Secretive interaction

We study the behaviour of n players engaged in a conversation, each attempting to predict the behaviour of the others. We prove that, under a certain notion of interaction, it is possible to learn all other players, but, generally, not without being learned back. We then focus on the interaction of two players and consider, for different methodological restrictions based on the philosophy of science, whether some form of secretive learning can be achieved.

Authors

• Eduardo Skapinakis

Keywords

• Inductive inference
• Interactive learning
• Secrecy
• Learning strategies