It hurts more to lose an unfair game. On dynamic psychological games of fairness. Michal Wiktor Krawczyk Center for Research in Experimental Economics and Political Decision-Making (CREED). Universiteit van Amsterdam. Roetersstraat 11, Amsterdam 1018 WB NL. e-mail: m.w.krawczyk@uva.nl Keywords: procedural justice, distributive justice, dynamic psychological games. March 31, 2006 In this paper I present a new model aimed at predicting behavior in games involving risk. The model is designed to capture the relative importance and interaction between procedural justice and distributive justice. Departing from the standard consequentialist perspective, I look beyond sheer outcomes of inter- actions by incorporating also expected payo/s, given strategies. While keeping the model parsimonious and avoiding explicit reference to players intensions, I am able to account for several regularities observed both in the lab and in the eld that jointly pose a challenge to classic models, social-utility models and intentions-based models alike. Dependence of the motivation function on the expected payo/s (which in turn depend on the strategies) cannot be accounted for by classical game theory (in which truncations of strategies to non-played paths are payo/-irrelevant). Therefore I make use of dynamic psychological game theory (Battigalli and Dufwenberg 2005). I begin by considering an n-person material payo/ game in extensive form hN [f0g; H; (y i ) i2N i; where N [f0g = f1; 2:::ng[f0g is the player set, whereby player 0 is interpreted as "nature", H is the set of feasible histories of the game and (y i : Z ! R + [f0g) i2N is a payo/ function, where Z  H is a set of terminal histories (end nodes ). A history of length l is a sequence h =(a 1 ; :::; a l ) 2 H where each a t =(a t 0 ;a t 1 ; :::; a t n ) represents the prole of actions chosen at stage t (1  t  l). Null history (before any actions are made) is dented by h 0 : The set of feasible actions for player i at history h is denoted by A i (h) and it may be a singleton, meaning that i is not active at h. A i (h) is empty if and only if h is a terminal history. Payo/ function (y i ) i2N determines for each terminal history z 2 Z a vector of length n representing non-negative material payo/s obtained by each agent. We let S i2N denote the set of (pure) strategies of player i. Individual strategy is denoted by s i =(s i;h ) h2HnZ , where s i;h 2 A i (h) is the action that would be 1