Most people use AI by asking it a question and judging its answer. But what if the AI pushed back? That is the idea behind Collaborative Human-AI Dialectics for Scientific Research, a new project led by Dr Bradley P. Allen of the Intelligent Data Engineering Lab (INDElab) at the University of Amsterdam. Together with the Alliance for Data Science and AI (ADSA), Vrije Universiteit Amsterdam and Old Dominion University, the project uses a system called Elenchus to explore whether AI can help scientists make their reasoning clearer, more precise and more consistent. The project is supported by the Alfred P. Sloan Foundation.

Scientists often need to turn specialist knowledge into formal descriptions that computers can work with. A biodiversity researcher, for example, may need to define exactly what counts as a biome, how it relates to habitats and species, and which exceptions matter. These structured definitions are called ontologies and help connect datasets, compare findings and make research more reproducible. Definitions can, however, contain hidden assumptions or lead to conclusions that do not quite fit together. Elenchus is designed to uncover those weak spots: an expert enters a claim or definition, and the AI flags a possible tension. The expert stays in control and can reject the challenge, revise the claim or explain why the apparent conflict is not a problem. Only decisions endorsed by the expert become part of the final knowledge record. The name refers to the Socratic method.

To test this, thirty biodiversity researchers will complete comparable tasks using a conventional free-form AI chat and using Elenchus’s structured dialogue, with the same person, task and language model. Independent experts will assess the results without knowing which method produced them. The researchers do not assume that more disagreement will automatically lead to better science. If structured challenge improves the quality of knowledge, it could inform the design of future AI systems for research. If not, that result is equally valuable, as it would show where an ordinary AI chat may be just as effective. Elenchus is being developed as open-source software.