Metraipolis › archive › dc_2fd031e0193f
Language-model groups overstate consensus when replaying human deliberation on a reasoning task — arXiv:2609.20543
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.20543
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-10T03:16:07.916Z
- SHA-256 of the archived text
15890b989c310b6ebe8516f4809df198a27506a2bb67e179f193b014ea26ad2a- Stable URL of this page
https://metraipolis.com/documento/dc_2fd031e0193f
How to cite
Language-model groups overstate consensus when replaying human deliberation on a reasoning task — arXiv:2609.20543 — arXiv (en), https://arxiv.org/abs/2609.20543 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2fd031e0193f (Metraipolis archive dc_2fd031e0193f, sha256 15890b989c310b6ebe8516f4809df198a27506a2bb67e179f193b014ea26ad2a)
BibTeX
@misc{dc2fd031e0193f,
title = {Language-model groups overstate consensus when replaying human deliberation on a reasoning task — arXiv:2609.20543},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.20543},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2fd031e0193f; sha256 15890b989c310b6ebe8516f4809df198a27506a2bb67e179f193b014ea26ad2a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20543. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Full-consensus rates are often treated as indicators of collective cognition, yet depend on how participation and final states are operationalized. We replayed 100 held-out human Wason groups with matched large language model (LLM) agent groups, seeding one belief-anchored agent per participant's pre-discussion answer and scoring agents and people with the same code. Across human scoring definitions, estimates ranged from 24.0% to 57.0%; about one fifth of participants never posted, whereas agents almost always did. Agent groups remained more consensual in two post-unblinding sensitivity analyses: the submit-based comparison (n = 98) yielded gaps of 34.0 and 43.9 percentage points for chat and reasoning modes, and the participation-matched comparison (n = 45) yielded gaps of 34.1 and 44.4 points. These complementary routes reduced different measurement asymmetries yet converged within 0.5 percentage points. The gap persisted without early stopping and under a reparameterization removing the memorizable answer; reasoning-mode groups then agreed nearly unanimously, mostly on incorrect answers. Simulated consensus did not track collective accuracy, and belief-anchored agent groups were biased estimators of the human group-outcome distribution in this setting. These analyses provide a scoring-explicit basis for assessing simulated-group estimates of human deliberative outcomes. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.MA). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Computers and Society (cs.CY); Multiagent Systems (cs.MA). Pagina del lavoro: https://arxiv.org/abs/2609.20543 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-10T03:03:51.722Z dalla pagina https://arxiv.org/catchup/cs.MA/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).