Reflective Oracles: A Foundation for Game Theory in Artificial Intelligence
Game theory treats players as special: A description of a game contains a full, explicit enumeration of all players. This isn’t a realistic assumption for autonomous intelligent agents. In this paper, we propose a framework in which agents and their environments are both modelled as probablistic oracle machines with access to a “reflective” oracle, which is able to answer questions about the outputs of other machines with access to the same oracle. These oracles avoid diagonalization problems by answering some queries randomly. Agents make decisions by asking the oracle questions about their environment, which they model as an arbitrary oracle machines. Since agents are themselves oracle machines, the environment can contain other agents as non-distinguished subprocesses, removing the special treatment of players in the classical theory. We show that agents interacting in this way play Nash equilibria.
KeywordsNash Equilibrium Cellular Automaton Turing Machine Pure Strategy Causal Decision Theory
Unable to display preview. Download preview PDF.
- 3.Fallenstein, B., Taylor, J., Christiano, P.: Reflective oracles: A foundation for classical game theory. Tech. rep., Machine Intelligence Research Institute (2015), http://intelligence.org/files/ReflectiveOracles.pdf
- 4.Hutter, M.: Universal Artificial Intelligence. Texts in Theoretical Computer Science. Springer (2005)Google Scholar
- 5.Kreps, D.M.: A Course in Microeconomic Theory. Princeton University Press (1990)Google Scholar
- 6.Rogers, H.: Theory of Recursive Functions and Effective Computability. McGraw-Hill (1967)Google Scholar
- 7.Solomonoff, R.J.: A formal theory of inductive inference. Part I 7(1), 1–22 (1964)Google Scholar
- 8.Weirich, P.: Causal decision theory. In: Zalta, E.N. (ed.) The Stanford Encyclopedia of Philosophy (2012)Google Scholar