Pub Date : 2025-10-06DOI: 10.1016/j.dss.2025.114554
Xingchen Chen , Hao Liu , Libo Liu , Kristijan Mirkovski , Marta Indulska , Katja Holtta-Otto
Customer-driven innovation relies on leveraging customer insights to develop or improve products that meet evolving customer needs and preferences. Central to this innovation is the ideation process that involves two key stages: identifying customer needs and generating new ideas. While user-generated content offers a rich source of consumer insights, existing approaches for automating the ideation process—including unsupervised learning, supervised learning, deep learning, text summarization and GenAI—face limitations that restrict their scalability and practical utility. Moreover, these approaches often address only isolated stages of the ideation process. Based on a design science methodology and grounded in the user innovation theory, this paper develops and evaluates an integrated GenAI-driven method that automates the ideation process. The method consists of two stages: (1) customer opinion knowledgebase construction and (2) GenAI-based idea generation. The proposed GenAI-driven method offers an adaptable, scalable, and comprehensive solution for advancing customer-driven innovation.
{"title":"An integrated GenAI-driven method for automating ideation with user-generated content","authors":"Xingchen Chen , Hao Liu , Libo Liu , Kristijan Mirkovski , Marta Indulska , Katja Holtta-Otto","doi":"10.1016/j.dss.2025.114554","DOIUrl":"10.1016/j.dss.2025.114554","url":null,"abstract":"<div><div>Customer-driven innovation relies on leveraging customer insights to develop or improve products that meet evolving customer needs and preferences. Central to this innovation is the ideation process that involves two key stages: identifying customer needs and generating new ideas. While user-generated content offers a rich source of consumer insights, existing approaches for automating the ideation process—including unsupervised learning, supervised learning, deep learning, text summarization and GenAI—face limitations that restrict their scalability and practical utility. Moreover, these approaches often address only isolated stages of the ideation process. Based on a design science methodology and grounded in the user innovation theory, this paper develops and evaluates an integrated GenAI-driven method that automates the ideation process. The method consists of two stages: (1) customer opinion knowledgebase construction and (2) GenAI-based idea generation. The proposed GenAI-driven method offers an adaptable, scalable, and comprehensive solution for advancing customer-driven innovation.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"199 ","pages":"Article 114554"},"PeriodicalIF":6.8,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-03DOI: 10.1016/j.dss.2025.114544
Niklas Wördehoff , Andreas Egger , Wolfgang Kratsch , Fabian König , Maximilian Röglinger
Process mining has developed into one of the most important research streams in business process management. Despite its successful application to improve process performance in industry, there is still substantial potential to be realized in the coming years. One of them is the use of unstructured video data to enable the analysis of previously unobservable parts of processes. Existing approaches derive event logs from video data by extracting a predefined set of potentially relevant activities. As this set is typically determined using a process model or input from process experts, rather than the available video data, current solutions are unable to identify activities that extend beyond the presumed process behavior, limiting transparency in process analysis. Therefore, this study aims to develop a solution that enables the extraction of actual process behavior from video data, as opposed to assumed process activities. Following a design science research methodology, we developed and evaluated the Reference Architecture for Video Event Extraction (RAVEE), which enables the identification of individual process steps in an unsupervised manner. We performed several evaluation activities to ensure the completeness and applicability of the RAVEE. A prototypical instantiation of the RAVEE further demonstrates its ability to extract process-relevant events from video data on two real-world datasets.
