Date of Award
8-2024
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Civil Engineering
Committee Chair/Advisor
Kapil Chalil Madathil
Committee Member
Kalyan Piratla
Committee Member
Vivek Sharma
Committee Member
Patrick Rosopa
Abstract
The Architecture, Engineering, and Construction (AEC) sector, characterized by its complex tasks, often faces challenges in adopting cutting-edge technologies, a trend that can significantly hinder productivity improvements compared to other industries. This dissertation explores the emerging integration of Artificial Intelligence (AI) and Augmented Reality (AR) within the AEC domain, investigating how effectively this can be implemented to potentially revolutionize industry practices. AI could be used in the AEC domain for advanced analytical and decision-making processes, while AR could be used for training, visualization, and remote collaboration.
In scenarios where construction site workers require expert guidance, these technologies could be helpful. AR, for instance, enables remote experts to offer real-time assistance, while AI can provide data-driven recommendations through AR interfaces, analyzing on-site challenges to suggest practical solutions. This integration of AI and AR harnesses the strengths of both technologies, potentially transforming the AEC landscape. However, this technological integration is not without its challenges, particularly concerning trust in AI, an important factor for the successful implementation of AI across domains.
The initial study of this dissertation investigates the effect of task complexity and AI recommender system reliability on trust, performance, and workload. Utilizing a Partially Observable Markov Decision Process (POMDP) model, the evolution of trust during interactions with recommender AI was modeled. The experimental study identified that the task complexity did not significantly impact trust in AI. Instead, the reliability of the AI agent emerged as a crucial determinant, with higher reliability correlating with increased trust. The trust trajectory predicted by the POMDP model aligned closely with experimental findings under most conditions, offering valuable insights for AI system designers for AR modules while completing construction tasks.
The second study investigated the effect of transparency and explainability on trust in the recommender AI. Specifically, it investigated the effects of varying levels of explanation and transparency on trust, performance, and workload while completing a construction task with the help of AR technology. The study identified that the overall change in trust after introducing a combination of transparency and explainability increased regardless of the reliability of the AI but with the cost of the time taken to complete the task. This combination of transparency and explainability was provided along with all the recommendations throughout the task. This led to the idea of the final study, which is to investigate the effect of providing the combination of transparency and explainability in a non-continuous way, especially in AR modules, to avoid visual clutter of information.
The third study investigated the effect of providing a combination of AI transparency and explainability along with the recommendation in an adaptive manner to maintain appropriate trust, compared to providing them continuously. The study identified that there was no significant reduction in trust level when transparency and explainability were provided when AI’s confidence was low and when provided during the first, middle and last steps of the task compared to providing them continuously with an advantage of time taken to complete the task while providing them intermittently.
This research attempts to offer a detailed exploration of the factors influencing trust in AI, particularly when interfaced with AR technology while completing a construction task. The findings are expected to guide the development of more effective, trust-enhancing AI tools, paving the way for their broader acceptance and implementation in complex construction environments.
Recommended Citation
Chandrika Bhanu, Aasish, "Integrating Artificial Intelligence and Augmented Reality For Enhanced Task Performance In the Construction Industry" (2024). All Dissertations. 3673.
https://open.clemson.edu/all_dissertations/3673
Included in
Architectural Engineering Commons, Civil Engineering Commons, Cognition and Perception Commons, Construction Engineering Commons, Construction Engineering and Management Commons, Experimental Analysis of Behavior Commons, Hydraulic Engineering Commons, Other Civil and Environmental Engineering Commons, Structural Engineering Commons