Decision Support Systems最新文献

Online games represent a rapidly growing and competitive global market for technology firms. Games are viewed as places where people can temporarily escape from reality. However, it is unclear how game escapism fosters game experience and game use, thus indicating a research gap. This gap keeps decision-makers (i.e., firms and policy-makers) in the dark regarding how game escapism affects gameplay, thus hindering effective decision-making. To fill this gap, uses and gratification theory is applied to build a model for explaining the mechanism underlying the influence of game escapism and telepresence on game experience and game use. We collect 1347 online gamer responses with which to test the model. The results indicate that game escapism improves all game experiences, while only enjoyment and concentration increase game use. Moreover, telepresence strengthens the impact of game escapism on enjoyment, concentration, and fantasy. Our findings offer insights for decision-makers, enabling them to leverage game mechanisms to either provide or negate the impact of game escapism, thus changing game use.

Pretrained language models (PLMs) are frequently employed in Decision Support Systems (DSSs) due to their strong performance. However, recent studies have revealed that these PLMs can exhibit social biases, leading to unfair decisions that harm vulnerable groups. Sensitive information contained in sentences from training data is the primary source of bias. Previously proposed debiasing methods based on contrastive disentanglement have proven highly effective. In these methods, PLMs can disentangle sensitive information from non-sensitive information in sentence embedding, and then adapts non-sensitive information only for downstream tasks. Such approaches hinge on having good sentence embedding as input. However, recent research found that most non-fine-tuned PLMs such as BERT produce poor sentence embedding. Disentangling based on these embedding will lead to unsatisfactory debiasing results. Taking a finer-grained perspective, we propose PCFR (Prompt and Contrastive-based Fair Representation), a novel disentanglement method integrating prompt and contrastive learning to debias PLMs. We employ prompt learning to represent information as sensitive embedding and subsequently apply contrastive learning to contrast these information embedding rather than the sentence embedding. PCFR encourages similarity among different non-sensitive information embedding and dissimilarity between sensitive and non-sensitive information embedding. We mitigate gender and religion biases in two prominent PLMs, namely BERT and GPT-2. To comprehensively assess debiasing efficacy of PCFR, we employ multiple fairness metrics. Experimental results consistently demonstrate the superior performance of PCFR compared to representative baseline methods. Additionally, when applied to specific downstream decision tasks, PCFR not only shows strong de-biasing capability but also significantly preserves task performance.

On accommodation-sharing platform, host self-description influence consumer behavior as an important information. Based on the Perceived Value Theory and the Expectation Confirmation Theory, we developed an analytical framework to investigate the relationship between host description strategies and consumer behavior of room booking and satisfaction. We measured host description strategies (honest description and positive description) using machine learning and rule-based text analysis methods. Then we verified the different effects of two host description strategies on each of consumer behaviors based on a panel dataset from Airbnb. Positive description and honest description have a positive impact on room booking and consumer satisfaction respectively. Room price moderates the relationship between host descriptions and consumer behaviors. A highly positive description strategy can promote bookings for high-priced listings but decrease satisfaction. The honest description strategy has a positive effect on the bookings of low-priced listings. This study contributes to tourism literature and property hosts in practice.

Online reviews play a crucial role in shaping buyers' purchase decisions. However, previous research has highlighted the existence of self-selection biases among buyers who contribute to reviews, which in turn leads to biased distributions of review ratings. This research aims to explore the further influences of self-selection bias on buyer satisfaction through agent-based modeling, considering two product differentiations: search and experience differentiation, as well as vertical and horizontal differentiation. Our findings reveal that self-selection bias can have varying positive and negative effects on the usefulness of online reviews in suggesting product quality (i.e., review utility) to buyers, thus affecting buyer satisfaction. While self-selection bias tends to decrease review utility in most scenarios, interestingly, it can also increase review utility by enabling a “screening” function of online reviews in addition to its normal “measuring” function. We also find that the varying effects of self-selection bias on buyer satisfaction are contingent upon the type of products under scrutiny and the interaction of different types of self-selection bias. This research makes valuable contributions to the existing literature on online reviews by introducing a novel theory to explain the effects of self-selection bias on buyer satisfaction.

Data integration plays a crucial role in business intelligence, aiding decision-makers by consolidating data from heterogeneous sources to provide deep insights into business operations and performance. In the big data era, automated data integration solutions need to process high volumes of disparate data robustly and seamlessly for various analytical needs or operational actions. Existing data integration solutions exhibit limited capabilities for capturing and modeling users' needs to execute on-demand data integration. This study, underpinned by affordance theory and the goal definition principles from the Goal-Question-Metric approach, designs and instantiates a goal-driven data integration framework for data analytics. The proposed innovative design automates data integration for non-technical data users. Specifically, it demonstrates how to elicit and ontologize users' data-analytic goals and addresses semantic heterogeneity, thereby recognizing goal-relevant datasets. In a structured evaluation using the context of counter-terrorism analytics, our design artifact shows promising performance in capturing diverse and dynamic user goals for data analytics and in generating integrated data tailored to these goals. Our research establishes a theoretical framework to guide future scholars and practitioners in building smart, goal-driven data integration.

