Reconsidering Meaningful Learning in a Bandit Experiment on Weighted Voting: Subjects’ Search Behavior

This paper clarifies subjects’ search behavior of correct options behind the experimental results shown by Guerci et al. (Theory Decis 83:131–153, 2017). In the experiment, subjects were asked to choose one of two weighted voting games repeatedly and their payoffs are stochastically determined for each of their choice according to a payoff-generating function that was hidden from subjects. The main results are as follows. (1) In the additional sessions conducted for the treatment without any payoff-related feedback information, it was reconfirmed that subjects learned to choose the correct option that generates higher expected payoffs for them and generalized what they had thought introspectively in a binary choice problem to a similar but different one. (2) Feedback information about payoffs given immediately after subjects’ choice often confused their inference on the relationship between nominal voting weights and actual payoffs so that they took the win-stay-lose-shift strategy in some sessions. (3) Immediate payoff-related feedback information sometimes induced subjects to randomly choose the runs of options.