usually collect information to serve specific goals andoften end up with samples that are unrepresentative of the un-derlying population. This can introduce biases on later judg-ments that generalize from these samples. Here we show thatgoals influence not only what information we collect, but alsowhen we decide to terminate search. Using an optimal stop-ping analysis, we demonstrate that even when learners have nocontrol over the content of a sample (i.e., natural sampling),the simple decision of when to stop sampling can yield sampledistributions that are non-representative and could potentiallybias future decision making. We test the prediction of thesetheoretical analyses with two behavioral experiments

Sampling biases are often assumed to arise from the type of information that learners sample (Fiedler, 2008), the possibility of negative payoffs (Denrell, 2001), or the prevalence of small samples (Kareev et al., 2002). Here, we show that even in a natural sampling situation (repeated Bernoulli trials), in which a learner’s only decision is when to stop sampling, different sampling goals can have an impact on sample composition and on inferences drawn from them. Specifically, we find that learners sampling with a binary goal (”more heads/tails?”) versus a distributional goal (”how many heads?”) end up with samples that differ not only in size but also content. Binary sampling leads to more samples with extreme distributions (many more heads or tails) compared to distributional sampling. In this project, we explore the impact of those sampling goals on subsequent decision-making on the basis of those samples.

People often make decisions with the goal of gaining information which can help reduce their uncertainty. However, recent work has suggested that people sometimes do not select the most diagnostic information queries available to them. A critical aspect of information search decisions is evaluating how obtaining a piece of information will alter a learner’s beliefs (e.g., a piece of information that is redundant with what is already known is useless). This suggests a close relationship between information seeking decisions on one hand, and belief updating on the other. This paper explores the deeper relationship between these two constructs in a causal intervention learning task. We find that patterns in belief updating biases are predictive of decision making patterns in tasks where people must make interventions learn about the structure of a causal system.

What is the best way of figuring out the structure of a causalsystem composed of multiple variables? One prominent ideais that learners should manipulate each candidate variable inisolation to avoid confounds (known as the “Control of Vari-ables” strategy). Here, we demonstrate that this strategy is notalways the most efficient method for learning. Using an opti-mal learner model which aims to minimize the number of tests,we show that when a causal system is sparse, that is, whenthe outcome of interest has few or even just one actual causeamong the candidate variables, it is more efficient to test mul-tiple variables at once. In a series of behavioral experiments,we then show that people are sensitive to causal sparsity whenplanning causal experiments.