Temporal epistemology in the age of AI: learning, knowing and the reorganization of chronological structure

Purpose This paper argues that generative artificial intelligence (AI) has forced a reckoning with a long-standing inadequacy in the temporal assumptions that have structured educational thought since the Enlightenment. It introduces post-linear epistemology as a framework for understanding how knowledge functions when learning is no longer adequately modelled as movement through a temporal sequence and develops concrete implications for curriculum design, assessment and teacher practice. Design/methodology/approach The paper is conceptual. It synthesizes scholarship on complexity in education, non-linear learning, technological acceleration, futures studies and critical artificial intelligence in education and uses this synthesis to specify the conditions under which a post-linear account of educational knowledge becomes necessary. Findings The paper identifies three interrelated shifts that, on this account, characterize the post-linear condition and are proposed here as claims for further examination rather than as established findings. First, chronologically ordered curriculum sequences are argued to be no longer adequate as the default architecture of educational design – though not as the only legitimate one. Second, cumulative derivation is argued to be increasingly displaced, in AI-mediated contexts, by probabilistic and relational forms of knowledge production. Third, temporal literacy – the capacity to identify when information was produced, how it has drifted and what time-frame it presupposes – is proposed as a distinct educational capacity, with three specifiable components, that these conditions make newly necessary. Research limitations/implications The paper is conceptual, not empirical. The claims about how recursive AI use affects student work and about whether temporal literacy can actually be taught and measured are presented as hypotheses for design-based research, not as findings. The framework is meant to make those questions easier to investigate, not to answer them. Practical implications Three changes follow. Curriculum design should look at which prerequisite sequences are actually necessary and which are inherited habit. Pedagogy should teach temporal literacy directly, using provenance critique, drift mapping and positioning essays, instead of treating it as a by-product of other work. Assessment should measure how students locate and situate knowledge, not just what they can reproduce. Social implications AI-generated text routinely blends sources from different periods into a single fluent voice. That makes the question of whose perspectives get amplified, and whose get smoothed away, a public concern rather than a specialist one. Temporal literacy is one of the capacities through which an ordinary reader can interrogate AI-mediated text for what it has left out. Originality/value Other accounts of non-linear learning (complexity theory, connectivism, postdigital thought) already cover much of what AI intensifies. None of them was built for one specific thing: AI generation flattens the temporal markers that normally tell sources apart. This paper picks out that specific feature, names the literacy needed to read against it and turns the conceptual claim into concrete recommendations for curriculum, assessment and teacher practice.