Information Systems Frontiers最新文献

Building on the perspectives of the uses & gratification (U&G) theory and stimulus-organism-response (S–O-R) model, this article develops and tests an integrative framework to examine the underlying factors influencing customers’ experiences with chatbots as a form of virtual conversational agent (VCA) in the UK and Vietnam. In addition to utilitarian and hedonic factors, anthropomorphism and social presence are also investigated, which are considered important experiential dimensions in a customer-machine relationship. We also explore how stimuli such as functionality, communication style similarity, and aesthetics indirectly affect outcomes like customer satisfaction and reuse intention, mediated by four types of customer experiences. Data collected from a sample of 417 and 359 participants in the UK and Vietnam respectively revealed that, in general, perceived informativeness, credibility, enjoyment, functionality, and communication style similarity are crucial for customer satisfaction in both countries. Interesting differences in the effects of customer experience between developed and developing countries were observed. For instance, the effects of anthropomorphism and social presence on satisfaction are only effective for customers from developed country, while those from developing country only need information provided by chatbots be transparent. Our findings offer a novel way to understand customer experience with chatbots and provide important theoretical and managerial implications.

Online health communities (OHCs) offer emotional and informational support to their users. However, past research has primarily treated these supports as separate, but they coexist in messages, making it essential to consider the emotional valence of text to understand the support being provided. This study examines how aligning questions and responses in OHCs reduces information gaps, and enhances support quality and perceived helpfulness. We use a labeled data set of question-response pairs to develop multimodal machine learning models to predict support interactions. Using explainable AI, we reveal the emotions within support exchanges, underscoring how emotional valence in the text determines informational support in OHCs and provide insight into the interaction between emotional and informational support. This study refines social support theory and establishes a foundation for decision aids and emotion-sensitive AI systems to deliver personalized social support tailored to users’ informational and emotional needs.

As per agenda-setting theory, political agenda is concerned with the government’s agenda, including politicians and political parties. Political actors utilize various channels to set their political agenda, including social media platforms such as Twitter (now X). Political agenda-setting can be influenced by anonymous user-generated content following the Bright Internet. This is why speech acts, experts, users with affiliations and parties through annotated Tweets were analyzed in this study. In doing so, the agenda formation during the 2019 European Parliament Election in Germany based on the agenda-setting theory as our theoretical framework, was analyzed. A prediction model was trained to predict users’ voting tendencies based on three feature categories: social, network, and text. By combining features from all categories logistical regression leads to the best predictions matching the election results. The contribution to theory is an approach to identify agenda formation based on our novel variables. For practice, a novel approach is presented to forecast the winner of events.

Network slicing is a 5G concept that virtualizes the physical network infrastructure to accommodate multiple service requirements on the same network, where each slice manages diverse needs and ensures their coexistence. In this work, we leverage blockchain technology to strengthen the security of handover authentication (HA) processes in network slicing systems.The proposed system addresses the challenge of reducing latency during handovers by incorporating a hybrid on-chain/off-chain model, optimizing the balance between security and speed. It employs the Raft consensus mechanism, which offers lower latency compared to more traditional consensus protocols such as PBFT. It establishes a decentralized registry for recording transfer events, streamlining user equipment (UE) identification verification, and improving HA efficiency. Moreover, we also introduce a three-component model: network slicing, user environments, and a Hyperledger Fabric (HLF) blockchain for authentication and authorization, which enhances the user experience by minimizing delays, ensuring data privacy, and providing scalability. By leveraging edge computing in conjunction with network slicing, the system further reduces latency, making it more efficient for real-time applications in dynamic mobile environments. Performance experiments indicate satisfactory scalability and maintained service quality under increasing throughput, affirming the suitability of the HLF-based system for managing network scenarios. Furthermore, the system’s modular design ensures compatibility with existing authentication protocols, such as AKA and EAP, enabling seamless integration with legacy systems. Consequently, this work enhances network security and service quality, especially in network slicing, HA, and employing HLF for privacy and security solutions. As 5G networks continue to evolve toward 6G, this system’s scalability and flexibility offer a promising approach to addressing future challenges in secure and efficient handover authentication.

With the rapid development of information technology, the gig labor marketplace is fast growing, with digital platform-based instant messaging (IM) playing an important role in raising freelancers’ orders, serving the intention for the crowdsourcing platforms to increase capacity to balance supply and demand. Using a large-scale field experiment on a crowdsourcing freight platform, this study investigates the impact of IM on freelancers’ response rate of orders. Our findings suggest the effects of IM depend on its content and information richness level. Task-relevant information in IM increases the freelancers’ response rate, especially for the priority commitment information, compared with order price information. In addition, although adding task-irrelevant information in IM decreases the freelancers’ response rate, it does not mean the less task-irrelevant information results in a weaker negative IM effect. Rather than that, including task-irrelevant information with a medium information richness level in IM harms the freelancers’ response to the most significant extent. Moreover, our findings reveal crowdsourcing platforms’ actions of IM to increase freelancers’ response rate are consistent with the actions to improve the order acceptance rate, thus demonstrating the critical role of increasing freelancers’ response rate in raising their interest in the final acceptance of the order serving. Our findings guide crowdsourcing platforms to design effective digital platform-based IMs to communicate with freelancers to arouse their response and interest in serving the orders. The capacity of crowdsourcing platforms thus can be dynamically adjusted and expanded to benefit their profitability.

