Why Even Basic A.I. Use Is So Bad for Students (www.nytimes.com)

Anastasia Berg, a philosophy professor at UC Irvine, reports that over half her large lecture class secretly used AI to write their final take-home exams, prompting her to warn that routine AI assistance is producing a “subcognitive” generation. Her argument goes beyond cheating: even seemingly benign uses — summaries, outlines, sentence polishing — short-circuit the repetitive, immersive linguistic practice students need to develop judgment, conceptual nuance and the capacity to follow and craft arguments. Berg traces the concern to a broader intellectual lineage (Plato’s worry about writing, debates over calculators) but insists this time the stakes include civic competence: a weakened facility with language undermines the ability to interpret news, medical forms or political claims and thus to participate in democracy. For the AI/ML community the piece is a practical challenge: language models that make summarization, drafting and stylistic editing trivial change the learning ecosystem. Technical implications include homogenized prose, diminished exposure to argumentative structure, and incentives for misuse. Responses Berg favors are pedagogical and policy-driven — tech-free spaces, clear incentives for original work — but the story also signals a need for product and research choices that support skill development: tool designs that scaffold learning (explain-this-step, graded hints), better detection and provenance, and collaboration with educators to integrate AI without replacing core cognitive practice.