{"title":"Beyond assumptions: A reference architecture to enable unsupervised process discovery from video data","authors":"Niklas Wördehoff , Andreas Egger , Wolfgang Kratsch , Fabian König , Maximilian Röglinger","doi":"10.1016/j.dss.2025.114544","DOIUrl":"10.1016/j.dss.2025.114544","url":null,"abstract":"<div><div>Process mining has developed into one of the most important research streams in business process management. Despite its successful application to improve process performance in industry, there is still substantial potential to be realized in the coming years. One of them is the use of unstructured video data to enable the analysis of previously unobservable parts of processes. Existing approaches derive event logs from video data by extracting a predefined set of potentially relevant activities. As this set is typically determined using a process model or input from process experts, rather than the available video data, current solutions are unable to identify activities that extend beyond the presumed process behavior, limiting transparency in process analysis. Therefore, this study aims to develop a solution that enables the extraction of actual process behavior from video data, as opposed to assumed process activities. Following a design science research methodology, we developed and evaluated the Reference Architecture for Video Event Extraction (RAVEE), which enables the identification of individual process steps in an unsupervised manner. We performed several evaluation activities to ensure the completeness and applicability of the RAVEE. A prototypical instantiation of the RAVEE further demonstrates its ability to extract process-relevant events from video data on two real-world datasets.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"199 ","pages":"Article 114544"},"PeriodicalIF":6.8,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145270821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-27DOI: 10.1016/j.dss.2025.114545
Lan Li, Noelle Li Ying Cheah, Seung Hyun Kim
As generative AI continues to transform industries, including the creative sector, it has become critical to understand how it interacts with legal frameworks. This study aims to investigate the effect of the landmark ruling issued by the U.S. District Court on August 18, 2023, which declared AI-generated art uncopyrightable to provide clarity to previously ambiguous legal standards on the AI-related services in online labor markets. Our findings reveal that prices for AI-related gigs on an online freelancer platform dropped by 32.97 % following the ruling, suggesting that the lack of copyright may have reduced the perceived value by limiting clients' residual rights. Furthermore, our research indicates that both freelancer experience and communication efficiency significantly moderate the relationship between AI art non-copyrightability and project pricing. In addition, the results show that large corporate clients were more affected by the ruling than individual clients. In contrast, prices for projects commissioned by small and mid-sized corporate clients did not change significantly. This suggests that large firms are more sensitive to intellectual property uncertainties because they rely heavily on formal rights to secure control and revenue from creative assets. This research contributes to a nuanced understanding of how legal frameworks for AI may shape the gig economy's AI art-related creative services, offering valuable guidelines for more informed decision-making by freelancers, clients, platform owners, and policymakers in this evolving landscape.
{"title":"AI art in the gig economy: Investigating the effects of non-copyrightability in online labor markets","authors":"Lan Li, Noelle Li Ying Cheah, Seung Hyun Kim","doi":"10.1016/j.dss.2025.114545","DOIUrl":"10.1016/j.dss.2025.114545","url":null,"abstract":"<div><div>As generative AI continues to transform industries, including the creative sector, it has become critical to understand how it interacts with legal frameworks. This study aims to investigate the effect of the landmark ruling issued by the U.S. District Court on August 18, 2023, which declared AI-generated art uncopyrightable to provide clarity to previously ambiguous legal standards on the AI-related services in online labor markets. Our findings reveal that prices for AI-related gigs on an online freelancer platform dropped by 32.97 % following the ruling, suggesting that the lack of copyright may have reduced the perceived value by limiting clients' residual rights. Furthermore, our research indicates that both freelancer experience and communication efficiency significantly moderate the relationship between AI art non-copyrightability and project pricing. In addition, the results show that large corporate clients were more affected by the ruling than individual clients. In contrast, prices for projects commissioned by small and mid-sized corporate clients did not change significantly. This suggests that large firms are more sensitive to intellectual property uncertainties because they rely heavily on formal rights to secure control and revenue from creative assets. This research contributes to a nuanced understanding of how legal frameworks for AI may shape the gig economy's AI art-related creative services, offering valuable guidelines for more informed decision-making by freelancers, clients, platform owners, and policymakers in this evolving landscape.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"199 ","pages":"Article 114545"},"PeriodicalIF":6.8,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145236140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-17DOI: 10.1016/j.dss.2025.114542
Yunchang Zhu, Xianghua Lu
Affable design is increasingly employed in AI conversational agents to foster smoother interaction and enhance user experience. However, a growing concern is that this overemphasis on social appeal often overlooks corrective interventions, particularly when users hold false or biased beliefs. Such omissions carry the risk of reinforcing user misconceptions and ultimately undermining the effectiveness of human–AI collaboration. Drawing upon the attribution theory, this study investigates whether the error-correction behavior of AI agents offset these risks and improve user engagement. Empirical evidence from three experimental studies verifies that AI agents' error-correction behavior indeed enhances users' perceived responsibility of AI agents and strengthens their engagement intentions. This effect does not appear to compromise social comfort, especially in the context where responsibility takes precedence, such as healthcare. This study further finds that the high expertise of AI agents amplifies the positive effects of error-correction behavior, while high entitativity diminishes these effects by blurring AI agents' responsibility. These findings offer important guidance for designing responsible AI agents and highlight the value of AI error-correction behaviors in human-AI interaction.
