We investigate the potential for Large Language Models (LLMs) to enhance scientific practice within experimentation by identifying key areas, directions, and implications. First, we discuss how these models can improve experimental design, including improving the elicitation wording, coding...

Eligibility criteria for interventions can induce an Ashenfelter Dip, and subsequent mean-reversion may result in improvement over time even absent the intervention. We investigate these dynamics for a food-as-medicine program to treat diabetes, where eligibility required elevated hemoglobin A1c...

We study the problem of estimating the average causal effect of treating every member of a population, as opposed to none, using an experiment that treats only some. We consider settings where spillovers have global support and decay slowly with (a generalized notion of) distance. We derive the...

Remote work policies remain controversial because of the perceived opportunity for increased shirking outside of the traditional office; a problem that is potentially exacerbated if employees work in a revenue-sharing team environment. Using a controlled experiment, where individuals are...

How does gender composition influence individual and group behavior? To study this question empirically, we assembled a new, national sample of United States city council elections and digitized information from the minutes of over 40,000 city-council meetings. We find that replacing a male...

Field experiments provide the clearest window into the true impact of many policies, allowing us to understand what works, what does not, and why. Yet, their widespread use has not been accompanied by a deep understanding of the political economy of their adoption in policy circles. This study...

In organizations, teams are ubiquitous. "Weakest Link" and "Best Shot" are incentive schemes that tie a group member's compensation to the output of their group's least and most productive member, respectively. In this paper, we test the impact of these incentive schemes by conducting two pilot...

We introduce DOSE--Dynamically Optimized Sequential Experimentation--to elicit preference parameters. DOSE starts with a model of preferences and a prior over the parameters of that model, then dynamically chooses a customized question sequence for each participant according to an...

We provide twelve best practices and discuss how each practice can help researchers accurately, credibly, and ethically use Generative AI (GenAI) to enhance experimental research. We split the twelve practices into four areas. First, in the pre-treatment stage, we discuss how GenAI can aid in...

We measure individual-level loss aversion using three incentivized, representative surveys of the U.S. population (combined N=3,000). We find that around 50% of the U.S. population is loss tolerant, with many participants accepting negative-expected-value gambles. This is counter to earlier...