Knowledge that Transforms

Social media’s dominant economic model relies on curating content that appeals to users’ emotions to capture and hold their attention, resulting in experiences that are primarily affective. User behavior drives content propagation, which is the core process underlying major social media phenomena—both beneficial and harmful. While IS research recognizes the importance of affective factors in content propagation, this research is marked by diverse theoretical foundations, a focus on content-level factors, and a complex landscape of findings. To unpack the affective factors underlying user behavior in content propagation and integrate insights from prior research, we develop an affective model of social media content propagation grounded in the affective response model (ARM) and extend ARM’s propositions to the omnibus context of social media. We examine two highly relevant, theoretically nuanced, and countervailing affective forces—humor and politics—by developing testable hypotheses in the discrete context of humorous and political memes. Specifically, we consider the role of human factors (mood, ideology), affective dispositions (sense of humor, affinity for political humor), content factors (political nature), and induced affective states (mirth) to predict propagation intentions. We test the model using an online experiment involving 289 participants, balanced across gender and political orientation, and 48 memes, balanced across presence and partisan leaning of political content. Using cross-classified mixed-effects regression, we find general support for our hypotheses and uncover interesting differences in politically relevant factors. The broader proposed model, supported by the study’s findings, offers a general multilevel framework to illuminate the role of affective processes in social media’s promise and peril, situate prior IS research, and support future studies of social media user behavior. We contribute to the broader social media literature by identifying emotionality as a contextual stimulus, and to ARM by explicating its multilevel nature, unpacking indirect effects, and contextualizing it to social media content propagation.

In this methods article, we unpack how researchers can foster trustworthiness in dimensionalization and category surfacing (DCS), a key method family within the genre of computational theory construction (CTC). Information systems (IS), management, and organizational scholars are increasingly leveraging DCS tools such as topic modeling, word embeddings, and clustering to surface latent categories and dimensions from textual data for theory construction. Yet they struggle because evaluations of such research often default to transparency, operationalized as replicability and accountability, which obscures the analytical choices that actually make DCS research rigorous. In this study, we recast transparency as a means toward trustworthiness. We treat researchers’ analytical moves as the primary unit of methodological reasoning in how they design, conduct, and disclose their choices across research phases. We develop a framework that authors, reviewers, and editors can use to construct and evaluate DCS research. The framework specifies how trustworthiness arises from the interplay of two research design choices: primacy to theoretical versus practice lexicons, and whether the content of texts or the structure of the corpus carries the theoretical load. We articulate expectations for conduct and disclosure across these design choices, clarifying how proportionate reasoning anchors trustworthiness. We conclude with implications for advancing trustworthiness within the broader CTC community and across other computational approaches to research.

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.

Despite its many benefits, widespread access to individuals’ personal data also causes severe privacy concerns for consumers, companies, and policymakers. This study proposes a novel framework that adapts the Shapley value-based feature attribution approach to the problem domain of data privacy by capturing the two crucial dimensions of data privacy—disclosure risk and data utility. Our proposed framework takes a holistic view of data masking through a fair feature attribution approach based on Shapley values. Different from the existing literature that mostly focuses on the risk-utility trade-off at the dataset level, the proposed framework addresses the trade-off at the feature level. Furthermore, the proposed framework is agnostic to data masking methods, statistical and machine learning methods, and data utility and disclosure risk evaluation metrics. Experimental results show that our proposed method can effectively reduce disclosure risk while preserving data utility.

At MIS Quarterly, our commitment to supporting a wide range of scholarly contributions, across theories, methods, topics, and geographies, is unwavering. Not only is it the right thing for the field; it is fundamentally important for the advancement of science. We believe that the quality and relevance of IS research improve when we broaden the conditions under which scholars from diverse backgrounds, regions, and traditions can meaningfully contribute. As a global journal, we actively seek to foster inclusive scholarship: work that reflects the diversity of our community and the complexity of the digital world it seeks to understand.

Many real-world process automation environments are rife with high-expertise and limited labeled data. We propose a computational design science artifact to automate systematic review (SR) in such an environment. SR is a manual process that collects and synthesizes data from medical literature to inform medical decisions and improve clinical practice. Existing machine learning solutions for SR automation suffer from a lack of labeled data and a misrepresentation of the high-expertise manual process. Motivated by humans’ impressive capability to learn from limited examples, we propose a principled and generalizable few-shot learning framework—FastSR—to automate the multistep, expertise-intensive SR process using minimal training data. Informed by SR experts’ annotation logic, FastSR extends the traditional few-shot learning framework by including (1) various representations to account for diverse SR knowledge, (2) attention mechanisms to reflect semantic correspondence of medical text fragments, and (3) shared representations to jointly learn interrelated tasks (i.e., sentence classification and sequence tagging). We instantiated and evaluated FastSR on three test beds: full-text articles from Wilson disease (WD) and COVID-19, as well as a public dataset (EBM-NLP) containing clinical trial abstracts on a wide range of diseases. Our experiments demonstrate that FastSR significantly outperforms several benchmarking solutions and expedites the SR project by up to 65%. We critically examine the SR outcomes and practical advantages of FastSR compared to other ML and manual SR solutions and propose a new FastSR-augmented protocol. Overall, our multifaceted evaluation quantitatively and qualitatively underscores the efficacy and applicability of FastSR in expediting SR. Our results have important implications for designing computational artifacts for automating/augmenting processes in high-expertise, low-label environments.

