Skip to main content

Experimental data: Inferring strategies from observations in long iterated prisoner’s dilemma experiments

Cite this dataset

Montero, Eladio (2022). Experimental data: Inferring strategies from observations in long iterated prisoner’s dilemma experiments [Dataset]. Dryad.


While many theoretical studies have revealed the strategies that could lead to and maintain cooperation in the Iterated Prisoner's Dilemma, less is known about what human participants actually do in this game and how strategies change when being confronted with anonymous partners in each round. Previous attempts used short experiments, made different assumptions of possible strategies, and led to very different conclusions. We present here two long treatments that differ in the partner matching strategy used, i.e. fixed or shuffled partners. Here we use unsupervised methods to cluster the players based on their actions and then Hidden Markov Model to infer what are those strategies in each cluster. Analysis of the inferred strategies reveals that fixed partner interaction leads to a behavioral self-organization. Shuffled partners generate subgroups of strategies that remain entangled, apparently blocking the self-selection process that leads to fully cooperating participants in the fixed partner treatment. Analyzing the latter in more detail shows that AllC, AllD, TFT- and WSLS-like behavior can be observed. This study also reveals that long treatments are needed as experiments less than 25 rounds capture mostly the learning phase participants go through in these kinds of experiments.


The data were collected for two treatments wherein participants played long IPD in two different pairwise configurations, i.e. fixed partners (FP) and shuffled partners (SP). The data from these experiments were collected in Brussels, Belgium, at the Brussels Experimental Economics Laboratory (BEEL), part of the Vrije Universiteit Brussel (VUB). Twelve sessions were held, in total 188 participants were recruited. The group of participants consisted of 48\% female and 52\% male. The average age in all sessions was 22+-8 years old (see the Supplementary Information for detailed data about the participants). The experiments were approved by the Ethical Commission for Human Sciences at the VUB (ECHW2015_3). The participants played the IPD on individual isolated laptops, designated to avoid any type of communication or arrangements by the participants, who remained anonymous throughout the experiment. At the end of the experiments, subjects were asked optional questions about their age, gender, and information about the game they just played (see the Supplementary Information for the detailed questions and answers). The FP treatment consisted of 6 sessions where 92 participants played pairwise IPD with the same opponent. The SP treatment consisted of 6 sessions where participants played the same IPD as FP but with a different opponent each round, hence the "shuffled" name. The IPD parameters are R = 3, T = 4, S = 0, P = 1. 

At the beginning of the experiment, every participant read a detailed instruction document with a small test at the end to make sure they understood the game dynamics (see Supplementary Information). At the end of the experiment, participants were given their gains (mean = 7.85 euros, SD = 1.67 in the SP treatment and mean = 10.61, SD = 3.87 in the FP treatment) and a show-up fee of 2.5 euros.

Usage notes

Please refer to the file README.txt for a full detail of the columns used in the dataset.


Vlaams AI-Onderzoeksprogramma, Award: 952215

Fund for Scientific Research, Award: 40005955

Fund for Scientific Research, Award: 31257234

Research Foundation - Flanders, Award: G.0391.13N

FuturICT 2.0, Award: FLAG-ERA JCT 2016

Service Public de Wallonie Recherche, Award: 2010235–ARIAC