{"title":"Being responsible or affable: Investigating the effects of AI error correction behaviors on user engagement","authors":"Yunchang Zhu, Xianghua Lu","doi":"10.1016/j.dss.2025.114542","DOIUrl":"10.1016/j.dss.2025.114542","url":null,"abstract":"<div><div>Affable design is increasingly employed in AI conversational agents to foster smoother interaction and enhance user experience. However, a growing concern is that this overemphasis on social appeal often overlooks corrective interventions, particularly when users hold false or biased beliefs. Such omissions carry the risk of reinforcing user misconceptions and ultimately undermining the effectiveness of human–AI collaboration. Drawing upon the attribution theory, this study investigates whether the error-correction behavior of AI agents offset these risks and improve user engagement. Empirical evidence from three experimental studies verifies that AI agents' error-correction behavior indeed enhances users' perceived responsibility of AI agents and strengthens their engagement intentions. This effect does not appear to compromise social comfort, especially in the context where responsibility takes precedence, such as healthcare. This study further finds that the high expertise of AI agents amplifies the positive effects of error-correction behavior, while high entitativity diminishes these effects by blurring AI agents' responsibility. These findings offer important guidance for designing responsible AI agents and highlight the value of AI error-correction behaviors in human-AI interaction.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114542"},"PeriodicalIF":6.8,"publicationDate":"2025-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158579","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-16DOI: 10.1016/j.dss.2025.114543
Yunhong Xu, Yitong Chen, Li Sun, Yu Chen
Unlabeled data and multi-source data provide unprecedented opportunities for the financial industry to improve credit scoring accuracy. When utilizing unlabeled data, existing credit scoring methods often suffer from unreliability issues due to improper clustering or the introduction of noise when predicting labels. When utilizing multi-source data, existing credit scoring methods based on federated learning frameworks fail to tailor models for different data distributions of different data sources due to the limitations of relying on a single global model. Moreover, recent studies have explored the individual value of unlabeled data and multi-source data, but they often fail to utilize both. To address these issues, we propose UMDCS (Unlabeled and Multi-Source data Driven Credit Scoring), a self-supervised credit scoring method that utilizes both unlabeled and multi-source data simultaneously. To utilize unlabeled data, we propose a novel sample masking function to generate pseudo-labels for unlabeled data and pre-train the encoder using the pretext tasks. To utilize multi-source data, we employ a horizontal federated learning framework to aggregate local encoders into a global model while preserving data privacy. The global encoder is concatenated with personalized predictors to form personalized credit scoring models for each data source. Five experiments and statistical significance tests show that UMDCS outperforms other baseline methods.
{"title":"Corporate credit scoring method based on unlabeled data and multi-source data","authors":"Yunhong Xu, Yitong Chen, Li Sun, Yu Chen","doi":"10.1016/j.dss.2025.114543","DOIUrl":"10.1016/j.dss.2025.114543","url":null,"abstract":"<div><div>Unlabeled data and multi-source data provide unprecedented opportunities for the financial industry to improve credit scoring accuracy. When utilizing unlabeled data, existing credit scoring methods often suffer from unreliability issues due to improper clustering or the introduction of noise when predicting labels. When utilizing multi-source data, existing credit scoring methods based on federated learning frameworks fail to tailor models for different data distributions of different data sources due to the limitations of relying on a single global model. Moreover, recent studies have explored the individual value of unlabeled data and multi-source data, but they often fail to utilize both. To address these issues, we propose UMDCS (Unlabeled and Multi-Source data Driven Credit Scoring), a self-supervised credit scoring method that utilizes both unlabeled and multi-source data simultaneously. To utilize unlabeled data, we propose a novel sample masking function to generate pseudo-labels for unlabeled data and pre-train the encoder using the pretext tasks. To utilize multi-source data, we employ a horizontal federated learning framework to aggregate local encoders into a global model while preserving data privacy. The global encoder is concatenated with personalized predictors to form personalized credit scoring models for each data source. Five experiments and statistical significance tests show that UMDCS outperforms other baseline methods.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114543"},"PeriodicalIF":6.8,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145120987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-15DOI: 10.1016/j.dss.2025.114540
Benjamin Gundersen , Saikishore Kalloori , Abhishek Srivastava
News recommender systems are decision support systems that exploit user-article interactions over a short duration of time to discover users’ interests and predict unseen news articles to generate a ranking of news articles that are relevant and interesting. In the news recommendation scenario, the relevance of articles decays quickly, and fresh articles are generated daily. Session based models are proposed using time-aware approaches to exploit interactions sequentially. Prior news recommender systems do not consider emotional information expressed in news articles within sessions for recommendations. Emotions play a key role in supporting decision-making and emotionally charged headlines can evoke curiosity or urgency, prompting users to click on certain articles. This paper presents an innovative decision support system for session based news recommendation, using expressed emotions from news articles, such as expressed in the title, abstract, and text, to improve user decision-making. We introduce a novel methodology that incorporates expressed emotions into three session based news recommendation models. Our results demonstrate that expressed emotion carries valuable information to improve session based news recommenders on various ranking metrics significantly and proved especially beneficial in scenarios with limited user interaction history, addressing the cold-start problem. The results show significant improvements in ranking metrics, emphasizing the utility of emotional features for dynamic decision-making support.
