Authors: Rai, A., Tian, J., and Xue, L.
Abstract: Artificial intelligence (AI)-automated decision systems encounter persistent, interdependent, and dynamic fairness tensions that traditional one-off interventions cannot resolve. Because these tensions persist due to interdependence and dynamic interaction, organizations require both a theory of the problem to explain their persistence and a theory of the solution to prescribe how they can be managed. Our design theory, FAIR (Fairness Adaptation through AI-augmented Responsiveness), provides a theory of the problem by reframing AI fairness as a sociotechnical paradox constituted within AI artifacts that automate decision tasks, through interdependent organizational, technical, and governance choices and their interaction with regulatory mandates and societal norms. Synthesizing four fairness perspectives (Ethics, Organizational Justice, Economic Fairness, and Rawlsian Justice), we identify three metatheoretical dimensions (principles, goals, foci) and show that the interdependence within and among these dimensions is the root, endogenous source that constitutes paradoxical fairness tensions. Building on this diagnosis, FAIR provides a theory of the solution by specifying an organizational capability grounded in three design foundations. First, the paradox lens motivates iterative adaptive cycles (Surfacing and Resolving) to continually surface and resolve AI fairness tensions. Second, design science in information systems and computer science distinguishes AI artifacts (the “what”) from the actors (the “who”) responsible for adapting them, establishing the basis for complementary human–AI agent collaboration in the adaptive cycles: AI agents execute monitoring to surface and refinement to resolve tensions, whereas human agents specify objectives, adjudicate trade-offs, and exercise contextual judgment and oversight. Third, the managing-with-AI literature informs how this human–AI agent collaboration should be governed. These foundations yield two reinforcing mechanisms: (i) artifact-level adaptation, achieved through structured human–AI agent collaboration, within and across the layers of the AI decision pipeline—Representation (data), Learning (model), and Calibration (decision); and (ii) portfolio-level, risk-tiered federated governance that structures how human–AI agent collaboration scales across tasks and artifacts, balancing process standardization with configuration choices and human control with AI autonomy based on task risk. Enabled by organizational “fairness complements”—namely, human skills to work with AI agents and structured stakeholder feedback—this sociotechnical design provides organizations with a sustained capability to harmonize global coherence and local flexibility in the responsive adaptation of AI fairness.

Authors: Zhao, K., Xiao, H., Rai, A., & Kim, J.
Abstract: Human mobility prediction underpins diverse applications in marketing, transportation, healthcare, network management, and public safety. Accurately forecasting movement patterns requires capturing both sequential regularities—habitual commutes and recurring visits—and contextual factors—dynamic influences such as weather, events, and social interactions that can reinforce or disrupt routine. This article presents a dual-perspective survey of mobility prediction, categorizing models based on their focus on routine, context, or a fusion of both. We review key components of human mobility prediction, explore their applications across multiple domains, survey statistical and machine learning predictors, and empirically evaluate their effectiveness using large-scale mobility data under varying conditions. Based on our empirical results, we offer practical guidelines for selecting mobility prediction techniques suited to different mobility patterns. This survey provides a comprehensive foundation for researchers and practitioners aiming to develop effective human mobility prediction systems across diverse real-world scenarios.

Authors: Lee, O. K. D., Park, Y., Choi, I., & Rai, A.
Abstract: As business competition is getting faster and more complex, taking timely and sufficient competitive actions by holistically utilizing key organizational resources and capabilities is critical for a firm’s survival. By extending the awareness, motivation, and capability (AMC) framework of competitive dynamics with information technology (IT), we investigated context-specific configurational mechanisms that explicate the simultaneous interactions among a firm’s IT and AMC factors for creating competitive actions. Using fuzzy-set qualitative comparative analysis (fsQCA), a set-theoretic method, we empirically analyze field survey data from 189 manufacturing firms. Our analysis uncovered multiple equifinal configurations, revealing nuanced, interdependent relationships among IT infrastructure and applications, awareness, motivation, and operational excellence and innovation capabilities. These relationships are key to generating a high frequency of competitive actions across diverse organizational and environmental contingencies. Based on the findings, we developed theoretical propositions of configurational causal recipes—namely, automation, autonomy, innovation, and integration—that explain which IT-AMC factors matter, how they interrelate, and the ways in which IT factors complement or substitute AMC factors to drive competitive actions within specific contexts of environmental speed, uncertainty, and firm size. Through interviews with top managers of diverse manufacturing companies, we validate the suggested configurational recipes in contemporary business environments. Additionally, we discuss the potential of refining or specializing the recipes to account for the role of emerging digital technologies. Finally, we conclude with the theoretical and practical implications of our findings.

