AI-driven scientific discovery requires scientific maturity beyond model opacity

Abstract

AI-driven discovery does not create the historical fact that results can precede explanation. It changes how that interval is produced and governed. Systems such as AlphaFold, GNoME, and FunSearch can generate high-value outputs at a rate and scale that require communities to allocate attention, validation resources, and scientific credit before the outputs have been fully integrated into domain reasoning. Existing anomaly typologies classify unusual data patterns, while breakthrough typologies distinguish how discoveries arise and relate to prior knowledge. Neither directly organizes the subsequent transition from a generated output to accountable scientific knowledge. This article defines the explanation crisis as a conditional mismatch between accelerated output production and slower processes of epistemic integration, rather than as a universal feature or an unprecedented historical rupture. It develops scientific maturity as a graded, non-compensatory reporting framework across four dimensions: domain anchoring, representational integration, failure articulation, and communal accountability. A contrastive analysis of AlphaFold, GNoME, and FunSearch shows how output form and receiving-field practices shape different maturation pathways. The GNoME controversy further demonstrates why computational stability, novelty, and successful generation cannot be collapsed into scientific discovery. The framework moves inquiry beyond model opacity through four questions: whether an output is validatable, inferentially usable, corrigible through failure, and answerable within a scientific community.

The authors' abstract, as published at the source. Discover Artificial Intelligence, 2026 · DOI ↗