Clément Herman —
PhD Candidate in Economics

Groups, not just individuals, carry out search over candidates, products, and projects, yet standard approaches treat search as an individual activity. In laboratory experiments, we study groups that sequentially sample alternatives, varying stopping-decision rules and the alignment of group members' preferences, and compare them to individuals under identical conditions. Individuals undersearch, whereas groups systematically examine more alternatives, especially under stricter stopping rules. With aligned preferences, longer group search improves outcomes by raising reservation thresholds and smoothing variability in strategies, thereby increasing the rejection of low-value alternatives. With misaligned preferences, it has detrimental effects: high-value alternatives are rejected more often because agreement is harder to reach, even though participants lower their standards. The paper thus identifies sequential exploration (rather than deliberation or information aggregation) as a novel mechanism explaining why groups outperform or underperform individuals, and challenges modeling group search as if a single “unitary” decision maker were at work.

In many applications, agents can influence how data are generated but cannot manipulate the data themselves. For example, students choose which classes to take but cannot alter their grades, and firms decide on internal accounting practices but cannot change the resulting financial statements. This paper experimentally studies information transmission when data are generated through an unobservable, strategically chosen process. We focus on settings where an informed sender—such as a student or a firm—privately selects a data-generating process (DGP)—a portfolio of classes or an accounting practice—to shape the beliefs of an uninformed receiver, such as an employer or an investor. Across treatments, we vary which DGPs are feasible and whether some come at a cost. These variations span different levels of information verifiability and allow us to capture, within a unified framework, the core insights of disclosure, cheap-talk, and signaling models. Our findings reveal three main patterns. First, while senders select their DGPs strategically in line with theoretical predictions, receivers often fail to account for this strategic selection. Second, introducing differential costs across feasible DGPs mitigates this "DGP selection neglect." Third, while receivers’ biases are robust across environments, their consequences vary: information transmission falters most when the evidence receivers observe is highly sensitive to the sender’s DGP choices. Finally, increasing transparency about the selected DGP—a policy often advocated in practice—does not, in general, improve information transmission.

In many real-life situations—such as searching for jobs, houses, or spouses—individuals must explore heterogeneous options sequentially before making a choice. This paper investigates how individuals search and when they stop searching in environments with heterogeneous alternatives. Drawing on Weitzman’s (1979) optimal search theory, we design a large-scale experiment that systematically explores a broad universe of search problems, in which participants face sequential search tasks involving costly inspections of alternatives (boxes) with varying reward distributions and search costs. While participants generally follow the direction of optimal search—favoring alternatives with higher expected value, lower cost, or greater variance—we document systematic deviations. These include pervasive undersearching, negative responses to variance in approximately one-third of participants, and menu-dependent behavior that violates the index-based predictions of Weitzman’s rule. Using structural modeling, we identify distinct behavioral types. We also show that specific combinations of box characteristics give rise to different search patterns, with more complex menus triggering greater departures from optimal behavior. Our findings highlight both the power and the limits of classical search theory in explaining actual search behavior.

How do people form beliefs about environments too rich to hold in memory? This paper develops and tests a framework in which decision-makers reduce large datasets to a small set of representative "cases"—a mental model—from which beliefs are then formed. In a lab experiment, participants view datasets of 12 items with 3 attributes each, construct 1–3 "guessing aids" summarizing what they observed, and later use only these aids to predict a hidden attribute value for unseen items. Comparing behavior to an optimal clustering (k-medians) benchmark, case selection largely tracks coverage as clustering predicts, but participants over-rely on literally observed and duplicated cases. Predictions track the nearest retained case, but moderates away from distant or extreme values. These deviations do not overturn the clustering benchmark's qualitative predictions about which datasets and statistics are well versus poorly recovered—they exacerbate them, producing worse accuracy for noisy datasets, outliers, extreme values, and higher-order statistical relationships like conditional correlations.