The Deliberative Deficit: An Empirical Critique of LLMs in Democratic Discourse
Large language models (LLMs) are increasingly used for collective reasoning on complex, value-laden problems, yet their performance on non‑verifiable, pluralistic tasks cannot be fully inferred from standard benchmarks. Using the Deliberative Reason Index (DRI), the study evaluated 1,980 five‑agent LLM runs across 12 citizen‑assembly topics and found that LLM groups produce procedural discourse comparable to humans but show limited gains in intersubjective consistency, reduced perspective diversity, and a reversal of human convergence patterns. Engineering diversity through persona prompting …
Large language models (LLMs) are increasingly used for collective reasoning on complex, value-laden problems, yet their performance on non‑verifiable, pluralistic tasks cannot be fully inferred from standard benchmarks. Using the Deliberative Reason Index (DRI), the study evaluated 1,980 five‑agent LLM runs across 12 citizen‑assembly topics and found that LLM groups produce procedural discourse comparable to humans but show limited gains in intersubjective consistency, reduced perspective diversity, and a reversal of human convergence patterns. Engineering diversity through persona prompting …
The findings are relevant to Swiss AI because they highlight the need for rigorous, context‑specific evaluation of AI tools in democratic decision‑making, aligning with Switzerland’s emphasis on responsible AI deployment.
Original source record. Open the original record to verify the underlying announcement.