We study changes in Twitter users' behavior associated with interacting with automated bots on the platform. Based on the list of malicious bots identified and shut down by Twitter in the wake of 2016 US presidential election, we are able to identify about 54 thousand human Twitter users who have interacted with automated accounts. We first establish the baseline pattern of user behavior in the period before interaction and then measure behavioral changes observed around the period of interaction. Using a quasi-experimental research design, we document economically and statistically significant quantitative and qualitative changes in users' behavior and discuss their implications.

No abstract