Perfect Prediction in Normal Form: Superrational Thinking Extended to Non-Symmetric Games

TL;DR

This paper introduces the Perfectly Transparent Equilibrium, a new solution concept for non-cooperative normal-form games where players have perfect prediction of each other's strategies, extending superrationality to non-symmetric games and ensuring Pareto optimality.

Contribution

It formalizes the Perfectly Transparent Equilibrium, proves its uniqueness and Pareto optimality, and extends superrationality to non-symmetric games within a non-Nashian decision framework.

Findings

The equilibrium is unique when it exists.

It coincides with Hofstadter's superrationality in symmetric games.

The equilibrium is Pareto optimal.

Abstract

This paper introduces a new solution concept for non-cooperative games in normal form with no ties and pure strategies: the Perfectly Transparent Equilibrium. The players are rational in all possible worlds and know each other's strategies in all possible worlds, which together we refer to as Perfect Prediction. The anticipation of a player's decision by their opponents is counterfactually dependent on the decision, unlike in Nash Equilibra where the decisions are made independently. The equilibrium, when it exists, is unique and is Pareto optimal. This equilibrium is the normal-form counterpart of the Perfect Prediction Equilibrium; the prediction happens "in another room" rather than in the past. The equilibrium can also be seen as a natural extension of Hofstadter's superrationality to non-symmetric games. Algorithmically, an iterated elimination of non-individually-rational strategy profiles is performed until at most one remains. An equilibrium is a strategy profile that is immune against knowledge of strategies in all possible worlds and rationality in all possible worlds, a stronger concept than common knowledge of rationality but also than common counterfactual belief of rationality. We formalize and contrast the Non-Nashian Decision Theory paradigm, common to this and several other papers, with Causal Decision Theory and Evidential Decision Theory. We define the Perfectly Transparent Equilibrium algorithmically and, when it exists, prove its uniqueness, its Pareto-optimality, and that it coincides with the Hofstadter's Superrationality on symmetric games. We also relate to concepts found in literature such as Individual Rationality, Minimax-Rationalizability, the Correlated Equilibrium. Finally, we specifically discuss inclusion relationships on the special case of symmetric games and contrast them with asymmetric games.