Human accuracy in social network analysis: The role of network characteristics in infection risk assessment

Understanding an individual’s human network assessment is fundamental for successful policy design when dealing with infection risk. Yet, how individuals cognitively compensate for their inability to process complex network models, remains poorly understood. Using an online best-worst ranking experiment, we investigate individuals’ perception of the risk of connecting to carefully constructed COVID-19-infected networks of 697 Dutch participants. Specifically, we consider three network characteristics that individuals can either directly access or implicitly estimate in real-life social network assessment. We show that the perceived infection risk in social networks is not solely based on the objective probability of this risk: easily assessable physical characteristics have stronger predictive power on the perceived risk than the objective probability. Heterogeneity assessment suggests that demographic differences such as the level of education interact with the order and strength of the network characteristics. The often-complex mental calculation underlying objective risk in networks is substituted by a heuristics-driven approach. Our findings facilitate more tailored, effective, and generalizable policy and economically optimal infection-mitigating campaign design.