Go to OpenReview Archive homepageCognitive Modeling for Computational Epidemiology
Strategic response options to the COVID-19 pandemic have been greatly influenced by predictive epidemiological models. Effects of non-pharma- ceutical interventions (NPIs; such as mask wearing) unfortunately are based on an abundance of very large uncertainties around the extent to which the popula- tion adopts risk reducing behaviors. The effects of NPIs appear to have large heterogeneity across regions, subgroups, and individual mindsets and capabili- ties. We hypothesize that these uncertainties can be improved with higher-fidelity computational modeling of the social-psychological reactions of individuals, groups, and populations. We build up the ACT-R theory and Instance-Based Learning Theory to formulate psychologically valid agents and develop a frame- work that integrates multi-level cognitive and social simulation with information networks analysis, and epidemiological predictions. We present initial results from analyses of beliefs and sentiments about COVID-19 NPIs induced from online social media that can provide inputs to seed and validate cognitive agents. We present illustrations of cognitive model hypotheses about the dynamics of behavior change in response to intentions, attitudes, messaging, and source cred- ibility. We present an example of social networks propagating attitude change in response to NPIs.