Adam Khurshid

Between Humans and AI: Authorship Verification and AI Detection for Cheating

Abstract

The undetectable use of AI to generate text has become a catastrophe. Despite the existence of AI detection tools, humans can still pass AI text as undetected using humanizers. Making the detector incorrectly recognize AI generated text as human generated. To minimize this catastrophe, we propose using authorship verification to determine if multiple writings have been produced by the same author. Rather than just examining if the literature is AI generated, we will contrast it with content we are aware has been produced by the supposed human writer. In this research, we test if author verification can detect AI generated text, and humanized AI generated text better than traditional AI detection. Our research outcome indicates that author verification is mostly not affected by humanizers while AI detection is negatively affected.