Attention And Reasoning From Experts In Smart Manufacturing With Digital Twin Implementations

Abstract

When it is well known that expert attention is a way to separate each attention head into its own expert in a Mixture-of-Experts design, we understand that, inside the implementation of the digital twin, we need to examine how expert attention and reasoning engagement can enable operative Human-Machine Partnership in Industry 6.0 environments in the near future. We focused on analysing how user-centric design elements moderate these relationships, based on data from the most successful factory projects. A multiple-case study approach was used to examine data from Siemens, GE, and Bosch, three major industry players, which were publicly available. The study used a structured qualitative review of performance measures, system designs, and reported results from Digital Twin applications in the Manufacturing sectors. Three key patterns appear throughout the study. First, expert attention requires intelligent information filtering rather than increased data availability, as the latter would only take up space. Second, the reasoning involvement depends on AI answers that are clear, build trust, and can allow control. Third, user-centric design increases the cognitive effects when matched to users’ expertise levels. All three companies in the study demonstrated double-digit improvements in efficiency, accuracy, and response times over cognitive-focused Digital Twin designs. Throughout the study, we found that companies should invest in explainable AI, match interface complexity to user expertise, and design for learning rather than performance to better implement Digital Twins. They should also prioritise attention support through intelligent filtering, reasoning support through transparent clarifications, and user-centric revision. This study advances Industry 6.0 from a conceptual vision to a practical reality, providing the first systematic analysis of the mechanisms of cognitive engagement in operational Digital Twin systems and offering a framework for experts and a basis for future research.

DOI: https://doi-ds.org/doilink/03.2026-22172216