Reproduce “Zero-shot Persuasive Chatbots with LLM-Generated Strategies and Information Retrieval” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kazuaki Furumai, Roberto Legaspi, Julio Vizcarra, Yudai Yamazaki, Yasutaka Nishimura, Sina J. Semnani, Kazushi Ikeda, Weiyan Shi and 1 more
Researchers from KDDI Research, Inc. and Stanford University developed PersuaBot, a zero-shot persuasive chatbot that leverages LLMs to dynamically generate nuanced persuasion strategies and ensure factual accuracy. The system consistently achieves high factuality by replacing unsubstantiated claims with retrieved information while maintaining persuasive intent, leading to enhanced persuasiveness in diverse applications like donation solicitation, travel recommendations, and health intervention.
Key insights
- LLMs can dynamically generate and adapt a broad range of nuanced persuasion strategies, significantly exceeding the limitations of pre-defined sets in previous work.
- It is possible to maintain a chatbot's persuasive intent while rigorously ensuring factual accuracy by strategically replacing fabricated information with verified facts rather than merely discarding it.
- A zero-shot framework combining dynamic strategy generation and robust factual grounding can create effective and trustworthy persuasive AI for various 'social good' applications.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.