Can AI Serve as a Mentor?

Embodiment and Testimonial Authority in the Age of Algorithms

Abstract

Some students in higher education turn to AI for mentoring support. Yet whether AI can genuinely serve as a mentor has received little scholarly attention. The existing research on AI in education has largely concentrated on instrumental effectiveness, with ethical concerns drawing considerably less scrutiny. An ontological analysis, however, remains absent altogether: can AI fulfill the fundamental conditions of mentoring in the first place? We argue that AI encounters two fundamental ontological limitations as a mentor. Drawing on the phenomenology of Merleau- Ponty, Pauline body theology, and Ricoeur’s hermeneutics of testimony, we argue: first, the absence of embodied co-presence precludes the intercorporeality essential for the transmission of tacit knowledge; second, the absence of an existential biography eliminates testimonial authority, which requires a readiness for vouching with one’s own being. Despite these limitations, students can perceptually interpret AI as a mentor. Drawing on Gibson’s concept of affordances and the Perceptual Bias Activation framework (Lee and Esposito 2025), we argue that AI generates false affordances. Constant availability can be misread as relational presence. Emotionless tone can be mistaken for impartial judgment. Adaptive responses can be perceived as genuine understanding. Our analysis reveals that AI produces a simulacrum of mentoring, not a technologically mediated relationship, but its simulation. For academic practice, this implies that universities should position AI as a knowledge assistant rather than a mentor substitute, in order to safeguard the formative dimension of higher education.