Computational Design Science Research to Solve Real-World Problems with High Impact

Abstract

As big data technologies continue to transform our lives, computational design science research (CDSR) is becoming an essential paradigm in the field of information systems (IS). Impact and relevance are central to CDSR, emphasizing the economic and societal impact of IS research. Nonetheless, it remains uncertain how design science research can contribute to the IS body of knowledge. Since design science differs from conventional sciences in its strong emphasis on problem-solving, there have been ongoing discussions about how design science can bring theoretical contributions to the IS field. This dissertation presents three studies that demonstrate how CDSR can contribute to the IS knowledge base while maximizing its broader impact and relevance. The first essay in the dissertation proposes a novel framework to provide individual-level explanations for any black-box machine learning classifier used in healthcare. The second essay develops and implements a fairness-aware graph representation learning for multiple patient outcome prediction in critical care, extending state-of-the-art methods in graph analytics and natural language processing. The third essay proposes a theory-driven graph representation learning framework for reciprocal recommendation in online matching platforms, which show distinct features from conventional user-item recommendation tasks.