{"title":"Emotion aware session based news recommender systems","authors":"Benjamin Gundersen , Saikishore Kalloori , Abhishek Srivastava","doi":"10.1016/j.dss.2025.114540","DOIUrl":"10.1016/j.dss.2025.114540","url":null,"abstract":"<div><div>News recommender systems are decision support systems that exploit user-article interactions over a short duration of time to discover users’ interests and predict unseen news articles to generate a ranking of news articles that are relevant and interesting. In the news recommendation scenario, the relevance of articles decays quickly, and fresh articles are generated daily. Session based models are proposed using time-aware approaches to exploit interactions sequentially. Prior news recommender systems do not consider emotional information expressed in news articles within sessions for recommendations. Emotions play a key role in supporting decision-making and emotionally charged headlines can evoke curiosity or urgency, prompting users to click on certain articles. This paper presents an innovative decision support system for session based news recommendation, using expressed emotions from news articles, such as expressed in the title, abstract, and text, to improve user decision-making. We introduce a novel methodology that incorporates expressed emotions into three session based news recommendation models. Our results demonstrate that expressed emotion carries valuable information to improve session based news recommenders on various ranking metrics significantly and proved especially beneficial in scenarios with limited user interaction history, addressing the cold-start problem. The results show significant improvements in ranking metrics, emphasizing the utility of emotional features for dynamic decision-making support.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114540"},"PeriodicalIF":6.8,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145120986","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-13DOI: 10.1016/j.dss.2025.114526
Subhajit Bag , Sobhan Sarkar , Indranil Bose
Assessing cybersecurity policies is crucial for any organization to combat evolving cyber threats. The absence of a comprehensive dataset has prevented previous studies from analyzing the risk of organizations’ cybersecurity policies. Past studies have not considered temporal information in the policies. Analysis of cybersecurity policies using attention mechanism requires automated determination of optimal number of attention units which remains unaddressed. Moreover, absence of interpretation in cybersecurity studies creates a barrier to understanding policy vulnerabilities and developing targeted solutions. To address these challenges, we develop a decision support system which (i) enhances risk classification of organization’s cybersecurity policies, (ii) develops a comprehensive cybersecurity policy dataset from the websites of 190 companies, transformed into a knowledge graph to capture entity relationships among various policies, (iii) integrates temporal information into the knowledge graph by incorporating time stamps from event sequences in cyberattack information, (iv) develops Explainable Factor Analysis based Multi-Head Attention mechanism, which automates the determination of the optimal number of attention units and optimizes data allocation across attention units using factor analysis, and (v) utilizes attention heatmaps and shapley values for interpretability. Our cybersecurity policy dataset is used as a case study with four benchmark datasets for further validation. Results reveal that our model outperforms the other state-of-the-art, achieving an 87.78% score, followed by robustness checking and statistical significance testing. Finally, Shapley values are used to interpret the model’s output to identify vulnerabilities within the organizational policies, providing crucial insights enabling decision-makers to enhance their cybersecurity policies and mitigate potential threats.
