Knowledge that Transforms

In this paper, we confront a paradox in the IS literature that even though our field focuses on the rapid pace of technological change and the dramatic scale of technology-enabled organizational and societal changes, we sometimes find ourselves studying these changes using—largely without question—constructs that were developed in a vastly different IT, user, and organizational environment. We provide guidelines to help assess whether an existing construct warrants updating and to structure the updating task if it is undertaken. Our three-step process provides for a theoretically grounded and comprehensive method that ensures we balance the need for construct updating against the need to sustain our cumulative tradition. We illustrate our guidelines using computer self-efficacy (CSE) as a case study. We document each of the steps involved in analyzing, reconceptualizing, and testing the revised construct information technology self-efficacy (ITSE). Our analyses show that the new construct better explains both traditional and contemporary constructs with a traditional (postal survey) and contemporary (online panel) sample. We discuss the implications of our work both for research on self-efficacy and more broadly for future updating of other important constructs.

Innovation breakthroughs prompt sensemaking discourses that promote community learning and socially construct the innovation. Through this discourse, interested actors advance diverse frames, appealing to consumers with disparate preferences but raising concerns for the coherence of that discourse. We unpack this diversity-coherence paradox by recasting coherence as the relatedness of innovation frames and spotlighting the role of discursive fields that circumscribe meaning. Our empirical context is the first six years of blockchain discourse across seven discursive fields. Our research offers three insights in furtherance of an ecological perspective on innovation discourse. First, framing diversity emanates from discursive fields rather than from actors. Second, fields play differentiated roles in the framing process. Enactment fields comprised of actors with direct experience with the technology limit diversity. They do so by erecting walls that circumscribe discourse through imprinting on their original frame and retracting from or abandoning frames learned from other fields. In contrast, mediated fields, in which actors lack direct experience with the technology, enhance diversity. They do so by imitating or learning from other fields and foreshadowing or anticipating the frames used by other fields, thereby building bridges. Third, rather than opposing each other, diversity and coherence coevolve as the diversity induced by mediated fields increases framing redundancies, synthesizing frames into a coherent community understanding of the innovation. Our research signals to the actors who serve as innovation ambassadors and gatekeepers that diverse views of an innovation are not only inevitable, given the many discourse fields in which those views are formulated, but can also be coherent and desirable.

The prominence of real-time ridesharing services, such as Uber and Lyft, has dramatically changed the landscape of traditional industries. This study provides a comprehensive analysis of the short-term wage elasticity of labor supply in real-time ridesharing markets using data from a major ridesharing platform in China. By exploiting an exogenous shock from uneven driving restrictions as an instrumental variable, we find a negative labor supply elasticity for ridesharing drivers, suggesting that drivers tend to drive less during days with a higher average hourly wage. Specifically, a percent increase in hourly wage will lead to a 0.931 percent decrease in daily working hours. This surprising finding is consistent with the behavioral income-targeting model based on the theory of reference-dependent preferences: Drivers have heuristic daily targets for total earnings and are more motivated to supply labor when they are below their income target than when they are above it. Therefore, they work less on days when earnings per hour are high and quit the market once their income target is reached. In addition, we find that taxi drivers are more rational and have positive labor supply elasticity, which implies that drivers are more rational when they have repeated opportunities for learning. Estimating labor supply elasticity is critical to understanding the economic efficiency of various surge pricing algorithms and driver subsidization programs for ridesharing platforms and policymakers. Our research suggests that a uniform price surging or driver subsidization approach for all ridesharing drivers may not incentivize the labor supply of drivers effectively.

We examine counterfactual explanations for explaining the decisions made by model-based AI systems. The counterfactual approach we consider defines an explanation as a set of the system’s data inputs that causally drives the decision (i.e., changing the inputs in the set changes the decision) and is irreducible (i.e., changing any subset of the inputs does not change the decision). We (1) demonstrate how this framework may be used to provide explanations for decisions made by general data-driven AI systems that can incorporate features with arbitrary data types and multiple predictive models, and (2) propose a heuristic procedure to find the most useful explanations depending on the context. We then contrast counterfactual explanations with methods that explain model predictions by weighting features according to their importance (e.g., Shapley additive explanations [SHAP], local interpretable model-agnostic explanations [LIME]) and present two fundamental reasons why we should carefully consider whether importance-weight explanations are well suited to explain system decisions. Specifically, we show that (1) features with a large importance weight for a model prediction may not affect the corresponding decision, and (2) importance weights are insufficient to communicate whether and how features influence decisions. We demonstrate this with several concise examples and three detailed case studies that compare the counterfactual approach with SHAP to illustrate conditions under which counterfactual explanations explain data-driven decisions better than importance weights.

Digital ventures must navigate a key tension as they design new digital market offerings—that is, products or services that are embodied in digital technologies or enabled by them. On the one hand, digital ventures pursue a vision that builds on what might be possible through the generative potential that digital technology offers; on the other hand, they face an environment in the here and now, with existing customer preferences, extant regulations, and legacy technology. Taking a designing view, we trace how six independent digital ventures in the German financial services industry dealt with this tension as they created their digital market offerings. Our findings suggest that digital ventures enact three designing mechanisms to resolve the tension: bounding the technology scope, transposing through digital objects, and probing the solution space. Through these mechanisms, digital ventures construct a buffer—one that has functional, material, and temporal dimensions—between the vision they gradually realize through their market offering and the here-and-now conditions of the environment that digital ventures enter.

