The Quadruple Loss of Academia: Two Pressures, Four Losses in AI-Accelerated Science

This paper treats four visible losses in academia as outcomes of two interacting pressures. The first is breadth pressure: questions that cross or precede disciplinary boundaries are difficult to evaluate, and the communities, criteria, journals, and careers that make them evaluable can pull inquiry toward local specialization. The second is throughput pressure: AI can increase hypothesis generation, analysis, and writing faster than human verification and judgment can scale, increasing pressure for automation and triage. Prior research has separately documented the evaluation of interdisciplinarity, field formation and specialization, verification bottlenecks in AI-driven science, and AI-induced illusions of understanding and epistemic convergence. Placed in one dynamic, these literatures reveal a further possibility: the repairs used to make breadth evaluable and rising throughput processable can generate second-order losses and can feed back across axes. In particular, author-level screening adopted to contain verification load can raise entry friction for solo, new, and non-traditional explorers, turning a throughput-side repair into a breadth loss. As a constructive counter-path, the paper develops distributed discovery with staged validation, using AI as a scaffold that connects heterogeneous explorers to literature and downstream expert verification while preserving human judgment.