Competitive intelligence is essential for operations management decision-making. Beyond traditional offline information channels, firms increasingly gather online data and resources to generate comprehensive competitive intelligence. This study derives competitive intelligence in large markets by developing an interpretable machine learning framework that integrates multifaceted user behavior data, including user favorites, user-commented products, and user textual comments. Considering the complementary nature of these data sources, we first combine latent features derived from user favorites and user-commented products to improve submarket inference. Using these inferred submarkets as supervised signals, we connect user-commented products and associated textual comments to uncover consumer perceptions. We estimate the model using multifaceted data on online user behavior in the automotive domain. The results demonstrate that our model effectively improves submarket identification, captures consumer perceptions, and predicts competitive positions for new entrants. The derived competitive intelligence helps managers make more informed decisions in product operations and marketing strategies.

Dataset recommendation is pivotal for streamlining data selection and accelerating scientific discovery. In this study, we propose the Sparse-Link Dataset Recommendation Model (SLDRM), an explainable framework that maps textual content, authors, and datasets into a unified topic space. Specifically, SLDRM captures the correlations among words, research communities, and dataset usage patterns by linking their respective latent topics. To handle the inherent sparsity of the research landscape, we incorporate a Spike-and-Slab prior. We validate our model using a real-world dataset collected from the PapersWithCode website. Experimental results show that our model not only improves recommendation accuracy but also enhances interpretability. The proposed model provides researchers with an efficient tool for dataset discovery and deepens the understanding of the knowledge production process in scientific networks.

Enterprise Systems (ESs) embed industrial best practices into adopting organizations through the usage of their system features. Yet, employees often engage in workaround use, either internal or external to the ES, which does not conform to prescribed use, but nonetheless could be beneficial for accomplishing work tasks. Building on Cultural Tightness-Looseness theory and research on system workarounds, we develop a model examining how group cultural tightness-characterized by strong adherence to enforced norms within a group-shapes employees' conforming and workaround use of ES and how these different forms of usage influence job performance. To test the hypotheses, we employ a mixed-method approach with Study 1 leveraging multilevel, longitudinal, and multisourced data collected from 228 employees within 57 groups in a Chinese company, along with Study 2 utilizing an online experiment involving 240 participants from the United States. Collectively, the studies provide compelling evidence supporting our research model, indicating that group cultural tightness plays an instrumental role in increasing conforming use while decreasing internal and external workaround use for individual employees across different national contexts. The findings of Study 1 further indicate that both conforming and internal workaround use have positive effects on employees' job performance, whereas external workaround use negatively impacts their performance. An additional study, using a comparable research design and surveying 220 employees across 59 groups within a foreign multinational corporation operating in China, yields similar results, supporting the generalizability of these findings across different organizational settings. Findings from our study thus provide generalizable insights into the relative influence of group cultural tightness on conforming and internal and external workaround use of ES, as well as the distinct effects of these different modes of system usage on job performance.

Web attacks are one of the major and most persistent forms of cyber threats, which bring huge costs and losses to web application-based businesses. Various detection methods, such as signature-based, machine learning-based, and deep learning-based, have been proposed to identify web attacks. However, these methods either (1) heavily rely on accurate and complete rule design and feature engineering, which may not adapt to fast-evolving attacks, or (2) fail to estimate model uncertainty, which is essential to the trustworthiness of the prediction made by the model. In this study, we proposed an Uncertainty-aware Ensemble Deep Kernel Learning (UEDKL) model to detect web attacks from HTTP request payload data with the model uncertainty captured from the perspective of both data distribution and model parameters. The proposed UEDKL utilizes a deep kernel learning model to distinguish normal HTTP requests from different types of web attacks with model uncertainty estimated from data distribution perspective. Multiple deep kernel learning models were trained as base learners to capture the model uncertainty from model parameters perspective. An attention-based ensemble learning approach was designed to effectively integrate base learners' predictions and model uncertainty. We also proposed a new metric named High Uncertainty Ratio-F Score Curve to evaluate model uncertainty estimation. Experiments on BDCI and SRBH datasets demonstrated that the proposed UEDKL framework yields significant improvement in both web attack detection performance and uncertainty estimation quality compared to benchmark models.

