The Duality of Persuasion: Between Personalization and Detection

This PhD project investigates the duality of artificial intelligence in persuasive communication, exploring both the generation of personalized content and the detection of persuasive strategies. Insights are drawn from four studies. The first demonstrates how large language models (LLMs) can generate personalized online advertisements targeting specific personality traits, such as openness and neuroticism, with effectiveness comparable to human-written ads. The second introduces a lightweight transformer model for persuasive text detection, achieving state-of-the-art performance in Subtask 3 of SemEval 2023 and applying it to Facebook political ads from the 2022 Australian Federal election campaign. The third, ongoing study extends to persuasive memes detection by employing LLMs for preprocessing and data augmentation before model analysis. Finally, a planned study will explore a dynamic personalization: Developing adaptive LLM frameworks to tailor personalized messages over time based on user feedback or interactions. These studies collectively contribute to the understanding and ethical application of AI in persuasive communication, offering insights for academia, industry, and policymakers while addressing societal implications of AI-driven persuasion.