Networks Blog Post 1
In a rather recent article, Economists Finally Test Prisoner’s Dilemma on Prisoners, published in the LA Times, journalist Geoffrey Mohan explores the results of a recent experiment where German economists test the prisoner’s dilemma, for the first time, on prisoners. While there have been previous studies of prisoners dilemma, the subjects have always been student recruits—this experiment marked the first time where actual prisoners were used in “prisoner’s dilemma”. The pair of German economists, Andreas Lange and Menusch Khadjavi, simulated iterations of prisoner’s dilemma on groups of female prisoners and a control group of female college students. The results were rather surprising: the data revealed that 1. The results did not conform to the expected Nash equilibrium and 2. That the German prisoner’s, were on average, more likely to cooperate than their student counterparts.
According to the classic model of prisoner’s dilemma, the Nash equilibrium is that both parties would defect for mutually lower returns. However, the data proves otherwise: “In the simultaneous game, a greater share of the prisoners solved their dilemma through cooperation: 55%… compared with 37% of students” (Mohan). How do we reconcile the discrepancy between the Nash equilibrium and experimental data? What do we make of the discrepancy between the behaviors of prisoners and students?
Although the article deliberately refrains from speculating on underlying reasons, it does recognize that unconsidered variables do result in an unexpected results in the experimental data. The explanation for the discrepancy between the traditional model of prisoner’s dilemma and the discrepancy between the students and prisoners boils down to various unconsidered variables. Analogous to many physical laws and theories, the Nash Equilibrium only holds true under simple cases that work under various assumptions. The experiment introduce numerous various variables such as: social preferences, reciprocity, varying degrees of risks and consequences, magnitudes of the rewards/losses will drastically affect the prisoner’s dilemma outcome. For example, a plausible unconsidered variable is that a prison’s social structure may provide incentive to cooperate: as the benefits for confessing, say a reduced sentence, are drastically reduced by consequences resulting from compromising one’s position in the prison social structure, amongst other associated risks. It’s clear that real situations do not necessarily conform to the textbook models. Although the Nash equilibrium is satisfactory in revealing insight into simple cases of games, it’s clear that refinements and adjustments must be made to the model, to consider more relevant variables, for the model to produce more accurate reflection of complex, experimental situations.
25, Geoffrey Mohan July. “Economists Finally Test Prisoner’s Dilemma on Prisoners.” Los Angeles Times. Los Angeles Times, 25 July 2013. Web. 30 Sept. 2013. <http://www.latimes.com/news/science/sciencenow/la-sci-sn-prisoners-dilemma-cooperation-20130725,0,4705696.story>.