Matthew Watts

Generative AI “companions” are increasingly integrated into higher education as always-available interlocutors that offer feedback, explanation, and encouragement. This article examines how such systems reshape the epistemic relations through which students speak, are heard, and come to count as knowers, and argues that companion-style AIs can disrespect epistemic relations in systematic ways. Students may become less able, and less willing, to testify on their own in genuinely reciprocal spaces (such as class discussions, office hours, and peer groups), habitually relying on their AI companions instead. Call this phenomenon testimonial outsourcing: a structural tendency to outsource one’s testimonial activity, which functions as a self-imposed, technologically mediated analogue of gaslighting. Students come to treat the AI’s outputs as more reliable than their own judgment and expressive capacities, thereby eroding their epistemic self-trust and cultivating a distorted conception of legitimate epistemic expectations. This article concludes by drawing implications for AI design and educational policy, placing constraints that preserve “non-mediated” spaces of speech, require explicit AI transparency, and build frictions that redirect students back toward interpersonal relations.

Semantic memory is often conceptualized as a storage space for an extensive assortment of explicit knowledge structures, formed as the result of a chronic, unconscious mechanism of abstraction and generalization. Against that, I argue that it is not the product of a dedicated system in which content is abstracted from experiences and stored, rather it’s best understood as split between what can be called mnemonic efforts and mnemonic habits. Mnemonic efforts are the effortful expression of semantic knowledge and often take the form of the determination of similarities among multiple episodic memory traces activated in parallel. Mnemonic habits on the other hand form as a result of fluency of constructive memory processes that can be broadly understood as embodied, or enacted.