Audience-Aware Argument Generation

During my Ph.D, I argued that effective arguments must be not only relevant to the discussion and responsive to an opponent’s reasoning but more importantly adapted to the audience’s interests and beliefs - an aspect that was not well addressed in the computational argumentation field. To this end, I developed methods to (a) extract key points of a debate, (b) model the opponent’s main claim and weak premises, and (c) tailor generated arguments to different audiences based on their stance or moral foundations. I also proposed approaches to counterargument generation—either by attacking weak premises or by jointly generating conclusions and counterarguments. Collectively, this work contributed to a better understanding of how to structure and condition arguments to achieve greater persuasive and communicative effectiveness, driving current technology at the time to be more considerate of its audience and social context. Further details about my thesis can be found here