Confusion and reinforcement learning in experimental public goods games (original) (raw)
Related papers
Confusion and learning in the voluntary contributions game
2013
We use a limited information environment to assess the role of confusion in the repeated voluntary contributions game. A comparison with play in a standard version of the game suggests, that the common claim that decision errors due to confused subjects biases estimates of cooperation upwards, is not necessarily correct. Furthermore, we find that simple learning cannot generate the kind of contribution dynamics commonly attributed to the existence of conditional cooperators. We conclude that cooperative behavior and its decay observed in public goods games is not a pure artefact of confusion and learning.
Equilibrium Play and Learning in a Stochastic Public Goods Experiment
In the game we study, the benefit received from a public good is not linear but varies stochastically with effort. We experimentally examine the effect of a quality bound under which group members receive negligible benefits from the public good, if the bound is not met. Our theory predicts that the presence of such a “pass mark” on quality may under certain conditions reduce the incidence of free-riding. We obtain very mixed support for this hypothesis using both mixed strategy as well as quantal response equilibrium benchmarks. In contrast, a reinforcement-learning model does better at predicting laboratory behaviour.
Revisiting confusion in public goods experiments
2000
Theories put forth to explain cooperation in public goods experiments usually assume either that subjects cooperate because they do not understand the game's incentives or that cooperation stems from social motives such as altruism and reciprocity. Recent research by Andreoni is an important first attempt to distinguish the roles of confusion and social motives in these environments.
Learning, signaling, and social preferences in public-good games
2006
ABSTRACT. This study compares the empirical performance of a variety of learning models and theories of social preferences in the context of experimental games involving the provision of public goods. Parameters are estimated via maximum likelihood estimation. We also performed estimations to identify different types of agents and distributions of parameters. The estimated models suggest that the players of such games take into account the learning of others and are belief learners.
Ignorance is not Always Bliss: Feedback and Dynamics in Public Good Experiments
SSRN Electronic Journal, 2000
In this paper we study the effects of providing additional feedback about individual contributions and earnings on the dynamics of contributions in a repeated public good game. We include treatments where subjects can freely choose whether to obtain additional information about individual contributions or individual earnings. We find that, in the aggregate, contributions decline less fast when additional information about contributions and earnings is provided on top of aggregate information. We also find that there exist substantial but intuitively appealing differences in the way individuals react to feedback. Particularly, individuals with a high propensity to contribute tend to imitate the highest contributor more often and are more inclined to obtain feedback about individual contributions than about individual earnings than individuals with a lower propensity to contribute.
Experimental Evidence from a Sequential Public Good Game
2008
We run a series of experiments in which subjects ha ve to choose their level of contribution to a pure public good. The design differs from the standard p ublic good game with respect to the decision proced ur . Instead of deciding simultaneously in each round, s ubjects are randomly ordered in a sequence which di ffers from round to round. We compare sessions in which s ubjects can observe the exact contributions from ea rlier decisions ("sequential treatment with information") to sessions in which subjects decide sequentially but cannot observe earlier contributions ("sequential treatmen t without information"). The results indicate that sequentiality increases the level of contribution to the public g ood when subjects are informed about the contributi on levels of lower ranked subjects while sequentiality alone has no effect on contributions. Moreover, we observe t hat earlier players try to influence positively the contributio ns of subsequent decision maker...
Feedback and dynamics in public good experiments
2012
In this paper we study the effects of providing additional feedback about individual contributions and/or earnings on contributions and the dynamics of contributions in a repeated public good game. We include treatments where subjects can freely choose whether to obtain additional information about individual contributions or individual earnings. We find that, in the aggregate, contributions are lower when feedback on earnings is provided compared to when feedback on contributions is provided. We also find that there exist substantial but intuitively appealing differences in the way individuals react to feedback. Particularly, individuals with a high propensity to contribute tend to imitate the highest contributor more often and are more inclined to obtain feedback about individual contributions compared to individuals with a lower propensity to contribute.► We study the effect of different types of feedback (contributions and/or earnings) on contributions and dynamics in a public good experiment. ► We find that including feedback about past earnings of group members decreases contributions and induces individuals to imitate the lowest contributor relatively more often as compared to the highest contributor. ► We find that the effects are driven by cooperative types who change the way they imitate.
Journal of Economic Psychology, 2009
In the recent literature, several hypotheses have been put forward in order to explain the decline of contributions in repeated public good games. We present results of an experiment which allows to evaluate these hypotheses. The main characteristics of our experimental design are a variation of information feedback and an elicitation of individual beliefs about others' contributions. Altogether, our data support the hypothesis of conditional cooperation with a selfish bias.
Journal of Economic Psychology, 2010
This study analyses the behavior in a repeated public goods game when subjects know about the possibility of existence of strict conditional cooperators. We employed a baseline treatment and a threat treatment in which subjects are informed about the possibility of being in a group together with automata playing a grim trigger strategy. We conjecture the resulting game allows for almost fully efficient outcomes. Contributions in the threat treatment increase by 40% before a surprise restart, and by 50% after the surprise restart. In line with the grim trigger strategy subjects contribute either all or nothing in the threat treatment.
2021
The strategy method is often used in public goods games to measure individuals’ willingness to cooperate depending on the level of cooperation by others (conditional cooperation). However, while the strategy method is informative, it risks being suggestive and inducing elevated levels of conditional cooperation that are not motivated by concerns for fairness, especially in uncertain or confused participants. Here we make 845 participants complete the strategy method two times, once with human and once with computerized groupmates. Cooperation with computers cannot rationally be motivated by concerns for fairness. Worryingly, 69% of participants conditionally cooperated with computers, whereas only 7% conditionally cooperated with humans while not cooperating with computers. Overall, 83% of participants cooperated with computers, contributing 89% as much as towards humans. Results from games with computers present a serious problem for measuring social behaviors.