{"title":"Enhancing cybersecurity risk assessment using temporal knowledge graph-based explainable decision support system","authors":"Subhajit Bag , Sobhan Sarkar , Indranil Bose","doi":"10.1016/j.dss.2025.114526","DOIUrl":"10.1016/j.dss.2025.114526","url":null,"abstract":"<div><div>Assessing cybersecurity policies is crucial for any organization to combat evolving cyber threats. The absence of a comprehensive dataset has prevented previous studies from analyzing the risk of organizations’ cybersecurity policies. Past studies have not considered temporal information in the policies. Analysis of cybersecurity policies using attention mechanism requires automated determination of optimal number of attention units which remains unaddressed. Moreover, absence of interpretation in cybersecurity studies creates a barrier to understanding policy vulnerabilities and developing targeted solutions. To address these challenges, we develop a decision support system which (i) enhances risk classification of organization’s cybersecurity policies, (ii) develops a comprehensive cybersecurity policy dataset from the websites of 190 companies, transformed into a knowledge graph to capture entity relationships among various policies, (iii) integrates temporal information into the knowledge graph by incorporating time stamps from event sequences in cyberattack information, (iv) develops Explainable Factor Analysis based Multi-Head Attention mechanism, which automates the determination of the optimal number of attention units and optimizes data allocation across attention units using factor analysis, and (v) utilizes attention heatmaps and shapley values for interpretability. Our cybersecurity policy dataset is used as a case study with four benchmark datasets for further validation. Results reveal that our model outperforms the other state-of-the-art, achieving an 87.78% <span><math><msub><mrow><mi>F</mi></mrow><mrow><mn>1</mn></mrow></msub></math></span> score, followed by robustness checking and statistical significance testing. Finally, Shapley values are used to interpret the model’s output to identify vulnerabilities within the organizational policies, providing crucial insights enabling decision-makers to enhance their cybersecurity policies and mitigate potential threats.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114526"},"PeriodicalIF":6.8,"publicationDate":"2025-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145097012","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-12DOI: 10.1016/j.dss.2025.114541
Hasan Mahmud , Najmul Islam , Satish Krishnan
Robo-advisors (RAs) are cost-effective, bias-resistant alternatives to human financial advisors, yet adoption remains limited. While prior research has examined user interactions with RAs, less is known about how individuals interpret RA roles and integrate their advice into decision-making. To address this gap, this study employs a multiphase mixed methods design integrating a behavioral experiment (N = 334), thematic analysis, and follow-up quantitative testing. Findings suggest that people tend to rely on RAs, with reliance shaped by information about RA performance and the framing of advice as gains or losses. Thematic analysis reveals three RA roles in decision-making and four user types, each reflecting distinct patterns of advice integration. In addition, a 2 × 2 typology categorizes antecedents of acceptance into enablers and inhibitors at both the individual and algorithmic levels. By combining behavioral, interpretive, and confirmatory evidence, this study advances understanding of human–RA collaboration and provides actionable insights for designing more trustworthy and adaptive RA systems.
{"title":"Human-Robo-advisor collaboration in decision-making: Evidence from a multiphase mixed methods experimental study","authors":"Hasan Mahmud , Najmul Islam , Satish Krishnan","doi":"10.1016/j.dss.2025.114541","DOIUrl":"10.1016/j.dss.2025.114541","url":null,"abstract":"<div><div>Robo-advisors (RAs) are cost-effective, bias-resistant alternatives to human financial advisors, yet adoption remains limited. While prior research has examined user interactions with RAs, less is known about how individuals interpret RA roles and integrate their advice into decision-making. To address this gap, this study employs a multiphase mixed methods design integrating a behavioral experiment (<em>N</em> = 334), thematic analysis, and follow-up quantitative testing. Findings suggest that people tend to rely on RAs, with reliance shaped by information about RA performance and the framing of advice as gains or losses. Thematic analysis reveals three RA roles in decision-making and four user types, each reflecting distinct patterns of advice integration. In addition, a 2 × 2 typology categorizes antecedents of acceptance into enablers and inhibitors at both the individual and algorithmic levels. By combining behavioral, interpretive, and confirmatory evidence, this study advances understanding of human–RA collaboration and provides actionable insights for designing more trustworthy and adaptive RA systems.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114541"},"PeriodicalIF":6.8,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221256","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-08DOI: 10.1016/j.dss.2025.114527
Weiyue Li , Ming Gao , Bowei Chen , Jingmin An , Yeming Gong
Social recommender systems help address data sparsity in user–product interactions by leveraging social relationships to infer user preferences. However, existing models often overlook the role of social capital that influence decision-making in social commerce. Social capital consists of structural, relational, and cognitive dimensions, all of which shape user preferences. To better understand these influences, we propose a multi-task learning framework named DeepSC that integrates social capital theory into preference modeling. Its user preference learning module extracts structural features through graph-based pre-training, learns relational features from dynamic user embeddings, and models cognitive features using a hypergraph attention network. Additionally, the dual graph-based product feature learning module enhances cognitive feature extraction by incorporating product co-interactions. DeepSC is optimized through a joint learning objective, combining point-wise and pair-wise learning with an auxiliary social link prediction task to refine user representations. Experiments on three e-commerce datasets demonstrate that DeepSC significantly outperforms the state-of-the-art recommendation models, highlighting the effectiveness of integrating social capital into social preference learning. Our research advances social recommendation by providing a social capital theory-driven approach to modeling user behavior in digital commerce.
