Unveiling the ethical positions of conversational AIs: a study on OpenAI’s ChatGPT and Google’s Bard

Abstract

AbstractIn an era where conversational AIs (CAIs) like OpenAI’s ChatGPT and Google's Bard are becoming integral to daily life, understanding their ethical positions is paramount. This research delves into the expressed moral values of these CAIs, exploring how their pre-training influences their ethical stances. The study aims to assess the articulated ethical positions of ChatGPT and Bard, uncovering whether these systems align with particular moral values. By understanding their ethical positions, the research seeks to provide insights into how these CAIs might respond to prompts and guide users in their selection and utilization. Utilizing O’Boyle and Forsyth’s Ethical Position Questionnaire (EPQ-5), the research evaluated the CAIs’ levels of idealism and relativism. The study also involved a third CAI, Anthropic’s Claude and an online human panel, to analyze the reasoning behind the responses, providing a more nuanced understanding of the ethical positions. The initial findings revealed that ChatGPT aligns more with an ‘absolutist’ position, endorsing strict adherence to moral principles, while Bard leans towards a ‘situationist’ stance, valuing flexibility and situational considerations. However, further analysis by Claude and humans suggested a more complex categorization, with ChatGPT fitting the 'exceptionist' categorization and Bard aligning with ‘absolutism.’ The research underscores the significance of recognizing the trained-in ethical positions of CAIs, as they are not neutral but reflect particular ethical leanings. Understanding these positions is vital for interpreting CAI outputs and using these systems effectively and ethically. The study calls for further exploration into how these ethical positions might influence real-world applications of CAIs.