The explosive spread of false news on social media has severely affected many areas such as news ecosystems, politics, economics, and public trust, especially amid the COVID-19 infodemic. Machine intelligence has met with limited success in detecting and curbing false news. Human knowledge and intelligence hold great potential to complement machine-based methods. Yet they are largely underexplored in current false news detection research, especially in terms of how to efficiently utilize such information. We observe that the crowd contributes to the challenging task of assessing the veracity of news by posting responses or reporting. We propose combining these two types of scalable crowd judgments with machine intelligence to tackle the false news crisis. Specifically, we design a novel framework called CAND, which first extracts relevant human and machine judgments from data sources including news features and scalable crowd intelligence. The extracted information is then aggregated by an unsupervised Bayesian aggregation model. Evaluation based on Weibo and Twitter datasets demonstrates the effectiveness of crowd intelligence and the superior performance of the proposed framework in comparison with the benchmark methods. The results also generate many valuable insights, such as the complementary value of human and machine intelligence, the possibility of using human intelligence for early detection, and the robustness of our approach to intentional manipulation. This research significantly contributes to relevant literature on false news detection and crowd intelligence. In practice, our proposed framework serves as a feasible and effective approach for false news detection.

Despite increased empirical attention, theory on bots and how they act to disseminate information on social media remains poorly understood. Our study leverages the conduit brokerage perspective and the findings of a multiple case study to develop a novel framework of algorithmic conduit brokerage for understanding information dissemination by bots and the design choices that may influence their actions. Algorithmic conduit brokerage encompasses two intertwined processes. The first process, algorithmic social alertness, relies on bot activity to curate and reconfigure information. Algorithmic social alertness is significant because it involves action triggers that dictate the kinds of information being searched, discovered, and retrieved by bots. The second process, algorithmic social transmission, relies on bot activity to embellish and distribute the information curated. Algorithmic social transmission is important because it can broaden the reach of information disseminated by bots through increased discoverability and directed targeting. The two algorithmic conduit brokerage processes we offer are unique to bots and distinct from the original conceptualization of conduit brokerage, which is rooted in human activity. First, since bots lack the human ability of sensemaking and are instead fueled by automation and action triggers rather than by emotions, algorithmic conduit brokerage is more invariant and reliable than human conduit brokerage. Second, automation increases the speed and scale of information curation and transfer, making algorithmic conduit brokerage not only more consistent but also faster and more extensive. Third, algorithmic conduit brokerage includes a set of new concepts (e.g., action triggers and rapid scaling) that are specific to bots and therefore not applicable to human conduit brokerage.

International dark web platforms operating within multiple geopolitical regions and languages host a myriad of hacker assets such as malware, hacking tools, hacking tutorials, and malicious source code. Cybersecurity analytics organizations employ machine learning models trained on human-labeled data to automatically detect these assets and bolster their situational awareness. However, the lack of human-labeled training data is prohibitive when analyzing foreign-language dark web content. In this research note, we adopt the computational design science paradigm to develop a novel IT artifact for cross-lingual hacker asset detection (CLHAD). CLHAD automatically leverages the knowledge learned from English content to detect hacker assets in non-English dark web platforms. CLHAD encompasses a novel Adversarial deep representation learning (ADREL) method, which generates multilingual text representations using generative adversarial networks (GANs). Drawing upon the state of the art in cross-lingual knowledge transfer, ADREL is a novel approach to automatically extract transferable text representations and facilitate the analysis of multilingual content. We evaluate CLHAD on Russian, French, and Italian dark web platforms and demonstrate its practical utility in hacker asset profiling, and conduct a proof-of-concept case study. Our analysis suggests that cybersecurity managers may benefit more from focusing on Russian to identify sophisticated hacking assets. In contrast, financial hacker assets are scattered among several dominant dark web languages. Managerial insights for security managers are discussed at operational and strategic levels.

This research highlights the circulative nature of digital platform ecosystem dynamics. Investigating these dynamics, we examine the mutual influence between participants’ product scope and product innovation over time and probe the moderating role of co-created collaborative networks. We distinguish between two types of product innovation: new product development and existing product updates. Our longitudinal analysis of the Hadoop software ecosystem indicates that participants covering a broader scope of the platform’s technological layers are less likely to develop new products but more likely to update existing products. In turn, participants with more frequent new product development are more likely to expand their product scope, whereas those with more frequent existing product updates are less likely to pursue scope expansion. Participants’ centrality in the ecosystem’s collaborative network amplifies the bidirectional link between product scope and existing product updates but weakens the link between product scope and new product development. Our findings offer a theoretical and practical understanding of temporal dynamics between participants’ product scope choices and different forms of product innovations in the co-created collaborative network environment.

Companies may face serious adverse consequences as a result of a data breach event. To repair the potential damage to relationships with stakeholders after data breaches, companies adopt a variety of response strategies. However, the effects of these response strategies on the behavior of stakeholders after a data breach are unclear; differences in response times may also affect these outcomes, depending on the notification laws that apply to each company. As part of a multimethod study, we first identified the adopted response strategies in Study 1 based on content analysis of the response letters issued by publicly traded U.S. companies (n = 204) following data breaches; these strategies include any combination of the following: corrective action, apology, and compensation. We also found that breached companies may remain silent and adopt a “no action” strategy. In Studies 2 and 3, we examined the effects of various response strategies and response times on the predominant stakeholders affected by data breaches: customers and investors. In Study 2, we focused on customers and present a moderated-moderated-mediation model based on the expectancy violation theory. To test this model, we designed a factorial survey with 15 different conditions (n = 811). In Study 3, we focused on investors and conducted an event study (n = 166) to examine their reactions to company responses to data breaches. The results indicate the presence of moderating effects of certain response strategies; surprisingly, we did not find compensation to be more effective than apology. The magnitude of the moderating effects of response strategies is contingent upon response time. We also found that the negative effects of data breaches disappear after six months. We interpret the results and provide implications for research and practice.