Abstract
Human-centered AI (HCAI), as a design philosophy, advocates prioritizinghumans in designing, developing, and deploying intelligent systems, aiming tomaximize the benefits of AI technology to humans and avoid its potentialadverse effects. While HCAI has gained momentum, the lack of guidance onmethodology in its implementation makes its adoption challenging. Afterassessing the needs for a methodological framework for HCAI, this paper firstproposes a comprehensive and interdisciplinary HCAI methodological frameworkintegrated with seven components, including design goals, design principles,implementation approaches, design paradigms, interdisciplinary teams, methods,and processes. THe implications of the framework are also discussed. This paperalso presents a "three-layer" approach to facilitate the implementation of theframework. We believe the proposed framework is systematic and executable,which can overcome the weaknesses in current frameworks and the challengescurrently faced in implementing HCAI. Thus, the framework can help put it intoaction to develop, transfer, and implement HCAI in practice, eventuallyenabling the design, development, and deployment of HCAI-based intelligentsystems.