Biased preferences through exploitation: How initial biases are consolidated in reward-rich environments.

In the current article, we test the prediction that an initial bias favoring 1 of 2 equally rewarding options-either based on a genuine contingency or a pseudocontingency in a small sample of initial observations-can survive over an extended period of further sampling from both options, when the reward structure fosters exploitation. Specifically, we argue and demonstrate that in reward-rich environments where two options predominantly-but equally frequently-yield positive outcomes, the initial bias should be upheld because exploitation of the allegedly superior option reinforces the biased preference. In contrast, in reward-impoverished environments, where both options yield predominantly negative outcomes, initial biases can be expected to be eradicated through exploration, which increases the chance of recognizing the equality of the initially nonpreferred option. In 3 experiments, initial evidence in a guided-sampling phase was set up for participants to perceive an actual contingency (Experiment 1) or infer a pseudocontingency (Experiment 2a and b) that made 1 option look more rewarding. In a subsequent free-sampling phase this led to a sustained bias toward this option when the environment contained mostly positive but not when it contained mostly negative outcomes. We argue that biased sampling in reward-rich environments could be responsible for false beliefs about the outcomes of behavioral options, and as such could be relevant to a broad range of topics including social interactions or health contexts. (PsycInfo Database Record (c) 2020 APA, all rights reserved).