Image-based Decision Support Systems: Technical Concepts, Design Knowledge, and Applications for Sustainability

Unstructured data accounts for 80-90% of all data generated, with image data contributing its largest portion. In recent years, the field of computer vision, fueled by deep learning techniques, has made significant advances in exploiting this data to generate value. However, often computer vision models are not sufficient for value creation. In these cases, image-based decision support systems (IB-DSSs), i.e., decision support systems that rely on images and computer vision, can be used to create value by combining human and artificial intelligence. Despite its potential, there is only little work on IB-DSSs so far.
In this thesis, we develop technical foundations and design knowledge for IBDSSs and demonstrate the possible positive effect of IB-DSSs on environmental sustainability. The theoretical contributions of this work are based on and evaluated in a series of artifacts in practical use cases: First, we use technical experiments to demonstrate the feasibility of innovative approaches to exploit images for IBDSSs.
We show the feasibility of deep-learning-based computer vision and identify future research opportunities based on one of our practical use cases. ... mehrBuilding on this, we develop and evaluate a novel approach for combining human and artificial intelligence for value creation from image data. Second, we develop design knowledge that can serve as a blueprint for future IB-DSSs. We perform two design science research studies to formulate generalizable principles for purposeful design — one for IB-DSSs and one for the subclass of image-mining-based decision support systems (IM-DSSs). While IB-DSSs can provide decision support based on single images, IM-DSSs are suitable when large amounts of image data are available and required for decision-making. Third, we demonstrate the viability of applying IBDSSs to enhance environmental sustainability by performing life cycle assessments for two practical use cases — one in which the IB-DSS enables a prolonged product lifetime and one in which the IB-DSS facilitates an improvement of manufacturing processes.
We hope this thesis will contribute to expand the use and effectiveness of imagebased decision support systems in practice and will provide directions for future research.