Authors: Chen, L., Rai, A., Wei, W., and Guo, X.
Abstract: Match formation is challenging in online matching platforms where suppliers are subject to dynamic capacity constraints. We provide a theoretical foundation for understanding how online matching platforms support the transmission and triangulation of multisource information for consumers to infer provider service quality and dynamic capacity states, and achieve desirable matching outcomes. Situating this study in the context of an online health consultation community (OHCC) and drawing upon signaling theory, we theorize how physicians’ owned and earned signals influence physicians’ voluntary online consultations with new patients they have not consulted with previously. Importantly, we articulate how these signaling effects are contingent upon physicians’ dynamic capacity in OHCC. We collected longitudinal data from a large OHCC in China and used a hidden Markov model (HMM) to characterize the dynamic physician capacity in the OHCC and test the hypotheses. Our findings reveal that service professionals’ owned and earned signals interactively work together to balance supply and demand dynamically, and thereby facilitating matchmaking. In OHCCs, where physicians provide voluntary service beyond their primary jobs at hospitals, we find that owned and earned signals increase patient consultations in different patterns contingent upon physicians’ capacity states. In addition, we discover the complementary and substitute relationships between owned signals and earned signals change when physicians are in different capacity states. The findings have significant implications for our understanding of online match formation under dynamic capacity constraints and the design of OHCCs.

Authors: Pye, J., Rai, A., and Dong, J.
Abstract: Hospitals have implemented health information technology (HIT) for clinical care to address rising operating costs in recent years. We integrate behavioral and institutional perspectives to explain how hospitals differentiate technological search relative to industry peers (i.e., search differentiation) for HIT portfolios. In the context of the U.S. healthcare industry, we theorize that hospitals’ search differentiation for HIT results jointly from idiosyncratic learning in response to cost-based performance shortfalls and isomorphic pressures in relation to changing policy uncertainty as the Health Information Technology for Economic and Clinical Health (HITECH) Act has unfolded. Based on a panel data set from 3,319 hospitals in 2007–2014, we demonstrate that when costs increase relative to aspiration level, a hospital differentiates its search for HIT by exploring more novel technologies for clinical care relative to peers. As policy uncertainty declines from the conceptualization phase to the enactment phase of the HITECH Act, a hospital’s search differentiation for HIT increases to a greater extent in response to cost-based performance shortfalls as lower uncertainty reduces the need to imitate peers’ search. As policy uncertainty further declines from the enactment phase to the enforcement phase of the HITECH Act and reaches its lowest level, however, the hospital’s search differentiation for HIT increases to a smaller extent in response to cost-based performance shortfalls because of policy incentives and professional norms to promote implementation of common technologies. Overall, we provide a more holistic picture of how uncertainty in a dynamic regulatory context intertwines with organizational learning from performance feedback in shaping search differentiation.

