Social influence and information cascades in experience-based decisions

Many consequential decisions are made in environments where outcome probabilities are not explicitly described but must be inferred from experience. In most modern environments, this information search takes place in the presence of others. Yet, the role of social influence in experience-based decisions has been largely overlooked, leaving it unclear how personal and social information combine to shape what people sample, when they stop searching, and what they finally choose. Here, we address all three decision stages. In a preregistered study, participants could freely sample two options before making a final choice, either alone (solo condition) or while observing the final choices of their group members (group condition). We developed a modeling framework and show that social influence left sampling unaffected but altered stopping and choice. Participants searched longer when the group was more divided and stopped earlier when the majority choice agreed with their own personal sampling information. Consequently, they drew fewer samples and encountered fewer rare events than solo participants. When choosing, group members copied others' choices only when their personal information was ambiguous. Comparing decision sequences in groups with a solo baseline, complemented with agent-based simulations, showed that this strategy gave rise to mild information cascades, with decisions in groups being more similar to each other than in a solo baseline. Our results reveal how social influence shapes search, stopping, and choice, and how these individual-level decision rules scale up to collective dynamics.