Computational Intelligence, Volume 18, Number 2, 2002 TOWARD A FORMALISM FOR CONVERSATION PROTOCOLS USING JOINT INTENTION THEORY SANJEEV KUMAR,MARCUS J. HUBER,PHILIP R. COHEN, AND DAVID R. MCGEE Department of Computer Science and Engineering, Oregon Graduate Institute Conversation protocols are used to achieve certain goals or to bring about certain states in the world. Therefore, one may identify the landmarks or the states that must be brought about during the goal-directed execution of a protocol. Accordingly, the landmarks, characterized by propositions that are true in the state represented by that landmark, are the most important aspect of a protocol. Families of conversation protocols can be expressed formally as partially ordered landmarks after the landmarks necessary to achieve a goal have been identified. Concrete protocols represented as joint action expressions can, then, be derived from the partially ordered landmarks and executed directly by joint intention interpreters. This approach of applying Joint Intention theory to protocols also supports flexibility in the actions used to get to landmarks, shortcutting protocol execution, automatic exception handling, and correctness criterion for protocols and protocol compositions. 1. INTRODUCTION Conversation protocols represent communication patterns in multi-agent interactions. These protocols are traditionally specified as finite state machines in which the transition arcs specify the communicative actions to be used by the various agents involved in a con- versation. Protocols are executed by performing these communicative actions and therefore, the communicative actions have come to be regarded as the central concept around which analyses of protocols are based. However, it is the states and not the state transitions that are key to the correctness and completeness of a protocol (Smith and Cohen 1996; Yolum and Singh 2000). In a previous paper on agent communication languages (Smith et al. 1998), we had suggested a landmark-based approach for formal analysis of conversation protocols wherein the most important aspect of a conversation protocol is not the communicative ac- tions involved in that protocol but the effects or the states that these actions bring about. The basic idea is that since protocols are used to perform certain tasks or to bring about certain state of affairs in the world, one might identify the important landmarks or the states of affairs that are brought about by and during the execution of a protocol. Conversation protocols can then be expressed at an abstract level as partially ordered landmarks where each landmark is characterized by propositions that are true in the state represented by that landmark. Several different actions can bring about the same state and therefore, the partially ordered landmarks represent a family of protocols. Communicative actions are the tools to realize concrete pro- tocols from a landmark-based representation. Besides contributing to formal analyses of protocol families, the landmark-based representation facilitates techniques similar to partial order planning (Minton et al. 1994) for dynamically choosing the most appropriate action to use next in a conversation, allows compact handling of protocol exceptions, and in some cases, even allows short circuiting a protocol by opportunistically skipping some intermediate landmarks. Limitations of the finite state machine representation for protocols have led agent re- searchers to explore various other techniques such as Definite Clause Grammars (Labrou and Finin 1997b), Colored Petri Nets (Cost et al. 1999), and enhanced Dooley graphs (Parunak 1996). Each of these alternatives attempts to remedy some shortcoming of the finite state Address correspondence to the authors at Oregon Graduate Institute, Department of Computer Science and Engineering, 20000 NW Walker Road, Beaverton, OR 97006. E-mail: {skumar, marcush, pcohen}@cse.ogi.edu. David R. McGee is now a researcher at Pacific Northwest National Laboratory operated by Battelle for US DOE. E-mail: david.mcgee@pnl.gov. C  2002 Blackwell Publishing, 350 Main Street, Malden, MA 02148, USA, and 108 Cowley Road, Oxford, OX4 IJF, UK.