The Laboratory for Collective Behavior

CoBe Lab

What we study

At the University of Chicago’s Collective Behavior (CoBe) Lab, we study how individual actions following simple local rules produce complex collective behaviors and how those collective behaviors, in turn, reshape the individuals who produce them.

Our work spans human and artificial agents, combining online experiments, large-scale behavioral data, and computational modeling to understand how people and AI systems learn from each other, form and revise beliefs, and coordinate at scale.

Across these efforts, we aim to advance theories of collective intelligence and complex systems: how, when, and why groups become wiser than their individual members, when they fail, and how the social and psychological systems mediating their interaction can be designed to support fairer, better-informed collective behavior.

Principal Investigator

Dr. Henry Dambanemuya

Assistant Instructional Professor

University of Chicago

Meet the research team →

Dr. Dambanemuya conducts theory-driven investigations of collective behavior in human and artificial agent populations across social and psychological systems.

His research draws on online experiments, surveys, crowdsourcing, and behavioral trace data, combined with computational methods (e.g., natural language processing, machine learning, and social graph mining) to study social influence, belief updating, and collective intelligence.

Current projects

Emergent influence and belief formation

When people argue their way toward a shared belief, two things shape where they land: the structure of who talks to whom, and where each person draws their information from (e.g., experts, majorities, elders, crowds). Two observations motivate this work. First, network structure and source effects are typically studied apart. Second, in real communication networks, messages rarely arrive labeled as an "expert opinion" or a "majority opinion"; these attributions are usually unobserved, and people might infer whom to trust from whom others already pay attention to. Influence, in other words, is not necessarily assigned but can and sometimes does emerge from network position. Two complementary paths follow from this.

The Network Scientist’s Path

Building on findings that network position predicts persuasiveness and is reinforced by it (Dambanemuya et al., 2024), and that imitation improves decisions only when those imitated have a genuine track record ( Dambanemuya et al., 2023), we trace how influence emerges in social networks, and when it might steer an unsuspecting crowd toward optimal collective outcomes rather than amplifying the opinions of the most vocal or loudest individuals.

Broad ImpactsSocial media platforms that mediate our public discourse are constantly deciding whose voices to amplify, yet emergent influence may track genuine expertise or merely reward early, loud, or well-positioned individuals. Understanding when network position rewards superior collective outcomes rather than noise speaks directly to how the online platforms that we interact with daily might surface trustworthy voices instead of entrenching the opinions of whoever happens to dominate a conversation.

The Experimentalist’s Path

We bring network structure and source effects research together to ask how attributed normative and descriptive information behaves once it enters a deliberating network. Our research design crosses network structure (centralized vs. decentralized), source attribution (experts, majorities, elders, crowds), and message type (injunctive vs. descriptive norms) to ask when deliberation turns reference information into accuracy rather than entrenchment on contested social, economic, or political issues.

Broad ImpactsIf source and framing determine whether a network amplifies or repairs collective error, then the same intervention can depolarize one group and entrench another. Identifying the conditions under which deliberation converts reference information into accurate collective beliefs, rather than noise or entrenchment, speaks directly to how the interventions deployed on contested issues (e.g., fact-checks, norm corrections, expert labels) might actually help reduce misperceptions rather than amplify them.