Hossein Jorjani
First publication: 2026 – 06 – 30
AI systems, now and in the future, may function as components of an external cognitive infrastructure for scientific inquiry. Their potential roles include: (1) storing and retrieving data as an augmented external memory; (2) structuring and organizing recorded experience, functioning as an inference machine that extracts patterns and generates preliminary generalizations; (3) constructing dynamic “tables of discovery,” that is, systematically arranging relevant variables, observations, absences, exclusions, and possible experimental pathways; and (4) proposing targeted experiments or simulations that narrow the field of inquiry to interventions most likely to be informative.
In this sense, AI may extend several Baconian functions: memory, ordering, comparison, exclusion, and directed experimentation. Yet such systems do not remove the need for human judgment. Their outputs require transparent provenance, explicit assumptions, rigorous validation, and sustained human oversight. AI can help organize the path of discovery, but it cannot by itself guarantee the truth of what is discovered.
A useful criterion for evaluating an AI as an inference machine is not whether it reproduces previously observed answers, but whether it can generate coherent answers to previously unseen questions by preserving the structural relations that characterize a particular conceptual framework.
Hossein Jorjani 