I am a fourth-year PhD candidate at Caltech. I study how to measure preferences and use them to personalize interventions.
I design theory-driven experiments that pin down what drives individual decisions and test the theories meant to explain them. My goal is to turn individual-level measurement into the basis for better-targeted policy.
Experimental economists routinely use incentivized tasks to elicit unobservables like preferences and beliefs — but these measurements are often a means to tailor later interventions, not an end in themselves. This paper asks whether anticipating those downstream consequences distorts how people respond to elicitation in the first place. I run an experiment in which subjects report a valuation in a BDM task for an induced-value object, followed by a binary choice problem. The treatment varies the link between the BDM report and the choice problem alternatives: in the control, the report does not affect the alternatives; in two treatments, it shifts the alternatives so the optimal report lies above or below the induced value. On average, reports move as predicted but fall short of the optimum. Two findings help explain this. First, when the BDM imposes a cost for misreporting, subjects retreat toward truthful reporting across all treatments. Second, a questionnaire separating understanding of the BDM, the binary choice, and their integration shows that different sources of characterization failure drive differences in responses. Together, these results show that anticipated consequences do shape elicited preferences, but the response is heterogeneous and only partially optimal; thus, the truth is not easily recovered even when downstream consequences are common knowledge.