Out and Abo(u)t: A Multi-Method Study on the Cognitive Mechanisms of Human Efforts in Detecting Social Bots

Social bot research has focused mainly on automated detection, with less attention to how users judge suspicious accounts in real-world interface settings. This study examines human performance and underlying cognitive processes in evaluating nonpolitical social media accounts, with emotional arousal and Need for Cognition (NFC) as key predictors. Fifty-seven participants viewed civil or uncivil political debate clips to induce lower or higher arousal and then assessed 21 Twitter account screenshots varying in bot likelihood. Galvanic skin response measured arousal, and eye tracking captured heuristic versus systematic processing. Participants achieved 78% overall accuracy and identified likely human accounts more reliably than likely bot accounts. Although the arousal manipulation succeeded, arousal did not affect accuracy. Users relied primarily on heuristic processing, and higher arousal reduced visual attention to systematic cues. NFC increased viewing time but did not predict accuracy or processing mode. These findings inform HCI research and design for online authenticity judgments under uncertainty by highlighting attention, arousal, and cognitive effort as key considerations.