This study constructs an explainable recommendation via combining product images, textual descriptions, user reviews, and user-item interactions. We propose a theory-based multimodal deep learning architecture. Specifically, we first extract the element-level features from product display information, including region-level visual and word-level textual features. To measure the impacts of these element-level features on user preferences, we introduce attention mechanisms. In addition, we inject the user reviews to disentangle the effects of visual and textual features on user preferences. Finally, we introduce a stick-breaking method to measure the asymmetrical influence of images and textual descriptions at the holistic level. To evaluate our model’s utility, we conduct experiments from four perspectives: recommendation performance, explanatory analysis, mechanism, and robustness. Experimental results show that our model can improve recommendation performance and give explanations. Our findings provide valuable insights for recommendation system design and offer guidance to marketers for optimizing product display pages

In recent years, virtual humans have been extensively used in the tourism industry to interact with tourists to enhance destination marketing effectiveness and improve services. However, there is limited research on the impact of virtual humans’ language use on tourists in interaction, especially the use of accents. To explore whether virtual tourist assistants’ accents affect tourists’ visit intentions, and the mechanisms through which destination accents and origin accents affect tourists’ visit intentions, we drew on symbolic interactionism theory and developed two mediation models. Two online experiments (Studies 1 and 2) and one laboratory experiment (Study 3) were conducted to test the hypotheses. The results showed that compared to a standard accent, the use of destination and origin accents by virtual tourist assistants increased tourists’ visit intentions. Perceived novelty and perception of destination distinctiveness serially mediated the effect of destination accent on visit intention. Wow-effect and destination involvement serially mediated the effect of origin accent on visit intention. Furthermore, the results revealed the moderating effects of destination type (natural vs. cultural) and need for cognition. Specifically, when the destination was a natural destination, the indirect effect of destination accent on visit intention was significant. However, the indirect effect was not significant when the destination was a cultural destination. When tourists have a high need for cognition, the indirect effect of origin accent on visit intention was augmented. These findings provide destination marketing organizations with insights into the utilization of accents.

While consumer-generated reviews deliver substantial business intelligence value for advancing recommender systems, they also create an attack surface in review-based recommender systems (R-RSs). Subtle textual perturbations through review tampering, forgery, and other adversarial attacks can manipulate recommendation outcomes. Nevertheless, prior scholars primarily focus on numerical ratings or visual-input adversarial manipulations, offering limited guidance for R-RSs that rely on discrete and semantically interdependent textual reviews. Following adversarial robustness theory, we develop a computational design science framework that jointly assesses and enhances adversarial robustness for R-RSs. For assessment, we design a novel Shapley value-guided Adversarial Review Generation (SARG) method that mimics rigorous adversarial environments by generating adversarial reviews aligned with the manipulation objective. For enhancement, we design an Attack-Lifecycle Robustness Enhancement (ALRE) method aligned with the attack reconnaissance and execution lifecycle, integrating stochastic recommendation process to reduce reconnaissance-phase information leakage, sensitivity-aware input dropout with certified robustness bounds to reduce overreliance on high-sensitivity tokens and adversarial contrastive retraining to strengthen representation robustness during attack execution-phase. Extensive experiments on Amazon and Yelp datasets demonstrate the severity of adversarial vulnerabilities in existing R-RSs and the effectiveness of our framework across diverse attack settings. This study advances the IS literature by offering a unified approach to assessing and enhancing adversarial robustness in R-RSs, with broader implications for online platforms reliant on user-generated textual content.

Fake news on social media platforms poses a significant threat to societal systems, highlighting the urgent need for advanced detection methods. The existing detection methods can be divided into machine-intelligence-based, crowd-intelligence-based, and hybrid-intelligence-based methods. Among these, hybrid-intelligence-based methods achieve the best performance but fail to consider the uncertainty issue in detection. In light of this, we propose a novel uncertainty-aware hybrid-intelligence (UAHI) method for fake news detection. Our method comprises three integral modules. The first module employs a Bayesian deep learning model to capture the inherent uncertainty within machine intelligence. The second module uses an item response theory-based user response aggregation to account for the uncertainty in crowd intelligence. The third module introduces a new distribution fusion mechanism, which takes the distributions derived from both machine and crowd intelligence as input and outputs a fused distribution that provides predictions along with the associated uncertainty. Experiments on the two datasets demonstrate the advantages of our method. This study has practical implications for three key stakeholders: internet users, online platform managers, and the government.

