational activities, monitor the status of such activities, gather information relevant to the organization, keep everyone in the organization informed, and so on. Teams of software agents can aid humans in accomplishing these tasks, facilitating the organization's coherent functioning and rapid response to crises and reducing the burden on humans. Based on this vision, this article reports on ELECTRIC ELVES, a system that has been operational 24 hours a day, 7 days a week at our research institute since 1 June 2000. Tied to individual user workstations, fax machines, voice, and mobile devices such as cell phones and palm pilots, ELECTRIC ELVES has assisted us in routine tasks, such as rescheduling meetings, selecting presenters for research meetings, tracking people's locations, organizing lunch meetings, and so on. We discuss the underlying AI technologies that led to the success of ELECTRIC ELVES, including technologies devoted to agent-human interactions, agent coordination, the accessing of multiple heterogeneous information sources, dynamic assignment of organizational tasks, and the deriving of information about organization members. We also report the results of deploying ELECTRIC ELVES in our own research organization.

We consider the problem of designing the the utility functions of the utility-maximizing agents in a multi-agent system so that they work synergistically to maximize a global utility. The particular problem domain we explore is the control of network routing by placing agents on all the routers in the network. Conventional approaches to this task have the agents all use the Ideal Shortest Path routing Algorithm (ISPA). We demonstrate that in many cases, due to the side-effects of one agent's actions on another agent's performance, having agents use ISPA's is suboptimal as far as global aggregate cost is concerned, even when they are only used to route infinitesimally small amounts of traffic. The utility functions of the individual agents are not ``aligned'' with the global utility, intuitively speaking. As a particular example of this we present an instance of Braess' paradox in which adding new links to a network whose agents all use the ISPA results in a decrease in overall throughput. We also demonstrate that load-balancing, in which the agents' decisions are collectively made to optimize the global cost incurred by all traffic currently being routed, is suboptimal as far as global cost averaged across time is concerned. This is also due to `side-effects', in this case of current routing decision on future traffic. The mathematics of Collective Intelligence (COIN) is concerned precisely with the issue of avoiding such deleterious side-effects in multi-agent systems, both over time and space. We present key concepts from that mathematics and use them to derive an algorithm whose ideal version should have better performance than that of having all agents use the ISPA, even in the infinitesimal limit. We present experiments verifying this, and also showing that a machine-learning-based version of this COIN algorithm in which costs are only imprecisely estimated via empirical means (a version potentially applicable in the real world) also outperforms the ISPA, despite having access to less information than does the ISPA. In particular, this COIN algorithm almost always avoids Braess' paradox.

Despite the significant progress in multiagent teamwork, existing research does not address the optimality of its prescriptions nor the complexity of the teamwork problem. Without a characterization of the optimality-complexity tradeoffs, it is impossible to determine whether the assumptions and approximations made by a particular theory gain enough efficiency to justify the losses in overall performance. To provide a tool for use by multiagent researchers in evaluating this tradeoff, we present a unified framework, the COMmunicative Multiagent Team Decision Problem (COM-MTDP). The COM-MTDP model combines and extends existing multiagent theories, such as decentralized partially observable Markov decision processes and economic team theory. In addition to their generality of representation, COM-MTDPs also support the analysis of both the optimality of team performance and the computational complexity of the agents' decision problem. In analyzing complexity, we present a breakdown of the computational complexity of constructing optimal teams under various classes of problem domains, along the dimensions of observability and communication cost. In analyzing optimality, we exploit the COM-MTDP's ability to encode existing teamwork theories and models to encode two instantiations of joint intentions theory taken from the literature. Furthermore, the COM-MTDP model provides a basis for the development of novel team coordination algorithms. We derive a domain-independent criterion for optimal communication and provide a comparative analysis of the two joint intentions instantiations with respect to this optimal policy. We have implemented a reusable, domain-independent software package based on COM-MTDPs to analyze teamwork coordination strategies, and we demonstrate its use by encoding and evaluating the two joint intentions strategies within an example domain.

Chapter 6 three motivating examples are considered.

As a result, they knowledge of an agent (that is, its epistemic coarse-grained level of abstraction, KBwould argue, it is not possible to discuss state) can be characterized as the Ss can be characterized in terms of two the knowledge of a system independently collection of all possible worlds that components: (1) a knowledge base, encoding of the task context in which are consistent with the knowledge the knowledge embodied by the system is meant to operate. I won't held by the agent. If the knowledge of the system, and (2) a reasoning engine, go into too many details here because the agent is complete, then the epistemic which is able to query the knowledge a detailed discussion of the declarative state contains only one world. A base, infer or acquire knowledge from versus the procedural argument is well nice feature of Levesque and Lakemeyer's external sources, and add new knowledge beyond the scope of this review. The treatment of epistemic logic is that to the knowledge base. Levesque important point to make is that in contrast to many other treatments and Lakemeyer's The Logic of Knowledge Levesque and Lakemeyer's approach is of modalities, the discussion is reasonably Bases deals with the "internal logic" of situated in a precise AI research easy to follow for people who are a KBS: It provides a formal account of paradigm, which considers knowledge not experts in the field. This is the result the interaction between a reasoning bases as declaratively specified, task-independent of two main features of this analysis: engine and a knowledge base.

The problem of reinforcement learning in large factored Markov decision processes is explored. The Q-value of a state-action pair is approximated by the free energy of a product of experts network. Network parameters are learned online using a modified SARSA algorithm which minimizes the inconsistency of the Q-values of consecutive state-action pairs. Actions are chosen based on the current value estimates by fixing the current state and sampling actions from the network using Gibbs sampling. The algorithm is tested on a cooperative multi-agent task. The product of experts model is found to perform comparably to table-based Q-Iearning for small instances of the task, and continues to perform well when the problem becomes too large for a table-based representation.

The problem of reinforcement learning in large factored Markov decision processes is explored. The Q-value of a state-action pair is approximated by the free energy of a product of experts network. Network parameters are learned online using a modified SARSA algorithm which minimizes the inconsistency of the Q-values of consecutive state-action pairs. Actions arechosen based on the current value estimates by fixing the current state and sampling actions from the network using Gibbs sampling. The algorithm is tested on a cooperative multi-agent task. The product of experts model is found to perform comparably to table-based Q-Iearning for small instances of the task, and continues to perform well when the problem becomes too large for a table-based representation.

This article gives an overview of current research on animated pedagogical agents at the Center for Advanced Research in Technology for Education (CARTE) at the University of Southern California/Information Sciences Institute. Animated pedagogical agents, nicknamed guidebots, interact with learners to help keep learning activities on track. They combine the pedagogical expertise of intelligent tutoring systems with the interpersonal interaction capabilities of embodied conversational characters. They can support the acquisition of team skills as well as skills performed alone by individuals. At CARTE, we have been developing guidebots that help learners acquire a variety of problem-solving skills in virtual worlds, in multimedia environments, and on the web. We are also developing technologies for creating interactive pedagogical dramas populated with guidebots and other autonomous animated characters.