ABSTRACT Proposals for the adoption of artificial intelligence (AI) in education are often accompanied by claims that it promotes ‘personalization’. However, the concept of personalization has been narrowly and ambiguously framed. This paper traces the genealogy of the notion of personalization in education, highlighting its roots in Anglo‐Saxon political‐economic frameworks that redefined citizens as ‘consumers’ of services, contrasting it with the Ibero‐American personalist tradition that emphasizes autonomy, dignity and integral human flourishing. While early 21st‐century policy initiatives and contemporary AI‐driven platforms promote what we term ‘Personalization 4.0’, they reproduce the limitations of modern instrumental rationality: reducing cognition to computation and freedom to choice‐optimization. Building on phenomenological and hermeneutic critiques of modern epistemology, the paper examines how AI personalization may foster depersonalization rather than deep personalization. We identify five paradoxes that reveal the tension between its promises and outcomes: (1) expanding horizons while narrowing knowledge through correlational standardization; (2) simulating uniqueness while fostering homogenization; (3) enabling personalization at the cost of consented surveillance; (4) serving as both oar and crutch, enhancing autonomy for the skilled while undermining development for novices; and (5) promising equity while deepening divides through the Matthew Effect. To address these tensions, we propose distinguishing between extrinsic personalization (externally provided adaptive conditions and tools) and intrinsic personalization (learner‐driven processes of agency, metacognition and self‐regulation). While AI contributes effectively to extrinsic personalization, it cannot substitute the intrinsic dimension, which remains the foundation of deep personalization. Genuine educational personalization requires their integration, yet current trends risk undermining intrinsic processes. The conclusion is that we need to depart from rationalist/modernist notion of rationality in order to fully understand how AI can contribute to deep personalization. Otherwise, there's a chance that current narratives, even when they claim to foster personalization, might actually hinder its true development.
Bellomo et al. (Fri,) studied this question.