Exploring the role of human centered augmented reality in construction: change detection, work efficiency, and human-AI collaboration

Abstract

The architectural engineering and construction (AEC) sector is characterized by its intricate project workflows and the constant exchange of extensive, detailed information among multiple stakeholders. This complexity necessitates the adoption of cutting-edge digital solutions that can handle vast data streams while improving communication, collaboration, and operational efficiency. This dissertation examined the Augmented Reality (AR) and Artificial Intelligence (AI) technologies in the construction industry, with a focus on their interaction with key human factors, including safety hazard detection, trust in technology, spatial cognitive skills, and workforce demographics. It responded to the challenges posed by increasingly complex construction projects and the need for enhanced efficiency, safety, and sustainability by emphasizing the importance of evaluating technological solutions not only for their technical merits but also for their effects on human performance and cognition. To achieve the research objective, 100 industry craft workers and 219 students participated in an assembly task on a full-scale Mechanical, Electrical, and Plumbing (MEP) model. First, the research investigated how AR head-mounted displays (HMDs) affect workers' ability to detect changes and safety hazards compared to traditional information formats in dynamic construction environments. Findings showed that AR HMD usage impaired hazard detection, especially as the density of information presented on the display increased. This impairment was notably greater among older workers, while spatial cognitive ability did not significantly moderate this effect. These results demonstrated the potential for immersive AR interfaces to distract users from critical environmental awareness, thereby raising essential safety concerns about their implementation. Furthermore, the research investigated how varying levels of detail (LOD) in AR models, compared to traditional paper-based instructions, affect work performance during assembly processes. The study demonstrated that higher-detail AR models significantly improved outcomes by speeding up task completion, reducing errors, and minimizing the need for rework. These performance gains were most pronounced among novice workers, indicating that AR has strong potential to bridge skill gaps and enhance inclusivity within the workforce. The research also explored the relationship between spatial cognitive ability and task performance, identifying that cognitive skills contributed positively to novices' success, while experienced workers compensated through expertise, resulting in a neutralizing effect on age- and cognition-related performance differences. Lastly, the study investigated trust in AI versus human-generated design information and its relationship to AR visualization. The findings revealed no significant overall difference in trust between AI and human sources; however, trust in AI declined significantly when errors involved safety-critical information, showing that participants' trust in the origin of design information was proportional to their perceived level of risk associated with that information. Importantly, AR visualization was found to reduce skepticism toward AI origin by enhancing understanding and confidence in the conveyed information. Work efficiency benefited from AR use regardless of the design source, though individual cognitive ability remained the primary predictor of task speed. Overall, this work highlights the crucial role of human factors in effectively integrating AR and AI technologies into construction workflows. It demonstrated AR can help a diverse workforce engage more effectively with complex project information, improve work performance, but at the same time, may compromise the user's ability to detect changes in a dynamic construction environment. Additionally, the findings provide practical guidance for developing transparent, reliable, and explainable AI systems, supported by intuitive AR interfaces, to foster trust and ensure human oversight for safe and widespread adoption of technology. By employing ecologically valid experimental methods that involved realistic assembly tasks and diverse user populations, this research advances knowledge about real-world human-technology interactions in the construction industry. It laid the groundwork for future construction ecosystems where digital innovations harmoniously support human cognitive diversity, enhancing safety, productivity, and worker experience. The proposed framework prioritizes cognitive factors, risk awareness, trust, and inclusivity as pillars for successful AR and AI integration, paving the way for a technologically advanced yet human-centered construction industry.