Authors: Mindel, V., Aaltonen, A., Rai, A., Mathiassen, L., and Jabr, W.
Abstract: Although online peer-production systems have proven to be effective in producing high-quality content, their open call for participation makes them susceptible to ongoing quality problems. A key concern is that the problems should be addressed quickly to prevent low-quality content from remaining in place for extended periods. We examine the impacts of two control mechanisms, bots and policy citations, and the number of contributors, with and without prior experience in editing an article, on the cleanup time of 4,473 quality problem events in Wikipedia. We define cleanup time as the time it takes to resolve a quality problem once it has been detected in an article. Using an accelerated failure time model, we find that the number of bots editing an article during a quality problem event has no effect on cleanup time; that citing policies to justify edits during the event is associated with a longer cleanup time; and that more contributors, with or without prior experience in editing the article, are associated with a shorter cleanup time. We also find important interactions between each of the two control mechanisms and the number of different types of contributors. There is a marginal increase in cleanup time that is larger when an increase in the number of contributors is accompanied by fewer bots editing the article during a quality problem event. This interaction effect is more pronounced when increasing the number of contributors without prior experience in editing the article. Further, there is a marginal decrease in cleanup time that is larger when an increase in the number of contributors, with or without prior experience in editing the article, is accompanied by fewer policy citations. Taken together, our results show that the use of bots and policy citations as control mechanisms must be considered in conjunction with the number of contributors with and without prior experience in editing an article. Accordingly, the number of contributors and their experience alone may not explain important outcomes in peer production; it is also important to find an appropriate mix of different control mechanisms and types of contributors to address quality problems quickly.

Authors: Long, Y., and Rai, A.
Abstract: Digital risk—or the likelihood of losses from key digital activities (i.e., information system [IS] sourcing, digital infrastructure, data management, IS applications, IS use, and digital product offerings)—constitutes a key consideration in firm valuation. Firms’ public disclosures (e.g., 10-K reports, earnings conference calls) are a key source of data to learn about digital risks. Although text analytics approaches (e.g., word frequency, topic modeling, and sentiment analysis) have been applied to a firm's public disclosures to assess various types of risk (e.g., political risk, tax risk, cybersecurity), they do not consider the structural linguistic relations embedded in the text that are potentially relevant in measuring risk. We apply a neural network approach to address this gap and extract linguistic relations from a firm's 10-K disclosure (Section “Item 1A”). We develop novel firm-level digital risk measures based on these linguistic relations. Specifically, we measure firm-level digital risk from three perspectives: (1) presence (whether digital risk is mentioned or not), (2) intensity (text coverage of digital risk relative to other issues), and (3) diversity (the types of digital risk mentioned). We validate our digital risk measures by demonstrating their significant correlation with firm risk, proxied by stock market volatility. Our research reveals that investors’ perceptions of digital risk diversity and digital risk intensity differ between IT and non-IT companies. First, across all firms, digital risk intensity is negatively associated with firm risk, indicating that investors do not incorporate intensity of digital risk when assessing firm risk. Second, in non-IT firms, digital risk diversity is positively associated with firm risk, suggesting that managers in these firms may influence investor perceptions through strategic disclosure of digital risk types. Overall, our findings suggest that text-based digital risk measurement is practically feasible, scalable, and economically meaningful.

Authors: Sambhara, C., Rai, A., Xu, S. X.
Abstract: Upcoming.

Authors: Tian, H., and Rai, A.
Abstract: As digital platforms evolve, app developers do more than passively participate; they actively reshape the competitive landscape through boundary-shifting moves (BSMs). By modularizing boundary resources at different levels of the technology stack, developers can provoke, delay, or neutralize competitive responses in hypercompetitive app markets. Drawing on an empirical analysis of mobile apps launched by leading Chinese internet firms, we uncover a striking asymmetry in how rival developers react to these moves. Specifically, higher-level app modules—those tailored to particular application domains—tend to delay rival responses, while lower-level technical modules—broadly applicable across domains—trigger faster competitive reactions. This effect is heightened among firms with a history of direct competition, as they respond more aggressively to changes at the lower levels of modularization. By foregrounding the strategic role of technology stack levels in shaping competitive interactions, our study advances a differential resourcing perspective and offers new insights into the dynamics of competition within digital platforms. These findings challenge the dominant host platform-centric and cooperative views on boundary resources, illuminating how app developers actively reshape platform ecosystems to gain temporary advantage.