We use slot machine games to study economic decision-making in complex high-uncertainty environments. We investigate the role of playful-exploratory and / or belief-based approaches to the game that may lead to suboptimal strategy choices compared to a situation where decision-makers play according to the Bayesian multi-armed bandit model. We explain this by a proposed model of expected-utility-maximizing agents choosing between a rational Bayesian and a curiosity-driven perspective.
A set of treatments capture and disentangle this deliberate rationality-curiosity trade-off: By changing key game features, such as the stakes, we change the opportunity cost of non-expected-utility-maximizing behavior. By offering additional, albeit payoff-neutral, choices, we allow players to engage in non-standard behaviors while playing according to the payoff-maximizing strategy at the same time.