Relational AI and Consciousness Impressions: Ethical Frontiers for Designing Artificial Consciousness
Rolando Calero, Gerardo Herrera preprint
Debates about whether large language models like GPT-4 could become conscious are intensifying. Rather than trying to build self-aware machines, a more practical and ethically sound approach is to design AI that convincingly simulates empathy and understanding to improve human–AI interaction. Drawing on phenomenology and theories of alterity, the argument is that relational simulation—creating AI that ethically engages through simulated otherness—offers greater social benefits than pursuing autonomous machine consciousness. Empirical findings on simulated empathy in LLMs are integrated with discussions of AI alignment and machine ethics. This framework defines consciousness impressions as human interpretations of AI behavior and provides a coherent pathway for AI development while probing the boundaries of mind and identity.