When AI Changes its Tone, Does Acceptance Follow?

Abstract

The integration of artificial intelligence (AI) into decision-making processes raises numerous questions about acceptance, particularly in hospital environments marked by strong professional identities. This working paper presents an ongoing experimental study that explores how the conversational tone of an AI agent might influence the acceptance of its recommendations, depending on the psychological profile of the healthcare professional. Building on a 2x2x2 model structured around three dimensions of ego (personal value, perceived competence and social role), the study aims to demonstrate that adapting the tone can significantly improve acceptance to algorithmic recommendations. The experimental protocol involves profession-specific scenarios in which participants assess a series of AI-generated messages, each introduced during the option evaluation stage of the decision-making process. By proposing a novel identity-based approach to AI design, this study seeks to contribute to both theory and practice. It is expected to open perspectives for the development of conversational agents capable of dynamically adjusting their tone to enhance acceptability and support professional autonomy in hospital settings.