If Multiagent Learning is the Answer, What is the Question?

Yoav Shoham
Stanford

Over the past year I (together with members of our research group)
have thought a little about work in multiagent learning, in game
theory as well as (more recently) in AI.  My (our) conclusion is that
there is an inhenrent unclarity in the problem or problems being
addressed, which until cleared away will impede progress in the area.
I'll try to elaborate on this, focussing on reinforcement learning as
an example but trying to draw broader lessons.