Applied Intelligence 24, 75–89, 2006 c  2006 Springer Science + Business Media, Inc. Manufactured in The Netherlands. Context-Centric Needs Anticipation Using Information Needs Graphs XIAOCONG FAN, RUI WANG, SHUANG SUN AND JOHN YEN School of Information Sciences and Technology, The Pennsylvania State University, University Park, PA 16802 zfan@ist.psu.edu rwang@ist.psu.edu ssun@ist.psu.edu jyen@ist.psu.edu RICHARD A. VOLZ Department of Computer Science, Texas A&M University, College Station, TX 77843 volz@cs.tamu.edu Abstract. Effective agent teamwork requires information exchange to be conducted in a proactive, selective, and intelligent way. In the field of distributed artificial intelligence, there has been increasing number of research focusing on need-driven proactive communication, both theoretically and practically. Among these works, CAST has realized a team-oriented agent architecture where agents, based on a computational shared mental model, are able to anticipate teammates’ information needs and proactively deliver relevant information to the needers in a timely manner. However, the first implementation of CAST takes little consideration of the dynamics of the anticipated information needs, which can change in various ways as the context develops. In this paper we describe a novel mechanism for organizing and managing the “context” of information needs. This allows agents to dynamically activate and deactivate information needs progressively. It has been shown that the two-level context-centric approach can enhance team performance considerably. Keywords: multi-agent systems, teamwork, information needs, contexts 1. Introduction Studies in cognitive science have shown that human team members tend to proactively share new informa- tion to achieve their joint goals [1–3]. “Proactive informa- tion delivery”—sharing relevant information without being asked—has been identified by some psychologists [4–6] as a key characteristic of effective human teams. This has been motivating researchers in Multi-Agent Systems field to also consider empowering software agents with the capability of proactive information delivery [7]. Proactive information delivery is important in multi-agent systems because effective teamwork relies on effective com- munication, which plays an essential role in dynamic team formation [8], in coordinating shared activities [7, 9, 10], and more theoretically, in the forming, evolving, and termi- nating of both joint intentions [11] and shared plans [12]. Researchers in cognitive science have tried to tie proactive information delivery to the capability of anticipating others’ future information needs based on certain shared mental models among team members [13]. Yen, Fan, and Volz [14] have formalized the notion of information needs and in- vestigated the classification of information needs typically emerging in agent teamworks. As they pointed, information needs can be derived from the knowledge of multi-agent information dependence [15] regarding future physical or epistemic commitments (i.e., activities or intentions). Al- though information dependence is very important for sup- porting social reasoning in multi-agent cooperations, the ex- isting researches on dependence theory or framework (e.g., [7, 16–18]) have virtually ignored the issue of dependence dynamics, assuming the models of dependence relations are fixed once established. However, oftentimes, the notion of information needs is context-dependent: the collection of information an agent needs to consider may change over time. It thus indicates that agents in a multi-agent system ought to track the dynamics (contexts) of others’ informa- tion needs in order to offer timely help without disturbing the recipients with information no longer relevant to their activities. CAST (Collaborative Agent architecture for Simulating Teamwork) is a teamwork model focusing on understanding proactive information exchange among teammates based on shared teamwork processes [7]. In the first implementation