Why Computational Cognitive Science? Specifying Social Science Theories as Cognitive Agents

Social science theories about how people and environments change each other are mostly stated in words. Verbal statements leave open the quantities and functional forms needed to derive outcomes. Tests of association cannot distinguish the individual rules that could have produced an association, and causal structures established separately are rarely combined. Computational cognitive science offers a remedy: it specifies theories as rules that generate individual judgment and choice, runs the rules, and checks before data collection whether rival rules can be distinguished. Cognitive agent modeling adapts this practice to theories of person–environment relations. Four causal structures recur in social science: the environment changes people, people select their environments, the two change each other over time, and environmental effects differ across people. Cognitive agent modeling specifies these structures as four components of one agent: learning, choice, feedback, and individual differences. The article compares the approach with other computational approaches and works through media effects theories as a case, giving an example specification and four propositions. In a simulation, three rival explanations of crime viewing and fear yielded positive mean correlations, and a design alternating assigned and freely chosen exposure identified the generating specification in 96% to 100% of data sets.