Key points are not available for this paper at this time.
Recent advancements in Generative AI (GAI) systems now demonstratehuman-like capabilities, presenting new opportunities in higher education mentorship. This study investigates college students’ ability to differentiate between mentor responses that were either AI-generated or human-authored, and how their perceptions influenced evaluations of helpfulness, caring, and willingness to engage again across social, personal, and academic domains. Participants (N = 147) rated and identified concealed-source responses from ChatGPT and faculty mentors. Results showed that participants were significantly less accurate at identifying AI-generated responses in the personal domain. Participants who perceived responses as human, regardless of the actual source, rated them higher across all domains, highlighting the activation of perceptual biases. Introducing the Perceptual Bias Activation (PBA) framework, we posit how these biases may be activated and influence behaviour. Addressing these biases is crucial for effectively integrating AI tools into education to complement traditional mentorship and enhance student learning. © 2026 Informa UK Limited, trading as Taylor & Francis Group.
Lee et al. (Thu,) studied this question.