This paper provides a comprehensive examination of the ongoing debate surrounding Artificial Intelligence (AI) and its societal implications, with a particular focus on job displacement. The release of generative AI tools for public use, particularly ChatGPT, has created numerous concerns on how these tools will be used and adverse impacts on society. Augmented Intelligence has been introduced as a concept utilizing AI to enhance human capabilities but its distinction as an assistive role is ill-defined. This research provides insights into the reconceptualization of AI as Augmented Intelligence examining their differences in terms of knowledge development, decision-making, and outcomes. Through three case studies, we demonstrate the assistive role of Augmented Intelligence and how it can serve as a catalyst for job creation and cognitive enhancement. We also explore the impact of AI and IA tools as a sociotechnical system and their effect on human cognitive abilities through the theoretical lens of the Dunning Kruger Effect. We conclude with a research agenda to stimulate future directions of research.

Online reviews are effective information-sharing tools due to their word-of-mouth characteristics. The extant literature has considered reviews as independent variables that influence business performance, while the environmental factors shaping these reviews remain under-explored. We examine the impact of COVID-19-related environmental uncertainties on changes in review prosumption (production and consumption) behavior. Based on the stimulus-response theory, with COVID-19 as the stimulus and prosumption as the response, we examined the changes in the characteristics of online reviews. Using the difference-in-differences methodology, we analyze online reviews of restaurants in two US cities that experienced different levels of COVID-19 impact. On the production side, we find an increased use of contextual terms and negative sentiments. On the consumption side, we find an increase in review usefulness and a decline in funniness. The results are robust, supported by coarsened exact matching and falsification tests. We conclude with a discussion of the study’s implications and contributions.

A recent trend in data management research investigates whether machine learning techniques could improve or replace core components of traditional database architectures, such as the query optimizer or selectivity and cardinality cost estimators. The preliminary approaches leverage cost-based optimizers and cost models to avoid a cold-start as they train and build learning models. In this work, we investigate whether learning could also be beneficial in rule-based optimizers, which instead of driving query execution decisions via a cost model they rely on a set of fixed rules and pre-defined heuristics. Our experimental testbed employs MonetDB, an open-source, column-store analytics data engine, and explore whether a learning model using Graph Neural Networks (GNNs) that is trained on a cost-based engine, such as PostgreSQL, could improve MonetDB optimizer’s decisions. Our initial findings reveal that our approach could improve significantly MonetDB’s query execution plans, especially as the query complexity increases whet it involves many join operators.

The well-organized structure of Policy Texts (PTs) is fundamental to intelligent governance, yet most PTs lack fine-grained category labels. PTs from different domains follow different classification systems, and traditional encoder-only models cannot directly handle scenarios where the label spaces of the source and target domains differ significantly, as their output layer typically is a fixed-dimensional classification head. Therefore, we propose a Cross-Domain Policy Text Classification (CDPTC) task. We introduce a method for the task called InstructCDPTC. This method, within an instruction tuning framework, transforms the classification task into a generation task, using the decoder-only model BigBird to predict masked tokens. We wrap the original PT within an instruction template containing a task description, a label description, and a mask sequence, which serve as input to BigBird. During training, we use the names of gold categories as the prediction targets for masked positions. During inference, we determine the final predicted category by computing the semantic distance between the averaged representations of the mask predictions and each candidate label. We constructed a dataset of 20,189 labeled policy texts from five different policy domains to evaluate InstructCDPTC. Experimental results demonstrate that InstructCDPTC achieves an F1 score of 0.824 under conditions where the sample distribution and label space of the target domain are entirely unseen, surpassing other baselines.

Shared services using digital platforms have increasingly gained prominence in recent times. Existing studies have studied several facets of ride-sharing services, but mobile app technology’s impact on user’s experience has not been explored meticulously. We attempt to study the technological artifacts which can signal about the capability of the service and thereby, reducing the informational asymmetry, stemming from lack of information and in-person communication. To address that, we adopt the Signaling Theory and Value Framework to understand the apps’ features, reflecting the shared mobility service quality to the users. We mine 212,000 and 150,000 user reviews on India’s two most extensively used shared mobility services- OLA and UBER, respectively and identify the factors affecting user experiences. We provide a novel framework by mapping these factors to theoretical lexicons. Multiple regression models show that time resources, monetary resources, perceived information protection, app usage controllability, perceived safety in e-payment mechanism, informational trust-related advantage, and participation in decision making influence the user experience of both the services significantly.