Adaptive Persuasion in Conversational AI: An LLM-Driven Framework for Dynamic Strategy Switching via Personality and Sentiment Analysis

Recent advances in persuasive communication and conversational AI have driven the development of adaptive frameworks that integrate dynamic strategy adjustment within large language model architectures. In this study, we introduce a novel framework that effectively integrates real-time sentiment analysis with comprehensive personality profiling based on the Big Five model to enable adaptive strategy switching during multi-turn dialogues. Our methodology also leverages a reinforcement learning-inspired weight adaptation mechanism that continuously readjusts the relative influence of stable personality traits and transient emotional states, thereby optimizing the selection of persuasive techniques. Furthermore, by incorporating a Retrieval-Augmented Generation (RAG) paradigm alongside the LLaMA model, the proposed framework produces contextually coherent and individually tailored responses that draw upon historical interaction data and strategic knowledge repositories. Experimental evaluations reveal that our adaptive system significantly enhances persuasive efficacy and user engagement, underscoring its potential to advance the state-of-the-art in personalized conversational agents.