Halpern helped create a model of knowledge representation which consisted of a set of states of the world, and each person has a partition into a collection of sets of states, where states are in the same partition if that person can't distinguish between those states. You can use this system to define knowledge and common knowledge, and model problems like muddy children. It also serves as a great framework for temporal logic.
Halpern led the creation of the Computing Research Repository (CoRR), a forerunner of arXiv, and would later moderate CS papers for arXiv.
Joe Halpern was the driving force behind the Theoretical Aspects of Rationality and Knowledge (TARK) conference, which attracts philosophers, economists, computer scientists and others to discuss what it means to know stuff. I had two papers in TARK 2009 in Stanford. But my favorite TARK memory came from a debate at the 1998 TARK conference at Northwestern.
Consider the centipede game, where two players alternate turns where each can either play to the right (R/r), or defect (D/d) to end the game immediately, with payouts in the diagram below.
The debate asked the following. Player 1 needs to think about the backward induction of the future moves, considering the case where player 2 played right in its first move. But this is an irrational move, so why should you assume player 2 is being rational when playing its second move later on?
Is the title correct? Or was this last year? Time is flying.
ReplyDeleteFixed.
Delete