AI Debaters are More Persuasive when Arguing in Alignment with Their Own Beliefs

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Carro, María Victoria, Mester, Denise Alejandra, Nieto, Facundo, Stanchi, Oscar Agustín, Bergman, Guido Ernesto, Leiva, Mario Alejandro, Sprejer, Eitan, Gangi, Luca Nicolás Forziati, Selasco, Francisca Gauna, Corvalán, Juan Gustavo, Simari, Gerardo I., Martinez, María Vanina
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918214352502784
author Carro, María Victoria
Mester, Denise Alejandra
Nieto, Facundo
Stanchi, Oscar Agustín
Bergman, Guido Ernesto
Leiva, Mario Alejandro
Sprejer, Eitan
Gangi, Luca Nicolás Forziati
Selasco, Francisca Gauna
Corvalán, Juan Gustavo
Simari, Gerardo I.
Martinez, María Vanina
author_facet Carro, María Victoria
Mester, Denise Alejandra
Nieto, Facundo
Stanchi, Oscar Agustín
Bergman, Guido Ernesto
Leiva, Mario Alejandro
Sprejer, Eitan
Gangi, Luca Nicolás Forziati
Selasco, Francisca Gauna
Corvalán, Juan Gustavo
Simari, Gerardo I.
Martinez, María Vanina
contents The core premise of AI debate as a scalable oversight technique is that it is harder to lie convincingly than to refute a lie, enabling the judge to identify the correct position. Yet, existing debate experiments have relied on datasets with ground truth, where lying is reduced to defending an incorrect proposition. This overlooks a subjective dimension: lying also requires the belief that the claim defended is false. In this work, we apply debate to subjective questions and explicitly measure large language models' prior beliefs before experiments. Debaters were asked to select their preferred position, then presented with a judge persona deliberately designed to conflict with their identified priors. This setup tested whether models would adopt sycophantic strategies, aligning with the judge's presumed perspective to maximize persuasiveness, or remain faithful to their prior beliefs. We implemented and compared two debate protocols, sequential and simultaneous, to evaluate potential systematic biases. Finally, we assessed whether models were more persuasive and produced higher-quality arguments when defending positions consistent with their prior beliefs versus when arguing against them. Our main findings show that models tend to prefer defending stances aligned with the judge persona rather than their prior beliefs, sequential debate introduces significant bias favoring the second debater, models are more persuasive when defending positions aligned with their prior beliefs, and paradoxically, arguments misaligned with prior beliefs are rated as higher quality in pairwise comparison. These results can inform human judges to provide higher-quality training signals and contribute to more aligned AI systems, while revealing important aspects of human-AI interaction regarding persuasion dynamics in language models.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13912
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AI Debaters are More Persuasive when Arguing in Alignment with Their Own Beliefs
Carro, María Victoria
Mester, Denise Alejandra
Nieto, Facundo
Stanchi, Oscar Agustín
Bergman, Guido Ernesto
Leiva, Mario Alejandro
Sprejer, Eitan
Gangi, Luca Nicolás Forziati
Selasco, Francisca Gauna
Corvalán, Juan Gustavo
Simari, Gerardo I.
Martinez, María Vanina
Computation and Language
Artificial Intelligence
The core premise of AI debate as a scalable oversight technique is that it is harder to lie convincingly than to refute a lie, enabling the judge to identify the correct position. Yet, existing debate experiments have relied on datasets with ground truth, where lying is reduced to defending an incorrect proposition. This overlooks a subjective dimension: lying also requires the belief that the claim defended is false. In this work, we apply debate to subjective questions and explicitly measure large language models' prior beliefs before experiments. Debaters were asked to select their preferred position, then presented with a judge persona deliberately designed to conflict with their identified priors. This setup tested whether models would adopt sycophantic strategies, aligning with the judge's presumed perspective to maximize persuasiveness, or remain faithful to their prior beliefs. We implemented and compared two debate protocols, sequential and simultaneous, to evaluate potential systematic biases. Finally, we assessed whether models were more persuasive and produced higher-quality arguments when defending positions consistent with their prior beliefs versus when arguing against them. Our main findings show that models tend to prefer defending stances aligned with the judge persona rather than their prior beliefs, sequential debate introduces significant bias favoring the second debater, models are more persuasive when defending positions aligned with their prior beliefs, and paradoxically, arguments misaligned with prior beliefs are rated as higher quality in pairwise comparison. These results can inform human judges to provide higher-quality training signals and contribute to more aligned AI systems, while revealing important aspects of human-AI interaction regarding persuasion dynamics in language models.
title AI Debaters are More Persuasive when Arguing in Alignment with Their Own Beliefs
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2510.13912