As AI begins to match or surpass human researchers on core scientific tasks, two pressing questions arise: why humans should continue doing science, and how they should employ AI when they do. The paper argues that the second question can only be answered after the first is resolved.
One indispensable aim of science is a durable, collective human understanding of the world, pursued in service of human ends and values. From this principle the authors derive a two‑part test for delegating a research task to AI.
The first part, human understanding, requires that the scientific community remain able to explain, defend, and contest the resulting claims, and retain the skills to do so. The second part, human direction, demands that judgments about which questions matter and what findings mean stay with accountable humans.
Applying the test to each phase of the Contributor Roles Taxonomy (CRediT) yields recommendations that fields can tailor to their specific questions and methods. The framework differs from existing guidance in three ways: (1) it rests on an account of what science is for rather than on integrity norms; (2) it holds even if AI outperforms human researchers, not merely because of AI’s current limitations; (3) it evaluates each AI use by its impact on human understanding and direction, across every research phase—including peer review and training of new researchers—rather than by tool type.
The paper concludes with a call for reforms in governance, education, and evaluation systems to keep science aimed at human understanding and direction.
Review