A latent dimension of Condorcet's jury theorem for multiple AI advisers

arXiv:2609.14438v1 Announce Type: new Abstract: When the same question is asked of multiple AI advisers, as in self-consistency and LLM-as-a-judge panels, Condorcet's jury theorem predicts that adding independent, competent advisers makes the majority more reliable. The theorem, however, has a latent dimension when viewed from the user's vantage: adding advisers also makes disagreement more visible. A binomial model reveals that this ``visible dissent'' becomes nearly inevitable as the number of advisers grows, and that reliability and disagreement both approach certainty but at different convergence rates. The two rates cross at an adviser accuracy of 4/5 (0.8). Below this value, visible dissent approaches certainty faster than reliability and, with enough advisers, becomes more likely than a correct majority. Even ideal panels of independent and competent advisers can be correct in aggregate but appear divided; such disagreement does not by itself indicate aggregation failure. The way advisers split also provides a common basis for predictive multiplicity, reconciliation load, and reliance miscalibration. These results indicate two distinct decisions when using multiple AI advisers: how many advisers to consult and how their verdicts should be presented and interpreted.

Theorem analysis shows visible dissent and reliability convergence rates differ.

This research reveals that aggregation of multiple AI advisers increases reliability but also visible disagreement, which can mislead users about decision correctness. This dual effect implies a nuanced social impact: potential confusion or mistrust balanced by improved accuracy. The findings do not imply direct harm or benefit but highlight challenges in interpreting AI advice ensemble outputs.

This is a mathematical/theoretical contribution about interpreting multi-adviser AI systems, not an empirical finding about real-world deployment or societal effects. It offers useful conceptual clarity for designing trustworthy AI systems but makes no direct claim about societal outcomes.

This paper presents a theoretical model about how the reliability of AI adviser panels and the visibility of their disagreement converge at different rates. It identifies challenges in interpreting such panels but does not claim any specific societal impact, either positive or negative.

The paper presents a mathematical model showing that using multiple AI advisers increases visibility of dissent. This can lead to more transparent and reliable AI decision-making. The findings suggest practical decisions for optimizing AI adviser usage.

SOURCE     arXiv cs.CY (Computers and Society) (tier 1)
   ↓
DOCUMENT   666daef2-3aa7-4ef9-aeeb-99b8b7c976ef
           https://arxiv.org/abs/2609.14438
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -24  (Adverse)
   ↓
CONFIDENCE 83%