Enhancing Data-Driven Decision Support with Multi-Perspective Solutions

Abstract

As digital systems become ubiquitous, providing all-around support for decision makers has become a significant part of contemporary information systems. To this end, numerous data-driven analytics techniques have been widely adopted by various platforms to facilitate decision making in a wide variety of application domains, e.g., product choice, employee recruitment, and medical diagnosis. The appropriate application of various data-driven methodologies for decision support in complex real-world contexts is crucial to gain benefits and to avoid unexpected consequences and, thus, the ability take into account multiple perspectives for better decision support represents an important challenge. In order to provide insights into this question, this thesis focuses on investigating some of the problems existing in decision support applications and attempts to provide various solutions and empirical evidence of the effectiveness of these solutions. Specifically, my thesis proposes to provide more nuanced decision support in different application domains by balancing different aspects of decision support models or by providing complementary sources of information for decision makers, e.g., balancing accuracy and long-tailness to address popularity bias in recommender systems; using individual prediction reliability to complement outcome prediction to support decision making in highly risk-sensitive domains like medical diagnosis or financial markets; providing complementary channels to fulfill online consumption decision support in the retailing industry. Solutions and findings provided by my thesis advance the understanding of decision support problems in multifaceted contexts, and have practical implications for information systems that adopt data-driven methods.