Answers are grounded in this paper and cite their sources.
Agreement has traditionally occupied a central position in conversational system design, particularly in systems optimized for user satisfaction, conversational coherence, collaborative problem-solving, and socially acceptable responses. However, the ability to explicitly express disagreement represents an equally important, yet conceptually distinct, interaction mechanism. A conversational participant may reject another participant's proposition, challenge an interpretation, oppose a recommendation, question an inference, or maintain an alternative position without necessarily initiating a formal debate or attempting to persuade the other party. Although existing research has investigated disagreement in human communication, user disagreement with large language models, sycophancy, multi-agent debate, and opinion dynamics among artificial intelligence agents, these research areas have generally examined different aspects of disagreement within separate theoretical and application contexts. This paper introduces the Explicit Disagreement Interaction Framework (EDIF), a unified theoretical framework that conceptualizes explicit disagreement as an independent, directional, and multidimensional interaction mechanism across human–AI, AI–AI, and human–human communication. The framework establishes a participant-oriented classification encompassing user-to-AI disagreement, AI-to-user disagreement, reciprocal human–AI disagreement, AI-to-AI disagreement, human-to-human disagreement, and multi-participant disagreement involving heterogeneous combinations of human and artificial intelligence participants. Beyond participant directionality, EDIF establishes a multidimensional analytical structure incorporating initiation, explicitness, evidential justification, disagreement intensity, persistence, reciprocity, participant multiplicity, positional stability, position modification, disagreement targets, and interaction outcomes. The framework further distinguishes disagreement from counterargument, debate, contradiction, criticism, and refusal, thereby establishing clearer conceptual boundaries for future research. To the best of our knowledge, this paper proposes the first unified, cross-participant, direction-sensitive theoretical framework specifically dedicated to explicit disagreement as a fundamental interaction mechanism across human–AI, AI–AI, and human–human communication. Rather than treating disagreement as a conversational failure or merely a component of debate, this study positions it as an independent interaction capability that may support epistemic independence, intellectual diversity, error identification, decision quality, and more transparent communication. The paper provides formal definitions, a conceptual model, research propositions, an evaluation methodology, and a research agenda for subsequent empirical investigation.
No takes yet. Share an insight, caveat, or question.
Jincheng Zhang (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: