Purpose Process Industry faces limits related to artificial intelligence (AI) adoption because of intrinsic dynamics in processes and feedstock. While proposing solutions, current studies neglect to consider the human role, which is deemed critical in Industry 5.0 perspective. This work aims to develop a procedure to evaluate the dynamic interactions between humans and adaptive AI controllers in real settings, addressing the need for robust systems that can integrate human expertise. Design/methodology/approach A novel framework is proposed, integrating Autonomic Computing principles with a classification of humachine interaction types. A three-step analytical procedure is then proposed and applied via an explanatory multiple case study of four industrial applications. Findings The analysis reveals a divergence between theoretical and current practice, where AI applications usually follow a human-as-a-master configuration, suggesting socio-technical barriers that prevent the adoption of higher automation levels. Originality/value This work offers a first attempt at a systematic procedure to evaluate humachine dynamics in Autonomic Computing applications in Process Industry. For practitioners, it constitutes a design tool for more effective socio-technical systems by aligning AI autonomy with operational challenges and the role of human supervision and control.
Cuzzola et al. (Thu,) studied this question.