This study focuses on multimodal topic modeling and attempts to separate public topics (shared across modalities) from private topics (unique to each modality) hidden in text and image data. To address this issue, we propose a novel Disentangled Multimodal Neural Topic Model (DMNTM). Specifically, we design the modality-specific encoder with an independence constraint to capture private topics, and the public encoder with a product-of-experts module to extract cross-modal shared topics. We conduct extensive experiments on six public datasets, including multimodal online reviews from Amazon, posts from Flickr, tweets from Twitter, and webpages from Wikipedia. Compared with state-of-the-art methods, we find that DMNTM significantly improves topic modeling performance in terms of perplexity, coherence, diversity, and topic quality over the best baseline. In two downstream tasks, including recommendation and sentiment classification, DMNTM further improves the performance. These results show that disentangling public and private topics effectively enhances both the quality and utility of multimodal representations.

With rapid advances in technology, marketers are eager to adopt innovative promotion tools to give consumers a detailed understanding of products. Augmented reality (AR) and live streaming are emerging technologies that retailers are focusing on, but it is unclear how they will affect consumers and what the difference is. Based on the functional mechanisms of online product presentations, we examined how AR and live streaming affect online consumers’ product quality and fit uncertainty, and the differences in their influence mechanism. A between-subject experiment (N = 553) was conducted. The findings show that (1) AR, live streaming, and AR + live streaming all significantly enhance interactivity and vividness. However, the three presentation modes differ in their paths of impact, and their combined use is not superior to the individual use; (2) the sequence of interactivity/vividness and spatial presence/social presence mediates the relationship between AR or live streaming and product uncertainty; and (3) product type moderates the relationship between product presentation mode and product uncertainty. This study reveals the influence mechanisms and boundary conditions of AR and live streaming on product uncertainty, providing insights into how online retailing can effectively utilize these two technologies.

Price promotions have been widely implemented by online group buying restaurants. However, the set menu, a popular form of price promotion, has received limited academic attention, particularly regarding its effectiveness compared to cash coupons in stimulating consumer reviews. Drawing upon mental accounting theory, this study investigates how set menus and cash coupons affect consumer review sentiment. Specifically, we propose that compared to cash coupons, set menus are more likely to generate less positive review sentiment. In addition, the influence of these price promotions on review sentiment is contingent on consumer regulatory focus and coupon price. To test the research model, we collected 4003 consumer reviews and transaction data from a leading online group buying platform. The findings show that set menus have a less positive impact on review sentiment, whereas cash coupons are associated with more positive sentiment. Moreover, prevention-focused consumers buying set menus with a low coupon price tend to post more positive reviews than promotion-focused consumers paying a high coupon price. Our study adds to the tourism and hospitality literature on price promotions and mental accounting, and provides guidance for online group buying marketers in designing effective promotions.

Large Language Models (LLMs) have brought unprecedented innovation opportunities to the marketing field. However, the practical applications of LLMs within the marketing landscape currently exhibit a fragmented and scattered nature. In this study, we aim to aggregate these scattered literature to create a holistic view of LLMs capabilities for marketing research. Specifically, we present an overview of LLMs using the evolution of LMs. Subsequently, we explore their application in the marketing domain across five distinct dimensions: data annotation, idea inspiration and content generation, substitution of human participants, user behavior learning and prediction, and evaluation of LLM feedback. Finally, we discuss the new trends and challenges for LLMs in marketing. This study enriches the theoretical foundations of integrating generative AI with marketing practices.

Daily-deal platforms closely cooperate with local retailers when issuing daily-deal coupons to profit from selling coupons online and redeeming them offline. However, most research on daily-deal business has only focused on online sales or the offline redemption process. We investigate the coherent two-phased process from selling coupons online to redeeming them offline, grounded in the lens of social judgment theory, to capture the full picture of the daily-deal business. By tracking the sales and redemption of 11,290 deals over a 13-month period on an online daily-deal platform and conducting various data analyses, we find that reputation and price curvilinearly affect the sold online of daily-deal coupons, which consequently positively affects coupon redemption offline. More specifically, the U test empirically indicates that the extreme point of the inverted U-shaped effect of reputation score is 86.0035 within the range [49.7353, 92.7551]. And the extreme point to price demonstrates a U-shaped effect is 399.6082 within the range [4.7060, 829.3651]. We further classify retailers’ daily deals into consumption on a group or individual level. Empirical data demonstrate that the inverted U-shaped effects of reputation and the U-shaped effects of price are weakened by group consumption. Furthermore, we investigate the moderating role of agglomeration on the relationship between daily-deal coupons sold online and redemption offline of daily-deal coupons. We also discussed the theoretical and practical implications.

