Algorithmic Optimisation or Relational Pedagogy? AI, Human Connection and the Formation of Self in the Learner-Teacher Relationship

This paper examines what a developing sense of self depends on when artificial intelligence (AI) is positioned within the learner-teacher relationship. AI-in-education research has emphasised personalisation, efficiency and human oversight, but has rarely engaged the developmental-psychology evidence that teaching is constituted through relational security, co-regulation and mentalisation rather than merely accompanied by them. Synthesising work across human-centred AI, developmental psychology and participation theory, the paper distinguishes a cognitive-support layer AI can occupy from a relational one it cannot; inference and synthesis of affective state approximate the outward form of attunement without the mind-reading and responsiveness that relational security depends on. Evidence that adults track identity continuity in AI companions and respond to its disruption with reactions characteristic of relationship loss suggests these self-formation stakes are not speculative when a system is placed where a relationship should be. The paper is theoretical rather than empirical, and its conclusions are provisional; it argues that AI may support thinking about teaching without constituting the relational work through which a learner's sense of self is formed, cautioning against both technological determinism and reflexive dismissal of AI's support functions.