The Problem with Puzzles

Published in Working Paper, 2026

How should a scientific community allocate its finite capacity for inquiry? The questions researchers ask condition what they can learn about the world. In political science, one widely invoked criterion for question selection is whether a question is motivated by a "puzzle," often understood as an empirical observation that contradicts prevailing expectations. This heuristic can generate important insights, but it also shapes inquiry in ways that are not fully appreciated. I argue that widespread reliance on this criterion creates a form of selection into inquiry. Questions expected to revise existing understanding receive attention, while questions expected primarily to confirm, refine, or increase confidence in existing knowledge are comparatively neglected. A discipline can accumulate an expanding collection of revisions without establishing how much confidence should be placed in the claims being revised. I illustrate this argument using a simple Bayesian model. The model demonstrates how puzzle- and surprise-driven research tends to privilege belief revision over the systematic refinement of existing claims---highlighting the distinction between expecting an answer and knowing it---with implications for how knowledge accumulates.

Recommended citation: Adriane Fresh. "The Problem with Puzzles." Working Paper.