Can Humans Detect the Authenticity of Social Media Accounts?

This study investigates the influence of verbal and non-verbal cues on people’s credibility judgments of fake Twitter profiles generated by an information hiding mobile app solely for transmitting secret messages. We tested the hypotheses that the trustworthiness conveyed by the profile picture, morality-related trait adjectives included in the profile summary and the profile owner’s gender would increase people’s credibility judgments of those fake Twitter profiles. 24 participants assessed 16 fake profiles on their credibility. They also expressed their confidence in their credibility judgements and they answered an open-ended question on which parts of the profile influenced their credibility judgements. The results showed that overall participants did not trust the Twitter profiles. Furthermore, confidence judgements were higher when profiles included competence-related traits in the profile summaries. Verbal rather than non-verbal cues had thus more influence on participants’ judgements. The openended responses revealed a large reliance on the content of the profile, which is what the mobile app relies on. We discussed these findings in light of the relative lack of credibility of the profiles generated by the mobile app. The new insights can help improve designs of systems depending on automated social media accounts and will provide useful clues about other applications where cognitive computing plays a role