Computer simulation can test hypotheses about large-scale social learning

Evidence

Over the past two decades, philosophers of science have taken models and ideas from fields such as economics and evolutionary biology and applied them to understand “epistemic communities”—that is, communities of people trying to gain knowledge about the world. Philosophers of science working with these models have tended to take communities of scientists to be their primary target, but as we argue later in the book, the models they have developed have much broader application: they can help us understand any community of people gathering evidence and sharing belief. … Large-scale social effects can be very difficult to study using observational or experimental methods. This is because the processes at work are widely distributed through time and space and often involve hundreds or thousands of people. It is hard to intervene or to ask questions about how things would have turned out under slightly different circumstances in the way scientists usually do. This is one reason that simulations and modeling can help: by building simple computer programs that can mimic groups of people sharing ideas, it is possible to test hypotheses about what sorts of factors can matter to how such groups of people learn. — Cailin O’Connor and James Owen Weatherall, The Misinformation Age, 47

Categorization

• XREF: The justification for agent-based models parallels the Milgrom game-theoretic approach — both turn social phenomena intractable to observation into tractable formal systems. — connects to: Even wildly unrealistic models can yield genuine insight into complex systems — Cailin O’Connor and James Owen Weatherall, The Misinformation Age: How False…, 176; Machine-learning proxies black-box complex social domains to enable intervention — Nick Couldry & Ulises A. Mejias, The Costs of Connection: How Data Is…, 450; Agent-based modeling reveals emergent group effects unpredictable from individual-level knowledge — Roderick M. Kramer, Geoffrey J. Leonardelli, Robert W. Livingston, Social Cognition, Social Identity, and…, 859 (+3 more)

Broader Connections

• Agent-based modeling reveals emergent group effects unpredictable from individual-level knowledge #corroborates (0.90) — Roderick M. Kramer, Geoffrey J. Leonardelli, Robert W. Livingston, Social Cognition, Social Identity, and…, 859
• Machine-learning proxies black-box complex social domains to enable intervention #expands (0.85) — Nick Couldry & Ulises A. Mejias, The Costs of Connection, 450
• Agents following identical optimal-distinctiveness rules can be simulated to test group formation #exemplifies (0.80)
• Pentland’s sociometer turned social interaction into instrumentable, predictively modelable data #exemplifies (0.78) — Shoshana Zuboff, The Age of Surveillance Capitalism, 1339