Abstract: Human inter-activity intervals often exhibit a power-law regularity. Such a robust regularity may be generated by an underlying mechanism. Previous statistical models have successfully captured this burstiness. However, most of them lack behavioral interpretaAon. In this study, we propose a novel two-stage neural autopilot model. The model combine the habitual choice and goal directed mode. The neural autopilot model not only fits the empirical data well but also offers deeper behavioral insights. The behavioral interpretation offered by the neural autopilot algorithm deepens our understanding of the mechanisms driving human activity and informs future policy considerations on how to effectively encourage—or avoid inducing—such behavior.

Abstract: Whether and under what conditions the strategy method creates behavioral distortions in the elicitation of choices in sequential games remains mildly understood in experimental methodology. In this paper, we compare behavior in six optimally designed centipede games, implemented under three different choice elicitation methods: the direct response method, the reduced strategy method and the full strategy method. These methods elicit behavioral strategies, reduced contingent strategies, and complete contingent strategies, respectively. We find significant behavioral differences across these elicitation methods—differences that cannot be explained by standard game theory, but are consistent with the predictions of the Dynamic Cognitive Hierarchy (DCH; Lin and Palfrey 2024) solution, combined with quantal responses.

Abstract: One of the most prevailing theories about female underrepresentation in academia, particularly in STEM, is the existence of recruiting bias against women, and gender-blind evaluations are often recommended to prevent such discrimination. Meanwhile, after a reform of Taiwan’s college admission, more programs implemented non-gender-blind evaluations (including application portfolio reviews, interviews, and others). Taking the adoption of non-gender-blind evaluations as a natural experiment, this study examines its impact on female college admission. The empirical result indicates that moving from fully gender-blind evaluations to fully non-gender-blind evaluations raises the female percentage of admitted applicants by 5.54 percentage points. Moreover, this pro-women effect surges up to a 10.06 percentage point increase in majors not directly linked to subjects taught in high school. This can be due to the interaction between non-gender-blind evaluations and gender differences in learning styles for unfamiliar subjects. This research contributes to the debate over which interventions can be a remedy to enhance women’s participation in fields where they are outnumbered, by shifting away from one-size-fits-all solutions and focusing on uncovering the most effective approach in particular circumstances.

Abstract: Nash equilibrium in mixed strategies incorporates equalized payoff and serial independence for undominated strategies. We document strong evidence that pitchers in Major League Baseball not only throw fastball-series pitches too often but over-switch between them and breaking balls, not complying with mixed strategies in the pitch-type selection. Moreover, overconfidence and tiredness, rather than commonly assumed experience, are two underlying behavioral mechanisms explaining such deviations. In particular, power pitchers deviate more and fatigued pitchers alternate pitch type less. To address the selection bias, propensity score matching confirms main results persist. Our empirical evidence enriches the real-life mixed strategy applicability.

Abstract: This systematic and large-scale reproduction effort tests the reproducibility and robustness of economics and political science. We reproduced and conducted robustness analysis of 110 articles recently published in leading economics and political science journals. We found that over 85% of published claims were computationally reproducible. In robustness checks, our re-analyses lead to 72% of statistically significant estimates to remain significant and in the same direction, and the median reproduced effect size is the same as the originally published effect size (that is, 100% of the published effect size). Fourth, six independent research teams examined 12 pre-specified hypotheses about determinants of reproducibility and robustness. They found a negative relationship between reproducers’ experience and reproducibility, but no relationship between reproducibility and author characteristics or data availability.

Abstract: This paper describes analyses of data from the Zearn K-5 math educational learning platform that were used to inform/motivate/inspire causal interventions, which were then tested with many other interventions in a mega-study. These analyses purely correlate student outcomes with teacher Zearn activity but are then hypothesized to possibly have causal effects. One analysis reduces the many kinds of teacher activity to a smaller number of dimensions (1-3) using principal-component analysis. The most diagnostic dimension is how often teachers absorb student failures and successes, which we term “empathy”. A second analysis finds that when teachers access Zearn on Fridays their students’ learning is higher in the future. An intervention in the mega-study based on empathy, from the principal-component analysis, is one of the most successful of many interventions. The Friday engagement-prompt works positively but less well. We draw conclusions from this example about how candidate nudges can be motivated by careful analysis of data in advance of nudge design.