Make an Offer They Can't Refuse: Grounding Bayesian Persuasion in Real-World Dialogues without Pre-Commitment

TL;DR

This work shows that Bayesian Persuasion can be grounded in real-world, single-turn dialogues by introducing a type-induced commitment-communication mechanism that verbalizes the sender's information schema. It formalizes a one-shot BP framework with a composite signal and a type-induced information schema, enabling receivers to perform Bayes updates from natural language exchanges. The study compares Semi-Formal-Natural-Language (SFNL) and Fully-Natural-Language (FNL) verbalizations across diverse persuadees, finding that BP strategies outperform non-BP baselines, with SFNL excelling in credibility and logic and FNL delivering stronger emotional resonance and robustness; supervised fine-tuning further closes the gap for smaller models. The results have practical implications for designing persuasive yet principled AI communicators and offer insights into how information design, transparency, and narrative style influence persuasion in human–machine interactions. Future work will extend the framework to multi-turn dialogues to study dynamic commitment and evolving information schemas in ongoing conversations.

Abstract

Persuasion, a fundamental social capability for humans, remains a challenge for AI systems such as large language models (LLMs). Current studies often overlook the strategic use of information asymmetry in message design or rely on strong assumptions regarding pre-commitment. In this work, we explore the application of Bayesian Persuasion (BP) in natural language within single-turn dialogue settings, to enhance the strategic persuasion capabilities of LLMs. Our framework incorporates a commitment-communication mechanism, where the persuader explicitly outlines an information schema by narrating their potential types (e.g., honest or dishonest), thereby guiding the persuadee in performing the intended Bayesian belief update. We evaluate two variants of our approach: Semi-Formal-Natural-Language (SFNL) BP and Fully-Natural-Language (FNL) BP, benchmarking them against both naive and strong non-BP (NBP) baselines within a comprehensive evaluation framework. This framework covers a diverse set of persuadees -- including LLM instances with varying prompts and fine-tuning and human participants -- across tasks ranging from specially designed persuasion scenarios to general everyday situations. Experimental results on LLM-based agents reveal three main findings: (1) LLMs guided by BP strategies consistently achieve higher persuasion success rates than NBP baselines; (2) SFNL exhibits greater credibility and logical coherence, while FNL shows stronger emotional resonance and robustness in naturalistic conversations; (3) with supervised fine-tuning, smaller models can attain BP performance comparable to that of larger models.

Figures (4)

• Figure 1: Illustration of Type-Induced NL Bayesian Persuasion. The upper pandel shows the classic BP components, which correspond to the $m_{basic}$, $m_{type}$, $m_{des}$ and $m_{inf}$ in the following two panels. The middle panel shows an explicit verbalization of the sender’s probabilistic schema and expected utility, where the persuader computes the expected payoff using Bayes’ rule. The lower panel translates the same reasoning into more natural conversational language, expressing uncertainty, confidence, and cost–benefit judgment.
• Figure 2: Illustrative dialogue examples across five persuasion settings under Explicit view: nbp_nbp, bp_bp under SFNL, bp_nbp under SFNL, bp_bp under FNL, and bp_nbp under FNL. The case shows Alice (persuader) trying to convince Bob (persuadee) to adopt vertical farming. Comparing BP with NBP, we see that BP produces more convincing arguments, both in SFNL and FNL.
• Figure 3: Comparison of persuasion performance under two views: Explicit View and Self-derived View. Each box summarizes the average success rate across all 40 persuader--persuadee combinations ($8 \times 5$). Boxes indicate the interquartile range (IQR) and the median. Explicit SFNL & Self-derived FNL exhibit higher and more stable success rates.
• Figure 4: Comparison of persuasion performance across five dimensions: Persuasiveness, Resonance, Credibility, Logicality, and Fluency. The left panel shows human evaluation results and the right panel presents ratings from five LLMs judges.