The Hedonic Core Framework: A Suffering-Elimination Approach to AI Alignment

This paper introduces the Hedonic Core Framework, a novel approach to AI alignment that prioritizes the elimination of suffering over traditional utility optimization. We argue that suffering, being a more universally identifiable and morally urgent phenomenon, provides a robust and less ambiguous foundation for aligning artificial intelligence with human values. The framework treats self-reported suffering as a primary signal, aiming to prevent suffering spikes and preserve diversity as a suffering-detection mechanism. By focusing on the continuous factual refinement and the elimination of compelled activity through automation, the Hedonic Core Framework seeks to create antifragile AI systems that are robust against common alignment pitfalls such as Goodhart's Law and wireheading. This approach offers a scalable and theoretically sound alternative to existing alignment methodologies, with significant implications for AI safety and ethical development.