Please login

How does testosterone influence decisions and choice-related neural computations in competitive environments? To address this question, we administered testosterone or placebo to female participants (n = 54) in a double-blind, randomized design. Following drug treatment, participants competed in a dot estimation task that manipulated opponent status (lower, equal, or higher) and outcome feedback (win or loss), after which they decided whether to compete against the same opponent again. All participants adjusted their behavior based on opponent status and outcome feedback. Participants who received testosterone, however, showed significantly greater sensitivity to outcome feedback: they were more willing to compete after winning and less willing after losing, and made those decisions faster - suggesting that testosterone increases the weighting of immediate, salient outcome information in competitive decision-making. At the neural level, a network comprising ventral striatum, vmPFC, bilateral TPJ and ACC processed outcome-related signals during the feedback period. Critically, neural prediction analyses at the trial-level revealed that activity in left ventral striatum and TPJ predicted subsequent decisions to compete, but only in participants who received testosterone. The direction of these effects mirrored the behavioral results: striatal activity amplified the tendency to re-compete after winning, whereas TPJ activity predicted renewed competition after losing. Together, these findings demonstrate that testosterone biases competitive decisions by amplifying the influence of outcome-related activity in reward and social cognition circuits.

Mitigating climate change requires urgent shifts toward sustainable consumption and production. While ecolabels have been widely studied as cost-effective tools for influencing consumer behavior, their effects on producers remain largely unexplored. We test the effect of ecolabels on sustainable production behavior in a dynamic common-pool resource dilemma with real economic and environmental incentives. In our lab experiment, groups of participants trade off short-term individual profits from resource extraction against long-term group-level gains and resource preservation. We find that ecolabels significantly increase sustainable production and extend resource longevity, improving both long-term environmental and economic outcomes. They facilitate early coordination on sustainable behavior which is then maintained through conditional cooperation. Importantly, these effects arise even when ecolabels are purely symbolic and provide no monetary benefit. Our findings uncover underexplored producer-side mechanisms of ecolabels and suggest that, beyond consumer nudges, ecolabels can serve as powerful policy instruments for promoting sustainable production.

How do people trade off individual versus group welfare in the face of uncertainty regarding private benefits of different actions? We propose a partial information revelation ('recommendation') policy designed to maximize group welfare, and we show its theoretical robustness to well-documented behavioral deviations from the risk neutral, Bayesian, and self-interested benchmark. In a large-scale online experiment with 2600 subjects, we then show that this policy fails to improve upon a full information benchmark even when individual and group objectives are aligned, as the recommended course of action is not followed often enough. In a setting where individual and group interests clash, the recommendation is followed less often, largely by subjects who misunderstand the policy. This provides suggestive evidence in favor of simplicity in information design in multi-agent strategic settings.

Amidst the growing popularity of AI language models, we study the potential impacts of AI-facilitated communication on trust and trustworthiness. With a laboratory experiment, we compare traditional communication with communication assisted by ChatGPT in a between-subject design. We find that participants with access to Chat-GPT more frequently make promises, while their promise-keeping rate diminishes. Overall we do not observe effects on trust and trustworthiness rates. However, we show that in the GPT treatment participants coordinate less often on the trust outcome and that promises no longer serve as a reliable indicator of honesty.

We study public persuasion in elections with binary outcomes, like referendums. In our model, one or multiple senders attempt to influence the election outcome by manipulating public information about a payoff-relevant state. We allow for a wide class of sender preferences, ranging from pure self-interest to a broad set of social welfare functions, including utilitarian preferences. Our main result identifies a single-crossing property that ensures the optimality of censorship policies, which reveal intermediate states while censoring extreme states. This holds in large elections under both monopolistic and competitive persuasion. The single-crossing property holds for all self-interested senders and more generally under a mild condition for the distribution of voters’ preferences. We analyze how a sender’s optimal censorship policy changes with his preferences and voting rules, and characterize the asymptotic properties as the electorate size goes to infinity. Finally,our results shed new lights on whether media competition maximizes voter welfare.