{"title":"Social capital matters: Towards comprehensive user preference for product recommendation with deep learning","authors":"Weiyue Li , Ming Gao , Bowei Chen , Jingmin An , Yeming Gong","doi":"10.1016/j.dss.2025.114527","DOIUrl":"10.1016/j.dss.2025.114527","url":null,"abstract":"<div><div>Social recommender systems help address data sparsity in user–product interactions by leveraging social relationships to infer user preferences. However, existing models often overlook the role of social capital that influence decision-making in social commerce. Social capital consists of structural, relational, and cognitive dimensions, all of which shape user preferences. To better understand these influences, we propose a multi-task learning framework named DeepSC that integrates social capital theory into preference modeling. Its user preference learning module extracts structural features through graph-based pre-training, learns relational features from dynamic user embeddings, and models cognitive features using a hypergraph attention network. Additionally, the dual graph-based product feature learning module enhances cognitive feature extraction by incorporating product co-interactions. DeepSC is optimized through a joint learning objective, combining point-wise and pair-wise learning with an auxiliary social link prediction task to refine user representations. Experiments on three e-commerce datasets demonstrate that DeepSC significantly outperforms the state-of-the-art recommendation models, highlighting the effectiveness of integrating social capital into social preference learning. Our research advances social recommendation by providing a social capital theory-driven approach to modeling user behavior in digital commerce.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114527"},"PeriodicalIF":6.8,"publicationDate":"2025-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145097011","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-07DOI: 10.1016/j.dss.2025.114528
Yang Du , Biao Li , Zhichen Lu , Gang Kou
The highly volatile nature of the stock market makes predicting data patterns challenging. Significant efforts have been dedicated to modeling complex stock correlations to improve stock return forecasting and support better investor decision-making. Although various predefined intrinsic associations and learned implicit graph structures have been discovered, they have limitations in fully exploring and leveraging both types of graph information. In this paper, we proposed a Hybrid Structure-aware Graph Neural Network (HSGNN) framework. Unlike models that rely solely on predefined or learned graphs, HSGNN utilizes money-flow graphs to complementarily learn implicit graph structures and applies sparse supply-chain graphs to jointly enhance stock return forecasting. Extensive experiments on real stock benchmarks demonstrate our proposed HSGNN outperforms various state-of-the-art forecasting methods, offering a robust decision-support system for financial stakeholders.
{"title":"Modeling hybrid firm relationships with graph neural networks for stock investment decisions","authors":"Yang Du , Biao Li , Zhichen Lu , Gang Kou","doi":"10.1016/j.dss.2025.114528","DOIUrl":"10.1016/j.dss.2025.114528","url":null,"abstract":"<div><div>The highly volatile nature of the stock market makes predicting data patterns challenging. Significant efforts have been dedicated to modeling complex stock correlations to improve stock return forecasting and support better investor decision-making. Although various predefined intrinsic associations and learned implicit graph structures have been discovered, they have limitations in fully exploring and leveraging both types of graph information. In this paper, we proposed a Hybrid Structure-aware Graph Neural Network (HSGNN) framework. Unlike models that rely solely on predefined or learned graphs, HSGNN utilizes money-flow graphs to complementarily learn implicit graph structures and applies sparse supply-chain graphs to jointly enhance stock return forecasting. Extensive experiments on real stock benchmarks demonstrate our proposed HSGNN outperforms various state-of-the-art forecasting methods, offering a robust decision-support system for financial stakeholders.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"198 ","pages":"Article 114528"},"PeriodicalIF":6.8,"publicationDate":"2025-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145050101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号