• Professor Luc Steels is the founder and director of the Artificial Intelligence Laboratory at the Vrije Universiteit Brussel and the Sony Computer Science Laboratory Paris. Luc Steels' research focuses on the origin and evolution of language.

• David Parkes, John L. Loeb Associate Professor of the Natural Sciences and Associate Professor of Computer Science, Harvard University

Semiotic Dynamics and the Recruitment Theory of Language Origins

Luc Steels
Vrije Universiteit Brussel (AI Lab) and Sony Computer Science Laboratory - Paris

Recent years have seen tremendous advances in understanding and modeling the dynamical processes whereby agents self-organise a communication system by grounded situated interactions. This requires not only mastering the nature of parsing and production processes (based on the state of the art in computational linguistics) but also understanding how agents might invent new bits of language as speaker, adopt them as hearer, and negotiate them in a collective dynamics leading to a shared language system. The results obtained so far have lead to the development of a new theory on the origins of language, which contrasts with the language as adaptation theory advanced by Pinker, Jackendoff, et al. This theory argues that the language faculty is a dynamically configured collection of mechanisms implementing strategies that increase communicative success and expressive power while decreasing effort (time, memory, processing, etc.). The mechanisms involved are not specific to language and genetic evolution is not taken to play a causal role in the evolution of language.

This talk first sketches the state of the art in the emerging field of semiotic dynamics. It then summarises the debate on the origins of language and gives examples of experiments showing how the recruitment of particular mechanisms makes a difference for building a better communication system.

Computational Mechanism Design: An AI Agenda

David Parkes

Computational mechanism design (CMD) brings together the concern in microeconomics about decision making in the context of distributed private information and self-interest with the concern in computer science about computational and communication complexity. CMD suggests an approach to modeling and solving problems in distributed AI, including those of multi-agent planning, multi-agent learning and distributed problem solving. CMD can be used to provide an approach to promote collaborative behavior amongst a team of self-interested computational agents. An active topic of research in the theoretical computer science community, there are many avenues of research in CMD that should have direct appeal to the artificial intelligence community.