Parental Trust in Automated Detection of Cyberpredators

Abstract

Online cyberpredators are a serious threat against children who are increasingly using social media and messaging systems to interact with strangers. At the same time, monitoring children's online activity is challenging for parents, due to the numerous platforms and ways a child can access them. Automated approaches could help detect dangerous conversations, but their adoption may prove difficult due to algorithm aversion - the tendency of humans to place less trust in recommendations by automated systems as compared to the judgment of other humans. Three online experiments investigate whether and under what conditions parents are willing to adopt automated systems for detection of cyberpredators, and when they are willing to trade potentially sensitive information about their children's online interaction, as well as individual control and agency. Factors tested for an effect on parental trust include accuracy of predictions by humans versus the algorithm, and storage and management of data in the cloud or local client. Implications for researchers and developers of automated systems of cybercrime detection are presented and discussed.