Steps Towards Meta-Theoretical Foundations for Resilient Human-AI Transformation in Industrial Systems

As artificial intelligence (AI) becomes increasingly integrated into industrial systems, new challenges emerge concerning transparency, worker agency, trust, and resilience. This article examines foundational principles for designing AI systems that support sustainable human-AI collaboration to aid in AI adoption in industries. Unlike the standard AI adoption in the high-tech sector, some industries have slightly different foci in terms of process and quality control, along with safe work. Specifically, there is a need for preservation of human skills and decision-making capacity, considering transparency, worker agency, and trust. The current example is set in a glass tempering industry, demonstrating how design choices in transparency and task allocation may lead to markedly different trajectories for operator agency, skill retention, and resilient operations. This case is used to demonstrate that, unlike the dominant view of AI implementation as “black box,” resilient-AI can be designed such that if failures occur, then the operators still possess sufficient process understanding to diagnose anomalies and implement corrections. Based on the theoretical themes and supporting AI example, the article provides a set of meta-theoretical design foundations for researchers, system designers, and decision-makers with actionable considerations for implementing human-centred AI solutions that are robust, explainable, and adaptable across changing industrial contexts.