We will explore issues common to three areas of investigation:
In addition to same-type learning (like animals interacting with animals), we are also interested in cross-type learning (like computers interacting with people). For learning to have effect, we focus on the repeated game setting, which gives the participants an opportunity to observe and predict its opponents.
We hope to gain a better understanding of the structure and dynamics resulting from interacting learning decision makers. We also would like to better understand how decision making can and should be carried out. For example, do people attach a value to a given action in a game, and consistently choose higher-valued actions? Or do they attempt to find patterns in the behavior of their opponent and respond to maximize scores against the perceived pattern? Perhaps the mechanism varies from person to person, or employs some as-yet undescribed process. These are some of the questions we hope are illuminated by examining current and ongoing research on interacting algorithms and behavioral experiments.