Comparing equilibria for game theoretic and evolutionary bargaining models

Game-theoretic models of bargaining are typically based on the assumption that players have perfect rationality and that they always play an equilibrium strategy. In contrast research research in experimental economics shows that in bargaining between human subjects, participants do not always play the equilibrium strategy. Such agents are said to be boundedly rational. In playing a game against a boundedly rational opponent, a player's most effective strategy is not the equilibrium strategy, but the one that is the best reply to the opponent's actual strategy. Against this background, this paper studies the bargaining behaviour of boundedly rational agents byusing genetic algorithms. Since bargaining involves players with different utility functions, we have two subpopulations - one represents the buyer, and the other represents the seller (i.e., the population is asymmetric). We study the competitive co-evolution of strategies in the two subpopulations for an incomplete information setting, and compare the results with those prescribed by game theory. Our analysis leads to two main conclusions. Firstly, our study shows that although each agent in the game-theoretic model has a strategy that is dominant at every period at which it makes a move, the stable state of the evolutionary model does not always match the game-theoretic equilibrium outcome. secondly, as the players mutually adapt to each other's strategy, the stable outcome depends on the initial population.