Human–artificial intelligence (AI) augmentation (HAA) is widely expected to enhance trust in AI-supported decision-making by combining algorithmic processing with human judgment. However, existing research on trust in HAA has focused primarily on AI users, providing limited insight into how trust is formed among affected groups that do not directly interact with the systems involved. This study develops a theory of asymmetric trust production, which proposes that, although HAA may strengthen trust among AI users, it can simultaneously constrain trust formation among affected groups by creating structural vulnerability, particularly in contexts characterized by equivocality and invisibility. This study employed a two-part design. Study 1 used a vignette experiment involving hiring, adjudication, and scholarship selection scenarios. The results showed that, under unfavorable outcomes, affected groups reported lower trust under HAA than under human-only decision-making. Higher perceived equivocality and invisibility were also associated with lower trust, particularly in pooled cross-scenario analyses. Study 2 used survey data from job applicants and structural equation modeling (SEM) to identify alternative trust pathways. Findings showed that institutional transparency and social influence shaped appealability, procedural fairness, and institutional accountability, which, in turn, supported trust in AI and HAA decisions, although several paths received only partial or no support. These findings extend research on trust in AI beyond direct users to structurally affected stakeholders by identifying both the conditions under which HAA can constrain trust and the institutional pathways through which trust may still be sustained. This study offers implications for the responsible design and governance of AI-supported decision-making.
Online knowledge collaboration among physicians in team services has become the key to improving the quality of online healthcare services. However, existing research lacks a discussion of the classification of different types of knowledge collaboration behavior and their impact on team performance. To fill this gap, this study collected public data from China’s leading online medical communities from December 2022 to October 2023, forming a panel dataset covering 510 teams, 2559 physicians, and 130,000 unstructured physician–patient interaction records. Based on the shared mental model, we first distinguished two modes of online knowledge collaboration: knowledge integration and knowledge validation. Second, our two-way fixed effects model revealed that both collaboration modes contribute to team economic returns, while knowledge validation is particularly important for enhancing online patient experience, especially when facing complex medical tasks. Furthermore, communication directionality strengthens the positive relationship between knowledge collaboration and team performance. This study provides valuable guidance for online medical teams to optimize their knowledge collaboration strategies based on specific performance objectives.
While research provides rich and detailed insights into how information technologies (IT) can foster mental health, concerns remain on how to conceptualize different types of mental health IT and explain the ways in which they support varying mental health related information needs. In response, we review extant literature on the use of IT to foster mental health. Guided by an information services lens, we identify three mental health IT configurations (self-support, group-support, expert-support) and describe how they apply to four mental health IT requirements (information, screening, treatment, and monitoring). Further, we define four IT capabilities (informational, exploratory, networking, and collaborative) and describe how they address mental health IT requirements of varying levels of uncertainty and equivocality. In conclusion, we discuss the contributions and implications of the resulting integrated theoretical framework, highlighting how it can help users evaluate and compare IT solutions for mental health, select appropriate solutions for specific mental health IT needs, and develop new and innovative solutions to better support mental health.
In an open innovation community, the popularity of an idea reflects the judgment by the community of the potential value of the idea and is a key precursor to idea implementation; however, the factors that drive the popularity of ideas remain underexplored. Drawing on Campbell’s theory of blind variation and selective retention, in this study, the novelty of the ideas submitted during the blind variation phase and the constructiveness of the feedback provided during the selective retention phase are posited to constitute two dimensions of knowledge contribution. By cluster analysis, four knowledge contribution patterns (i.e., user roles), namely, balanced contributors, creative innovators, constructive feedback providers, and minimal contributors, are identified. Community participation by users has dual effects: Internally, it shapes their cognitive frameworks and innovation capabilities, and externally, it leads to the accumulation of visible community indicators (e.g., points) that function as attention cues. By integrating the theory of limited attention, in this study, we examine how user roles formed in the early stages of participation influence idea popularity and how attention cues serve as boundary conditions that moderate this relationship. Using panel data from Xiaomi’s MIUI Forum, we observe that balanced contributors generate more popular ideas than do users in all other roles, despite not excelling on any single knowledge contribution dimension. Point gains are positively associated with idea popularity, with balanced contributors benefitting more from point gains than minimal contributors and creative innovators do. Theoretical and practical implications are discussed.
Session-based recommendations (SBRs) play a pivotal role in reducing users’ information overload and increasing sales and revenue on e-commerce platforms. Understanding users’ behavioral patterns and fine-grained preferences is crucial for the development of effective SBRs. Drawing insights from consumer behavior literature, we examine two key behavioral states—fixed and exploratory states—that characterize users’ distinct behavioral patterns in evaluating product attributes. Specifically, the fixed state reflects a behavioral tendency to maintain consistency with past choices, while the exploratory state captures users’ inclination to seek diverse and novel options. Methodologically, we propose a novel Attribute-aware Session-based Recommendation method that distinguishes between the Fixed and Exploratory states, called ASR-FE. Our method features three novel components: (1) a time-weighted sequence variance mechanism that identifies users’ fixed and exploratory states for each attribute sequence, and two heterogeneous decoders—(2) a fixed state decoder and (3) an exploratory state decoder—to predict users’ attribute preferences based on the corresponding states. Extensive experiments on two real-world datasets demonstrate that incorporating the fixed and exploratory states not only enhances recommendation accuracy but also improves the interpretability of our model, yielding crucial implications for e-commerce platforms.