Music recommender systems play a critical role in music streaming platforms by providing users with music that they are likely to enjoy. Recent studies have shown that user emotions can influence users' preferences for music moods. However, existing emotion-aware music recommender systems (EMRSs) explicitly or implicitly assume that users' actual emotional states expressed through identical emotional words are homogeneous. They also assume that users' music mood preferences are homogeneous under the same emotional state. In this article, we propose four types of heterogeneity that an EMRS should account for: emotion heterogeneity across users, emotion heterogeneity within a user, music mood preference heterogeneity across users, and music mood preference heterogeneity within a user. We further propose a Heterogeneity-aware Deep Bayesian Network (HDBN) to model these assumptions. The HDBN mimics a user's decision process of choosing music with four components: personalized prior user emotion distribution modeling, posterior user emotion distribution modeling, user grouping, and Bayesian neural network-based music mood preference prediction. We constructed two datasets, called EmoMusicLJ and EmoMusicLJ-small, to validate our method. Extensive experiments demonstrate that our method significantly outperforms baseline approaches on metrics of HR, Precision, NDCG, and MRR. Ablation studies and case studies further validate the effectiveness of our HDBN. The source code and datasets are available at https://github.com/jingrk/HDBN.

Background:With the development of online health care platforms, patient reviews have become an important source for assessing medical service quality. However, the critical aspects of quality dimensions in textual reviews remain largely unexplored. Objective:This study aims to establish a comprehensive medical service quality assessment framework by leveraging online review data. Such a framework would support large service providers, such as online platforms, to assess the quality of many doctors efficiently. Methods:We adopted a text-mining approach with theory-driven topic extraction from online reviews to develop a service quality assessment framework. The framework is based on topic and sentiment classification methods. We conducted an empirical analysis to assess the validity of the framework. Specifically, we examined if patients' sentiments regarding our extracted dimensions affect demand (number of consultation requests) due to quality signals reflected in these dimensions. Results:We develop a 5-dimensional health care service quality framework (HSQ-5D model). In the empirical study, patient demand is affected by these dimensions, including expertise (coefficient=1.12; P<.001), service delivery process (coefficient=5.60; P<.001), attitude (coefficient=0.82; P<.001), empathy (coefficient=2.65; P<.001), and outcome (coefficient=0.26; P<.001; through patients' perceived quality from reviews). The 5 dimensions can explain 85.52% of the variance in patient demand, while all information from online reviews can explain 85.67%. The results show the validity and the potential practical value of the proposed HSQ-5D model. Conclusions:This study explores how online reviews can be used to evaluate health care services, offering significant implications for health care management. Theoretically, we extend existing service quality frameworks by integrating text-mining analysis of online reviews, thereby enhancing the understanding of service quality assessment in the digital health context. Practically, the framework can allow health care platforms to identify and reveal doctors' service quality to reduce patients' information asymmetry and strengthen patient-provider relationships, ultimately contributing to a more effective and patient-centered health care system.

The advent of Artificial General Intelligence (AGI) has heralded a new era in e-commerce, empowering recommendation systems with its advanced capabilities, yet concerns about fairness in these systems have emerged. This paper presents a comprehensive study examining user gender fairness in various recommendation algorithms and domains, with a particular focus on AI-enabled and LLM-based recommendation systems. Concretely, we conduct experiments on four datasets from distinct domains to evaluate and compare the gender fairness of eleven recommendation models from six families under several fairness metrics, such as Absolute Difference, Item Coverage, and Gini coefficient. Our findings reveal significant disparities in recommendation accuracy and diversity between male and female users, highlighting the need for fair and unbiased recommendation services in e-commerce. Notably, the latest LLM-based recommendation model demonstrates promising fairness in terms of Item Coverage and Gini coefficient between male and female users, suggesting its potential in mitigating gender bias in recommendations. This study contributes to the understanding of gender fairness in different families of recommendation systems and provides insights for recommendation system design in e-commence platforms.

Virtual humans have been widely adopted to serve as alternatives to human staff for user services. Although there is extensive research on the effect of virtual human realism on user responses, it remains unknown how virtual humans conveying virtual social touch affects user perceptions. Our study shows that the presence (vs. absence) of virtual social touch significantly triggers stronger perceived creepiness and intrusiveness toward virtual humans, thereby having a negative effect on the intention to interact. This effect is moderated by the individual social anxiety as well as the virtual human appearance and language style. For users with lower social anxiety, virtual humans with anime-like appearance and using informal language style, this negative perception caused by virtual social touch can be alleviated. These findings enrich the application of mind perception theory and provide theoretical and practical insights for the design and use of virtual humans in marketing interactions.

