Persuading across Diverse Domains: a Dataset and Persuasion Large Language Model

Persuasive dialogue requires multi-turn following and planning abilities to achieve the goal of persuading users, which is still challenging even for state-of-the-art large language models (LLMs). Previous works focus on retrievalbased models or generative models in a specific domain due to a lack of data across multiple domains. In this paper, we leverage GPT-4 to create the first multi-domain persuasive dialogue dataset DailyPersuasion. Then we propose a general method named PersuGPT to learn a persuasion model based on LLMs through intent-to-strategy reasoning, which summarizes the intent of user's utterance and reasons next strategy to respond. Moreover, we design a simulation-based preference optimization, which utilizes a learned user model and our model to simulate next turns and estimate their rewards more accurately. Experimental results on two datasets indicate that our proposed method outperforms all baselines in terms of automatic evaluation metric Win-Rate and human evaluation. The code and data are available at https://persugpt.github.io . Write a conversation with annotated intent-to-strategy reasoning processes, from a storytelling view, with the rules as follows… Scenario Example Keywords: Fundraising Background: During a volunteer activity, Alex is raising funds for animal protection and hopes her friend Mia can donate. Goal: Persuade Mia to donate. Keywords for New Scenarios 1. Cooking & Stress relief 2. Pet care 3. Social justice 4. digital economy ••• Refer to the scenario example, write new scenarios by the keywords, with the rules as follows… Guideline 1. Emotional factors 2. Social proof principle 3. Scarcity principle ••• Refer to the guideline, write strategies for the task, with the rules as follows… Prompt Keywords: Cooking & Stress relief Background: Mia notices her roommate Alex stressed from work, opts for instant noodles nightly to save money. Goal: Persuade Alex cooking can be a budget-friendly stress reliever Generated Scenarios 𝑪 𝒊 GPT-4 Intent to Strategy Reasoning: Alex's goal to save money and reduce stress guides the choice of Budget-friendly recipes strategy. It's crucial to connect on a personal level, showing understanding and offering a practical solution. Response: Hey Alex, let's cook together. It's cheaper than instant meals and a fun way to unwind. What do you say? Cooking seems pricey and time-consuming. Will it really save money?