The United States Air Force (USAF) continues to be plagued by a chronic pilot shortage, one that could be exacerbated by an accompanying shortfall in the commercial airlines. As a result, efforts have increased to alleviate this shortage by finding methods to reduce pilot training attrition. We contribute to these efforts by setting forth a decision support system (DSS) for pilot candidate selection using modern machine learning techniques. In view of the recent Responsible Artificial Intelligence Strategy published by the United States Department of Defense, this research leverages interpretable and explainable machine learning methods to create traceable and equitable models that may be responsibly and reliably governed. These models are used to regress candidates’ average merit assignment selection system scores based on information available for selection and prior to training. More specifically, using data provided by the USAF from 2010 to 2018, this paper develops and analyzes multiple interpretable models based on Gaussian Bayesian networks, as well as multiple black-box models rendered explainable by SHAP values and conformal prediction. A preferred pair of interpretable and explainable models is selected and embedded within a DSS for USAF pilot candidate selection boards: the Air Force Pilot Applicant Selection System. The utilization of this DSS is explored, the analyses it enables are discussed, and relevant USAF policymaking issues are examined.

Users can increasingly delegate to information systems (IS) – that is transferring rights and responsibilities regarding certain tasks – even to the degree that IS can act autonomously (i.e., without the intervention or supervision of users). What is more, IS increasingly offer to assume the rights and responsibilities for a task not only in response to user prompts (i.e., user-invoked delegation) but also without user prompts (i.e., IS-invoked delegation). Yet, little is known about whether, how, and why users agree to delegation when they are asked by the IS in contrast to when they self-initiate the delegation. Drawing on self-affirmation theory, we investigate user acceptance of IS- versus user-invoked delegation in two complementary online experiments in software development. Our core findings reveal that IS-invoked (vs. user-invoked) delegation increases users' perceived self-threat and thus decreases their willingness to accept delegation. This threatening effect is larger the less (vs. more) the user perceives control after the potential delegation. Taken together, we uncover defensive user responses to IS-invoked delegation. Furthermore, we shed light on the underlying and moderating mechanisms representing the reasons and contextual features that explain and mitigate these defensive measures. These findings have significant implications for IS designers seeking to improve user-IS collaboration and outcomes by employing IS-invoked delegation.

Categorization is one of the basic tasks in machine learning and data analysis. Building on formal concept analysis (FCA), the starting point of the present work is that different ways to categorize a given set of objects exist, which depend on the choice of the sets of features used to classify them, and different such sets of features may yield better or worse categorizations, relative to the task at hand. In their turn, the (a priori) choice of a particular set of features over another might be subjective and express a certain epistemic stance (e.g. interests, relevance, preferences) of an agent or a group of agents, namely, their interrogative agenda. In the present paper, we represent interrogative agendas as sets of features, and explore and compare different ways to categorize objects w.r.t. different sets of features (agendas). We first develop a simple unsupervised FCA-based algorithm for outlier detection which uses categorizations arising from different agendas. We then present a supervised meta-learning algorithm to learn suitable (fuzzy) agendas for categorization as sets of features with different weights or masses. We combine this meta-learning algorithm with the unsupervised outlier detection algorithm to obtain a supervised outlier detection algorithm. We show that these algorithms perform at par with commonly used algorithms for outlier detection on commonly used datasets in outlier detection. These algorithms provide both local and global explanations of their results.

Although artificial intelligence can contribute to decision-making processes, many industry players lag behind pioneering companies in utilizing artificial intelligence-driven technologies, which is a significant problem. Explainable artificial intelligence can be a viable solution to mitigate this problem. This paper proposes a research model to address how explainable artificial intelligence can impact decision-making processes. Using an experimental design, empirical data is collected to test the research model. This paper is one of the pioneer papers providing empirical evidence about the impact of explainable artificial intelligence on supply chain decision-making processes. We propose a serial mediation path, which includes transparency and agile decision-making. Findings reveal that explainable artificial intelligence enhances transparency, thereby significantly contributing to agile decision-making for improving cyber resilience during supply chain cyberattacks. Moreover, we conduct a post hoc analysis using text analysis to explore the themes present in tweets discussing explainable artificial intelligence in decision support systems. The results indicate a predominantly positive attitude towards explainable artificial intelligence within these systems. Furthermore, the text analysis reveals two main themes that emphasize the importance of transparency, explainability, and interpretability in explainable artificial intelligence.

Accurate assessment of financial distress of SMEs is critical as it has strong implications for various stakeholders to understand the firm's financial health. Recent studies start to leverage network data and suggest the effect of event propagation for predicting financial distress. Yet such methods face methodological challenges in determining and explaining event propagation due to heterogeneous entities and events. In this research, we propose to extend graph contrastive learning and interpretable machine learning in the context of a firm network formed by distinct entities (e.g., firms and persons) and events (i.e., positive and negative), and employ the propagation influence of events in firm networks for financial distress assessment of SMEs. To this end, we design a novel artifact, i.e., adaptive interpretable heterogeneous graph contrastive learning, by drawing on homophily and social learning theories. Our experimental results demonstrate the effectiveness of the proposed artifacts and suggest the differing effects of positive vs. negative events on the financial distress of SMEs. This research contributes to the IS and explainable graph AI literature by improving the assessment and interpretability of network-based financial distress of SMEs.