New Behavioral Science Approach Combines Experiments and Computational Models
Summary: Researchers outline a combined method using controlled experiments and computer-based models to better understand group and organizational behavior.
Researchers from North Carolina State University and Northwestern University describe an integrated approach that links laboratory experiments with agent-based and other computational models to generate deeper insights into how groups and organizations behave.
Behavioral science has long relied on both controlled experiments and computational modeling, yet these tools are rarely used together in a systematic way. William Rand, a computer scientist and assistant professor of business management at NC State’s Poole College of Management, and Ned Smith, an associate professor of management and organizations at Northwestern University’s Kellogg School of Management, are co-authors of a paper that explains how combining these methods can sharpen our understanding of macro-level social patterns that emerge from micro-level human behavior.
How the combined approach works
The method begins with well-designed laboratory experiments that probe how individuals or small groups make decisions under specific conditions. These experiments reveal the decision rules, heuristics, and response tendencies that people use when faced with a task or social situation. Instead of stopping at laboratory findings, researchers can encode these empirically observed micro-level behaviors into a computational model—often an agent-based model—that simulates many interacting agents over time and at larger scales.
Once the model incorporates experimentally derived rules, it can be used as a predictive testbed to explore how those micro-level behaviors scale up. For example, the model can simulate thousands of individuals interacting in varied environments, revealing macro-level patterns such as organization-wide coordination, information diffusion, or emergent inequalities. Model outputs can point to conditions under which groups adopt one strategy over another, or highlight long-run consequences that are impossible or impractical to observe directly in the lab or in the field.
Iterative cycle between experiments and models
The strength of this approach lies in iteration. Experimental data inform model construction; the model generates hypotheses and predictions about large-scale behavior; those predictions motivate further experiments that test competing mechanisms or boundary conditions. This iterative loop helps researchers refine both theory and measurement, improving confidence that lab-observed mechanisms actually drive broader social outcomes.
For instance, laboratory work might show that groups tend to resolve a problem using either collaborative discussion or competitive individual effort. A calibrated model can then simulate many groups across different network structures, incentive schemes, or information environments to predict when one approach will dominate. If the model predicts that dense social networks favor collaboration while sparse networks favor competitive strategies, researchers can design follow-up experiments that manipulate network connectivity to test the prediction.
Practical relevance and applications
This mixed-methods strategy has practical implications across management, public policy, and any domain where group behavior matters. Computational models provide a platform to test how sensitive lab results are to contextual factors and to explore policy interventions that could steer social systems toward better outcomes. For organizational leaders and policymakers, this means insights drawn from controlled human-subject studies can be assessed for robustness and scalability before costly real-world implementation.
By combining empirical rigor with computational exploration, the approach helps bridge the gap between small-scale behavioral studies and the complex dynamics observed at organizational or societal scales. It enables researchers to adjudicate between competing explanations, estimate effect magnitudes beyond the lab setting, and surface long-run implications of subtle micro-level tendencies.
Abstract (concise)
Simulating Macro-Level Effects from Micro-Level Observations. Integrating agent-based modeling (ABM) with lab-based human experiments leverages the strengths of both methods. Lab experiments reveal causal relationships and behavioral mechanisms at the individual level, while ABM uses explicit decision rules to reproduce and explore macro-level empirical patterns. The combination benefits modelers by providing empirically validated mechanisms and benefits experimenters by offering tools to assess validity, compare competing mechanisms, and examine long-run, system-level consequences of micro-level observations. The paper illustrates this mixed-method approach with examples related to status, social networks, and job search, and it outlines methodological guidance for future integration efforts.
About this research
Source: North Carolina State University, reporting on collaborative work with Northwestern University.
Publisher: Organized coverage by NeuroscienceNews.com. Image credited as public domain.
Original research: The paper is titled “Simulating Macro-Level Effects from Micro-Level Observations” and appears in a peer-reviewed management science outlet. It presents both conceptual arguments and an applied example demonstrating how lab-derived behavioral rules can be scaled via simulation to generate testable macro-level predictions.
This combined experimental-and-modeling approach offers a practical roadmap for researchers who want to translate detailed laboratory findings into reliable, scalable insights for organizational decision-making and public policy design.