When AI Explains Its Decision, Humans May Stop Thinking Independently
CIO, Thursday, August 20th, 2026
Research finds LLM-generated rationales suppress productive disagreement and push evaluators to reject good ideas.
CIO reports on research testing AI's influence on humans reviewing innovation proposals. Researchers found that AI recommender tools were persuasive enough to convince evaluators to reject decisions made by independent human experts.
The effect caused evaluators to pass on promising innovations they would otherwise have advanced. Adding LLM-generated rationales to recommendations suppressed productive human disagreement rather than informing it.
The finding matters for any organization inserting AI recommendations into a human review process on the assumption that explanations improve judgment.