Prisoner's Dilemma Task

Background

The Prisoner's Dilemma Game (PDG) is a decision-making game about cooperation and competition in resource allocation in the absence of knowing what the other party's decision will be. The PDG is a non-zero-sum game where the rewards are not a straightforward 'I win what you lose (or vice versa)' but rather a more varied pattern of resource allocation depending on the cooperative vs. competitive choices of both parties. The dilemma is to decide between cooperation and competition (aka self-interest) when competition might yield the highest (immediate) reward but loses to cooperation if both parties decide to compete. The name of the dilemma is derived from the
famous thought experiment in which two criminal partners who are interrogated in different rooms must decide whether to betray the other person (in which case they themselves might get a free pass while the other one serves a long prison sentence) or keep silent (in which case they both get a more lenient sentence). The dilemma is that in the absence of knowing what the other person decides to do, a betrayal committed by both will lead to a higher prison sentence for each one of them than had they both decided to stay silent.

The game is a popular game in socioeconomics and is traditionally played using a mathematical payoff matrix that maps out the choices ('strategies') available to both players and the corresponding numbers (payoffs) they receive based on their joint decisions. Instead of specific prison sentences, researchers often simply use abstract numbers representing utility or rewards, where higher numbers are better for the player.

Millisecond's PDG 'with Outcome Matrices' is based on the publication by Lyn Ellett and colleagues from 2013 who studied the relationship between paranoia and behavior in the PDT using such mathematical payoff matrices. A second implementation of the PDG, Millisecond's PDG with reciprocity, is based on the study of Raoul Bell and colleagues from 2017 who studied reciprocal behavior in a PDG setting within an explicit 'investment framework'.

Task Procedure

In the PDG 'with Outcome Matrices', participants first learn about the general set up of the task and work through several examples of
PDG payoff matrices. Choices are simply referred to as strategy 'X' and 'Y'.

A short practice round presents four PDG matrixes and requires participants to select the correct responses for all four possible combinations of choices. For example, a question might be "if you select strategy X and your partner selects strategy Y, what would your partner get?". In this script, values used during practice can either follow the typical PDG pattern (default) or can be assigned randomly. If they follow PDG patterns, strategy X aligns with the 'Cooperative' and strategy Y aligns with the 'Competitive' strategy. Practice is repeated until all 4 questions are answered correctly or 10 practice blocks have run. After 10 practice blocks the script aborts if participants do not provide perfect understanding of the matrix set-up.

By default settings, the test runs one game. Participants are told that they are paired with a co-player. By default, the 'pairing' takes 5s. For each round a new PDG matrix (if more than one game is played) is presented and participants have to make a choice by selecting a button labelled 'X' or 'Y'. After they make their choice, a random delay is introduced that simulates the decision time of the coplayer (sometimes the co-player takes more time, sometimes less). Once both decisions are 'known', feedback is provided and the participant is told what strategy the co-player chose, and how many points they both won. If applicable. the feedback also includes an update of the total number of points won across all games played.

The script implements three patterns to calculate the values presented in the PDG matrices. By choosing to play more than one game, the script provides the opportunity to manipulate the choices the computer will make (more competitive, more cooperative etc.) which allows researchers to study reciprocal behavior and retaliation.

What it Measures

The Prisoner's Dilemma Game is a measure of cooperation and trust

Psychological domains

• Self-Interest: The drive to maximize one's own gains, even at the expense of others or regardless of social consequences
• Risk Assessment: Calculating the reward and cost of betrayal versus the cost and reward of cooperation
• Reciprocal Altruism: Evolutionary drive to help others with the expectation that the favor would be returned

Main Performance Metrics

• number of X-strategy choices: Absolute number of selecting cooperation strategy (X); main measure of cooperation and trust in the co-player

Psychiatric Conditions

Research has identified significant behavior differences in the following groups when tested with a PDG:

• Borderline Personality Disorder (BPD)
• Schizophrenia
• Major Depression (MDD)
• Autism Spectrum Disorder (ASD)

Test Variations

References

Search Google Scholar for peer-reviewed, published research using the Inquisit Prisoner's Dilemma Task.

---

Martin Nowak, & Karl Sigmund. (1993). A strategy of win-stay, lose-shift that outperforms tit-for-tat in the Prisoner's Dilemma game. Nature, 364(6432), 56-8.

Mokros, A., Menner, B., Eisenbarth, H., Alpers, G., Lange, K., Osterheider, M., & Watson, David. (2008). Diminished Cooperativeness of Psychopaths in a Prisoner's Dilemma Game Yields Higher Rewards. Journal of Abnormal Psychology, 117(2), 406-413.

Reed, Zeglen, & Schmidt. (2012). Facial expressions as honest signals of cooperative intent in a one-shot anonymous Prisoner's Dilemma game. Evolution and Human Behavior, 33(3), 200-209.

Lyn Ellett , Rhani Allen-Crooks , Adele Stevens , Tim Wildschut & Paul Chadwick (2013) A paradigm for the study of paranoia in the general population: The Prisoner's Dilemma Game, Cognition and Emotion, 27:1, 53-62

Jahng, Kralik, Hwang, & Jeong. (2017). Neural dynamics of two players when using nonverbal cues to gauge intentions to cooperate during the Prisoner's Dilemma Game. NeuroImage, 157, 263-274.

Bell R, Mieth L, Buchner A (2017). Separating conditional and unconditional cooperation in a sequential Prisoner's Dilemma game. PLoS ONE 12(11)