Ongoing efforts to combat the global pandemic of COVID-19 via public health policy have revealed the critical importance of understanding how individuals understand and react to infection risks. We here present a model to explore how both individual observation and social learning are likely to shape behavioral, and therefore epidemiological, dynamics over time. Efforts to delay and reduce infections can compromise their own success, especially in populations with age-structure in both disease risk and social learning - two critical features of the current COVID-19 crisis. Our results concur with anecdotal observations of age-based differences in reactions to public health recommendations. We show how shifting reliance on types of learning affect the course of an outbreak, and could therefore factor into policy-based interventions.

28Early analysis of outbreaks of novel pathogens to evaluate their likely public health 29 impact depends on fitting predictive models to data gathered and updated in real-time. 30Both transmission rates and the critical 0 R threshold (i.e. the pathogen's 'reproductive 31 number') are inferred by finding the values that provide the best model fit to reported 32 case incidence. These models and inferred results are then the basic tools used for public 33 health planning: how many people expected to be infected, at what scales of time and 34 space, and whether potential intervention strategies impact disease transmission and 35spread. An underlying assumption, however, is that the ability to observe new cases is 36 either constant, or at least constant relative to diagnostic test availability. 37 We present a demonstration, discussion, and mathematical analysis of how this 38 assumption of predictable observability in disease incidence can drastically impact model 39 accuracy. We also demonstrate how to tailor estimations of these parameters to a few 40 examples of different types of shifting influences acting on detection, depending on the 41 likely sensitivity of surveillance systems to errors from sources such as clinical testing 42 rates and differences in healthcare-seeking behavior from the public over time. Finally, 43we discuss the implications of these corrections for both historical and current outbreaks. 44 45

New COVID-19 diagnoses have dropped faster than expected in the United States. Interpretations of the decrease have focused on changing factors (e.g. mask-wearing, vaccines, etc.), but predictive models largely ignore heterogeneity in behaviorally-driven exposure risks among distinct groups. We present a simplified compartmental model with differential mixing in two behaviorally distinct groups. We show how homophily in behavior, risk, and exposure can lead to early peaks and rapid declines that critically do not signal the end of the outbreak. Instead, higher exposure risk groups may more rapidly exhaust available susceptibles while the lower risk group are still in a (slower) growth phase of their outbreak curve. This simplified model demonstrates that complex incidence curves, such as those currently seen in the US, can be generated without changes to fundamental drivers of disease dynamics. Correct interpretation of incidence curves will be critical for policy decisions to effectively manage the pandemic.