THE CO-EVOLUTION INDEX (CEI): A HUMAN–AI–ENVIRONMENT (HAE) FRAMEWORK FOR EVALUATING RESPONSIBLE GENERATIVE AI SYSTEMS

Generative Artificial Intelligence (GenAI) is advancing at an unprecedented pace and is increasingly transforming education, scientific research, healthcare, economics, and public governance. However, most existing evaluation approaches remain focused on the technical performance of AI models through metrics such as accuracy, reasoning capability, benchmark performance, safety, and fairness. While these measures provide valuable insights into model capabilities, they do not fully capture the broader interactions among humans, AI systems, and the environments in which they operate. This study proposes the Human–AI–Environment (HAE) Framework as a systemic model for understanding the co-evolutionary dynamics among humans, artificial intelligence, and their surrounding environments through continuous feedback and resonance processes. Building upon this framework, the study introduces the Co-Evolution Index (CEI) as a novel evaluation metric designed to assess the degree of harmonious development within the entire Human–AI–Environment ecosystem rather than focusing solely on AI performance. The proposed CEI consists of five core dimensions: human capability, AI capability, environmental richness, feedback quality, and resonance. Together, these dimensions provide a holistic assessment of ecosystem development and adaptive capacity. Through illustrative case studies involving the ChatGPT ecosystem, AI-assisted education, and AI-supported scientific discovery, the study demonstrates that the practical value of AI depends not only on the technical capabilities of AI models but also on the quality of interactions among humans, AI systems, and their environments. The findings suggest that feedback and resonance play central roles in shaping the emergent intelligence of Human–AI–Environment ecosystems. The study contributes a new perspective to human-centered AI research by shifting the focus of evaluation from isolated AI models to co-evolving socio-technical ecosystems. Furthermore, it provides an initial foundation for developing next-generation AI evaluation metrics that support responsible, sustainable, and human-centered AI development.