Artificial intelligence, the knowledge problem, and academic judgment: an Austrian perspective

Recent advances in generative artificial intelligence have renewed claims that university teaching and research can be largely automated. This article argues that such claims confuse information processing with judgment. Drawing on Hayek’s theory of dispersed and tacit knowledge, Huerta de Soto’s analysis of computation and entrepreneurial discovery, and the theory of judgment under Knightian uncertainty, it explains why greater computational capacity does not eliminate the knowledge problem. Contemporary AI can retrieve, recombine, and generate large volumes of context-sensitive information, while memory and retrieval mechanisms allow earlier interactions to shape later outputs. These capabilities expand the range of plausible explanations and possible courses of inquiry, but they do not provide a purpose-independent criterion for determining which interpretation is relevant, which assumptions are defensible, or which course of action should be pursued. The article applies this distinction to higher education. Artificial intelligence is likely to substitute for some standardized and information-intensive academic tasks, including routine content delivery, retrieval, and preliminary synthesis. At the same time, it increases the relative importance of judgment-intensive functions: diagnosing misunderstanding, selecting worthwhile research questions, evaluating methods and evidence, mentoring students, and assigning responsibility for scholarly claims. The argument is therefore not that every existing university practice will persist unchanged, but that artificial intelligence shifts the comparative advantage of professors and universities toward the cultivation, evaluation, and institutional coordination of judgment under uncertainty.