Source-linked AI summary
Heuristics guide the implementation of social preferences in one-shot Prisoner's Dilemma experiments
Valerio Capraro, Jillian J. Jordan, David G. Rand
TL;DR
The motivations behind cooperation in one-shot anonymous Prisoner’s Dilemmas remain unclear, particularly because standard social-preference models do not readily explain the observed distribution. The paper examines cooperation across b/c levels and participants’ stated reasons; it finds trimodal giving, persistent half-cooperation alongside b/c-related shifts, and a positive association with later Dictator Game giving.
Problem
Standard social-preference models do not readily explain the trimodal cooperation distribution, particularly its substantial 50% mode.
Source: Results · p. 4, Discussion · p. 6Method
The paper examines cooperation distributions across b/c levels and participants’ end-of-study explanations for choosing 50% cooperation.
Source: Results · p. 4, Discussion · p. 7Results
Cooperation is trimodal, b/c shifts probability from zero to full without changing half-giving, and PD cooperation is strongly associated with subsequent DG giving.
Source: Results · p. 5, Discussion · p. 6Takeaways & Limitations
The findings support heuristics as guides to social preferences and suggest prosocial tendencies extend across PD and DG play.
Source: Results · p. 5, Discussion · p. 8The between-subjects design cannot track individuals changing behavior, so the study cannot distinguish personal b/c thresholds from shifts in who cooperates.
Source: Discussion · p. 7
Abstract
from arXiv · showhide
Cooperation in one-shot anonymous interactions is a widely documented aspect of human behaviour. Here we shed light on the motivations behind this behaviour by experimentally exploring cooperation in a one-shot continuous-strategy Prisoner's Dilemma (i.e. one-shot two-player Public Goods Game). We examine the distribution of cooperation amounts, and how that distribution varies based on the benefit-to-cost ratio of cooperation (b/c). Interestingly, we find a trimodal distribution at all b/c values investigated. Increasing b/c decreases the fraction of participants engaging in zero cooperation and increases the fraction engaging in maximal cooperation, suggesting a role for efficiency concerns. However, a substantial fraction of participants consistently engage in 50% cooperation regardless of b/c. The presence of these persistent 50% cooperators is surprising, and not easily explained by standard models of social preferences. We present evidence that this behaviour is a result of social preferences guided by simple decision heuristics, rather than the rational examination of payoffs assumed by most social preference models. We also find a strong correlation between play in the Prisoner's Dilemma and in a subsequent Dictator Game, confirming previous findings suggesting a common prosocial motivation underlying altruism and cooperation.
Results
Cooperation was concentrated at three discrete levels across benefit-to-cost ratios. Higher b/c was associated with less zero cooperation and more maximal cooperation, while mean cooperation rose and PD cooperation was positively associated with later Dictator Game giving.
Results
- Across b/c values, cooperation clustered at giving nothing, half, or everything; overall, 22.4% gave nothing, 19.2% half, 52.3% everything, and 6.2% other amounts.Source: Figure 1 · p. 4, Results · p. 4
- As b/c increased, participants were less likely to give nothing and more likely to give everything, while the probability of giving half did not change significantly.Source: Results · p. 4
- Mean cooperation increased with b/c (coeff = .239, p = .003).Source: Figure 2 · p. 5, Results · p. 5
- PD cooperation and subsequent DG giving were positively associated (r = .561, p < .001), with few DG givers who did not cooperate in the PD.Source: Results · p. 5, Figure 3 · p. 6
Discussion
Cooperation was trimodally distributed in the one-shot continuous-strategy Prisoner’s Dilemma, and increasing b/c shifted participation away from zero and toward full cooperation. The persistent half-cooperation peak points to heuristic reasoning, while play also correlated with giving in a subsequent Dictator Game.
Discussion
- Cooperation was trimodally distributed, with peaks at zero, half, and full cooperation; increasing b/c reduced zero cooperation and increased full cooperation, but its influence was limited.The pattern was observed across the investigated b/c values.Source: Discussion · p. 6
- The 50% peak and modest b/c response depart from standard rational social-preference predictions; fairness or midpoint-anchoring heuristics could produce persistent half-giving.In this Prisoner’s Dilemma, giving half does not generally create equal outcomes because transfers are multiplied by b/c and the partner also chooses.Source: Discussion · pp. 6–7
- 24% of statements cited fairness, 15% expected the partner to give half, 22% described uncertainty, and 16% described compromise as reasons for half-giving.The authors interpret participants’ post-experimental descriptions as direct evidence that heuristics motivated 50% cooperation.Source: Discussion · p. 7
- The between-subjects design cannot show individual behaviour changes; consistency and contagion effects may also complicate clean measurement of payoff-structure effects.Population-level shifts could reflect personal b/c thresholds or within-person probabilistic changes, which the design cannot distinguish.Source: Discussion · pp. 7–8
- PD giving strongly correlated with later Dictator Game giving; DG givers were almost entirely a subset of PD cooperators, suggesting shared and additional PD motives.The authors suggest that motives such as inequity aversion occur in both games, while efficiency concerns or the partner’s choice may add motives in the PD.Source: Discussion · pp. 6, 8
- The study illuminates how heuristics implement social preferences, not how internalized norms may originate those preferences.The authors also report evidence of a domain-general proclivity to cooperate across games.Source: Discussion · p. 8
Methods
Participants completed a continuous-strategy Prisoner’s Dilemma followed by a Dictator Game, then described their choices. The between-subjects manipulation varied the PD’s transfer multiplier, and blinded assistants categorized free responses from half-givers.
Methods
- Participants were recruited through MTurk, received a $0.35 show-up fee, and earned an average of $0.47 in additional income.The $0.47 was average additional income based on the games.Source: Methods · p. 9
- In the first-stage continuous PD, each player received $0.10 and chose a transfer multiplied by k, with k=2, 3, 4, 5, or 10 and b/c=k.The multiplier varied between participants.Source: Methods · p. 9
- Participants had to pass payoff-comprehension questions; they made their PD choice before proceeding without learning their partner’s decision.The procedure was intended to ensure payoff understanding and prevent contagion effects from the partner’s PD choice.Source: Methods · p. 9
- In stage two, participants transferred any amount from a separate $0.10 endowment to a different worker in a standard Dictator Game, then completed free responses and a demographic survey.Dictator Game transfers were not multiplied, and recipients had no endowment or transfer decision.Source: Methods · p. 9
- 308 US residents participated after 140 people who failed comprehension questions were excluded; condition sizes were 66 (k=2), 56 (k=3), 60 (k=4), 61 (k=5), and 65 (k=10).Each participant was randomly assigned to one multiplier condition.Source: Methods · p. 9
- Two research assistants, unaware of the study hypotheses, coded half-givers’ free responses into eight categories describing their motivations.The categories included fairness, expectations about the partner, uncertainty, compromise, intuition, maximizing group payoff, restating the choice, and avoiding selfishness.Source: Methods · pp. 9–10