A Human-Centric AI Governance Framework for Smart Manufacturing Systems

Abstract

Increased usage of AI in smart manufacturing allows predicting equipment failures, performing automated quality checks, optimizing processes, implementing intelligent robotics, and making decisions based on data. At the same time, increased AI autonomy creates a number of issues related to the transparency, accountability, data reliability, safety, and decision-making ability of humans. This paper presents the Human-Centric AI Governance Framework for Smart Manufacturing Systems that helps to adopt AI in a responsible way but still retains the possibility for humans to retain control over operations. The framework comprises five layers that are interconnected with each other. These layers include: AI and data governance, risk and decision classification, human oversight and decision authority, operational monitoring and control, and accountability with continuous improvement. One of the distinguishing features of the framework is that it uses the risk-based approach where the amount of human involvement increases with respect to the risks associated with the potential consequences of the decision made by means of AI. The practical use of the proposed framework is demonstrated by means of an example related to the use of AI in the process of predictive maintenance of equipment. The integration of data validation, risk classification, human involvement, performance monitoring, and feedback in the process of using AI is illustrated on the given example.