AI for Compassion: How Large Language Models Could Support Moral Circle Expansion and Suffering Reduction

AI for Compassion: How Large Language Models Could Support Moral Circle Expansion and Suffering Reduction examines whether large language models can support moral reflection and contribute to moral circle expansion from a suffering-focused ethical perspective. The paper explores psychological and ethical barriers to concern for suffering, considers potential mechanisms through which generative AI could encourage perspective-taking and reflection, and proposes a conceptual framework for a system called CompassionGPT. The paper also examines risks and limitations, including misinformation, emotional manipulation, speciesist biases in training data, over-reliance on AI-generated moral reasoning, and uncertainty surrounding artificial sentience. It concludes with safeguards and an evaluation agenda for future empirical research. This is a conceptual, literature-informed research paper. It does not report an original participant study or claim demonstrated causal effectiveness of generative AI in expanding moral concern.