When chatbots are deployed to automate customer service, it is nearly inevitable that situations will arise in which they struggle to understand customer requests. Unfortunately, the onus of resolving such conversational breakdowns tends to fall on either the customer or the chatbot alone, turning customer-chatbot interaction into a frustrating and often unsuccessful guessing game. Despite indications that customers would be open to collaboration, we know little about repair strategies that involve the customer and chatbot working together to resolve breakdowns. Our research addresses this gap by investigating the design and impact of collaborative repair strategies in customer-chatbot interaction. Drawing upon an integration of the theory of least collaborative effort with research on human-machine communication and customer service chatbots, we propose a novel repair strategy design; we instantiated it in the chatbot of a large insurance company and conducted a naturalistic summative evaluation through a randomized field experiment. Overall, our results suggest that a collaborative repair strategy can lead to more breakdowns being resolved and mitigate the negative impacts of breakdowns on key customer outcomes. Our research offers a new way of thinking about customer-AI service interactions by shifting the narrative from confrontation to collaboration, extends the theory of least collaborative effort by integrating the perspective of customer-chatbot interaction, and provides in-depth insights into breakdown and repair in real-world conversations between customers and chatbots.

Content validation is an essential aspect of the scale development process that ensures that measurement instruments capture their intended constructs. However, researchers rarely undertake this core step in behavioral research because it requires costly data collection and specialized expertise. We present RATER (replicable approach to expert ratings), a free web-based system (www.contval.org) that can help the broader research community (scientists, reviewers, students) gain quick and reliable insights into the content validity of measurement instruments. Guided by psychometric measurement theory, RATER evaluates whether a scale’s items correspond to their intended construct, remain distinct from other constructs, and adequately represent all aspects of the construct’s content domain. The system employs two unique artificial intelligence models, RATERC and RATERD, which leverage psychometric scales from 2,443 journal articles spanning eight disciplines and two state-of-the-art large language model architectures (i.e., BERT and GPT). A set of six complementary studies confirms the RATER system’s accuracy, reliability, and usefulness. We find that RATER can augment the scale development and validation process, increasing the validity of findings in behavioral research.

The research on online social movements generally concludes that collective identity, i.e., the sense of we-ness that individual protesters in a movement share, is not only unattainable but also dispensable, even though it is considered a defining feature of traditional movements. In this paper, we explore one of the boundary conditions of these findings, namely the riskiness of protest practices. Analysing the high-risk social movement, My Stealthy Freedom (MySF), which contests compulsory hijab in Iran in a way that hybridizes online and offline protest practices, we show that a sense of collectiveness can be instantiated in online social movements, why it is critical to the success of high-risk activism, and how it is (re)produced. Comparing and contrasting three instantiations of MySF, each of which was enacted on a different social media platform, we develop a theoretical model of how feelings of collectiveness are enacted in high-risk online activism. In addition to providing guidance for online movements where collective identity is desirable, our study challenges prior research on online activism by theorizing the role of embodiment, affect, and the dialectic between activists’ personal and the movement’s collective identity.

Optimizing multiple objectives simultaneously is an important task for recommendation platforms to improve their performance. However, this task is particularly challenging since the relationships between different objectives are heterogeneous across different consumers and dynamically fluctuate according to different contexts, resulting in a Pareto-frontier in the result of recommendations, where the improvement of any objective comes at the cost of others. Existing multi-objective recommender systems do not systematically consider such dynamic relationships; instead, they balance between these objectives in a static and uniform manner, resulting in only suboptimal recommendation performance. In this paper, we propose a Deep Pareto Reinforcement Learning (DeepPRL) method, where we (1) comprehensively model the complex relationships between multiple recommendation objectives; (2) effectively capture personalized and contextual consumer preferences for each objective; (3) optimize both the short-term and the long-term recommendation performance. As a result, our method achieves significant Pareto-dominance over the state-of-the-art baselines across four offline experiments. Furthermore, we conducted a controlled experiment on Alibaba's video streaming platform, where our method simultaneously improved three conflicting business objectives significantly over the latest production system, demonstrating its tangible economic impact in practice.