What motivates human behaviour in social dilemmas? The results of public goods games are commonly interpreted as showing that humans are altruistically motivated to benefit others.  However, there is a competing ‘confused learners’ hypothesis: that individuals start the game either uncertain or mistaken (confused), and then learn from experience how to improve their payoff (payoff-based learning).  We: (1) show that these competing hypotheses can be differentiated by how they predict contributions should decline over time; and (2) use meta-data from 237 published public-goods games to test between these competing hypotheses. We find, as predicted by the confused learners hypothesis, that contributions declined faster when individuals have more influence over their own payoffs. This prediction arises because more influence leads to a greater correlation between contributions and payoffs, facilitating learning. Our results suggest that humans, in general, are not altruistically motivated to benefit others, but instead learn to help themselves.

For the Simulation analyses: Conducted entirely in R studio.

For the Black box experiment: a behavioural economics experiment on human participants at the Centere for Experimental Social Sciences (CESS), Nuffield College, Oxford. The data are the output files from the experimental software (z-Tree). Data are analyzed in R studio. Ethics and consent forms are included also.

For the Comparative Regression data: We searched for relevant studies in three ways. First, we searched the “Web Of Knowledge” database. In May 2014, we searched with the phrases “public good* game*”, and “voluntary contribution mechanism”. In October 2017 we searched for additional articles between 2014-2017 inclusive, with the phrase “public good* game*”, refined by TOPIC: “experiment” AND “voluntary contribution mechanism”, and with the phrase “repeated public good* game*” refined by TOPIC: “experiment”. Second, we searched for suitable papers cited in three reviews on social dilemmas. Third, we looked for other papers cited in the papers that we had found.

Our dependent/response variable was the mean percentage contribution, to one decimal place, per round. When studies presented their data in graphical form we extracted the average level of contribution with WebPlotDigitizer  (Rohatgi, A. (https://automeris.io/WebPlotDigitizer/, 2017).

The article has three separate investigations: a) a simulation, b) an experiment (black box), and c) a comparative regression.

The associated files are stored in three zip folders, a, b and c., with Read_Me files.