Argumentation Effect of a Chatbot for Ethical Discussions about Autonomous AI Scenarios

Abstract

This study presents a German-language AI chatbot adept at fostering argumentative dialogues across various topics. A core component is the Argumentative Knowledge Graph that powers argument understanding and generation. We showcase the Argumentation Effect of our chatbot in a comprehensive user study. This effect is characterized by the chatbot's successful ability to reshape users' attitudes towards ethically challenging topics, such as autonomous AI systems acting as medicine doctors, by imparting novel perspectives and insights. This led to a substantial gain in user knowledge and acceptance, underscoring the transformative role dialogue systems can play in promoting more informed discussions on complex ethical matters. In tandem with this, we introduced a preliminary methodology for advancing the chatbot's Argument Recognition Unit (ARU). By harnessing real-world user arguments and generated samples from an advanced language model, we achieved significant improvements in the confidence and accuracy of argument prediction. In summary, our research demonstrates how an AI chatbot can instigate a powerful Argumentation Effect, reshaping attitudes, widening knowledge, and facilitating informed discourse on ethically charged topics. This is further amplified by improved argument recognition capabilities, opening promising avenues for more effective, and thus more engaging, user-chatbot dialogues.