Abstract: Artificial intelligence social experiments have the distinctive feature of context-driven innovation, and the “responsibility gap”, which is the most disturbing problem in the application of artificial intelligence in the former, is generated and solved in specific contexts.By analyzing the technical logic of AI-enabled context-driven innovation and the intrinsic requirements of context innovation for AI ethics, we can know that there is a common problem of “responsibility gap” between the “ethics of AI social experiments” and the “ethics of AI”.Therefore, it is a new urgent task to eliminate the problem of “responsibility gap” in AI social experiments, which will directly contribute to the formulation of AI ethical principles and regulations. According to the different experimental contexts, the imputation types of experimental ethical challenges can be subdivided according to the differences in nature and boundaries, namely, the culpability, moral accountability, public accountability, and active responsibility.By introducing “meaningful human control”, the most promising guiding idea and practical framework in the field of AI global governance, we can provide a new way of thinking to comprehensively address the disconnect between the cognitive, embodied, and moral forms of control and their multiple forms of responsibility in AI social experimentation contexts.