Despite the growing popularity of virtual live streaming, limited research has explored how the alignment between a virtual anchor's traits and a brand's image affects consumer outcomes. In this study, we conduct three lab experiments to examine the effect of matching the virtual anchor's appearance, voice, and language style with a brand's image on consumer purchase intention. We identify two key mediators-processing fluency and perceived affinity-that explain how this alignment influences purchase intention. Our findings show that both the fit between the virtual anchor's appearance and the brand image, as well as the fit between the anchor's voice and the brand image, positively affect consumers' purchase intention. When it comes to language style, figurative language increases purchase intentions for brands perceived as warm, while literal language does not enhance purchase intentions for brands perceived as competent. Additionally, processing fluency and perceived affinity mediate the relationship between virtual anchor appearance-brand image fit and consumer purchase intention. Specifically, processing fluency mediates the effect of the fit between the anchor's voice and brand image on purchase intention, while perceived affinity plays a moderating mediating role between the warm and competent brand images. Perceived affinity also mediates the effect of language style fit on purchase intention, while processing fluency does not serve as a mediator in this context.

Composed image retrieval (CoIR) involves a multi-modal query of the reference image and modification text describing the desired changes, allowing users to express image retrieval intents flexibly and effectively. The key of CoIR lies in how to properly reason the search intent from the multi-modal query. Existing work either aligns the composite embedding of the multi-modal query and the target image embedding in the visual domain through late-fusion or converts all images into text descriptions and leverage large language models (LLM) for text semantic reasoning. However, this single-modality reasoning approach fails to comprehensively and interpretably capture the users’ ambiguous and uncertain intents in the multi-modal queries, incurring the inconsistency between retrieved results and ground truth. Besides, the expensive manually annotated datasets limit the further performance improvement of CoIR. To this end, this article proposes an LLM-enhanced Intent Uncertainty-Aware Linguistic-Visual Dual Channel Matching Model (IUDC), which combines the strengths of multi-modal late-fusion and LLMs for CoIR. We first construct an LLM-based triplet augmentation strategy to generate more synthetic training triplets. Based on this, the core of IUDC consists of two matching channels: the semantic matching channel is responsible for intent reasoning on the aspect-level attributes extracted by an LLM, and the visual matching channel accounts for the fine-grained visual matching between multi-modal fusion embedding and target images. Considering the intent uncertainty presented in the multi-modal queries, we introduce Probability Distribution Encoder (PDE) to project the intents as probabilistic distributions in the two matching channels. Consequently, a mutually enhanced module is designed to share knowledge between the visual and semantic representations for better representation learning. Finally, the matching scores of two channels are added to retrieve the target image. Extensive experiments conducted on two real datasets demonstrate the effectiveness and superiority of our model. Notably, with the help of the proposed LLM-based triplet augmentation strategy, our model achieves a new record of state-of-the-art performance among all datasets.

Since the 1990s, China has gradually implemented price reform. The increasing-block pricing policy (referred to as "IBP") began to be applied in different kinds of household resource-based products so as to produce resource-saving effect. However, when evaluating policy effects, the information factors are often ignored. This paper concentrates on evaluating the residential increasing-block water pricing policy (referred to as "IBWP") and divides the information into two categories: policy information and bill information. Among them, policy information is divided into implementation information and rule information, and bill information is subdivided into three aspects: price, volume and fee. Based on the household data of the Chinese Household Water Use Behavior Survey 2019, this paper uses the propensity score matching (PSM) method to investigate the impact of households' information cognition on water consumption. The results show that residents with policy information will increase their water consumption, while the availability of rule information will reduce the increase in water consumption; bill information encourages households to take behavioral water-saving measures more frequently, and among different types of bill information, residents are the most sensitive to water fee information. Through further analysis, the shadow price of household water is much higher than the current water price. Hence it can be considered that the current water price is distorted to a large extent, which leads to the deviation of the policy information received by consumers. Therefore, the water-saving effect of IBWP at the present stage is very limited. The policy implication shows that IBWP needs to further optimize its structure, the government should gradually increase the current residential water price; relevant departments ought to strengthen the policy publicity, and improve the acquisition rate of previous bills for residents.