Language Evolution and Computation Bibliography

Abstract Learning to communicate is an emerging challenge in AI research. It is known that agents interacting in decentralized, stochastic environments can benefit from exchanging information. Multi-agent planning generally assumes that agents share a common means of communication; however, in building robust distributed systems it is important to address potential miscoordination resulting from misinterpretation of messages exchanged. This paper lays foundations for studying this problem, examining its properties analytically and empirically in a decision-theoretic context. We establish a formal framework for the problem, and identify a collection of necessary and sufficient properties for decision problems that allow agents to employ probabilistic updating schemes in order to learn how to interpret what others are communicating. Solving the problem optimally is often intractable, but our approach enables agents using different languages to converge upon coordination over time. Our experimental work establishes how these methods perform when applied to problems of varying complexity.

Learning to communicate is an emerging challenge in AI research. It is known that agents interacting in decentralized, stochastic environments can benefit from exchanging information. Multiagent planning generally assumes that agents share a common means of communication; however, in building robust distributed systems it is important to address potential mis-coordination resulting from misinterpretation of messages exchanged. This paper lays foundations for studying this problem, examining its properties analytically and empirically in a decision-theoretic context. Solving the problem optimally is often intractable, but our approach enables agents using different languages to converge upon coordination over time.

This paper presents an algorithm for learning the meaning of messages communicated between agents that interact while acting optimally towards a cooperative goal. Our reinforcement-learning method is based on Bayesian filtering and has been adapted for a decentralized control process. Empirical results shed light on the complexity of the learning problem, and on factors affecting the speed of convergence. Designing intelligent agents able to adapt their mutual interpretation of messages exchanged, in order to improve overall task-oriented performance, introduces an essential cognitive capability that can upgrade the current state of the art in multi-agent and human-machine systems to the next level. Learning to communicate while acting will add to the robustness and flexibility of these systems and hence to a more efficient and productive performance.