International Journal of Information Management最新文献

Machine learning algorithms enable advanced decision making in contemporary intelligent systems. Research indicates that there is a tradeoff between their model performance and explainability. Machine learning models with higher performance are often based on more complex algorithms and therefore lack explainability and vice versa. However, there is little to no empirical evidence of this tradeoff from an end user perspective. We aim to provide empirical evidence by conducting two user experiments. Using two distinct datasets, we first measure the tradeoff for five common classes of machine learning algorithms. Second, we address the problem of end user perceptions of explainable artificial intelligence augmentations aimed at increasing the understanding of the decision logic of high-performing complex models. Our results diverge from the widespread assumption of a tradeoff curve and indicate that the tradeoff between model performance and explainability is much less gradual in the end user’s perception. This is a stark contrast to assumed inherent model interpretability. Further, we found the tradeoff to be situational for example due to data complexity. Results of our second experiment show that while explainable artificial intelligence augmentations can be used to increase explainability, the type of explanation plays an essential role in end user perception.

The introduction of machine learning (ML), as the engine of many artificial intelligence (AI)-enabled systems in organizations, comes with the claim that ML models provide automated decisions or help domain experts improve their decision-making. Such a claim gives rise to the need to keep domain experts in the loop. Hence, data scientists, as those who develop ML models and infuse them with human intelligence during ML development, interact with various ML stakeholders and reflect their views within ML models. This interaction comes with (often conflicting) demands from various ML stakeholders and potential tensions. Building on the theories of effective use and wise reasoning, this mixed method study proposes a model to better understand how data scientists can use wisdom for managing these tensions when they develop ML models. In Study 1, through interviewing 41 analytics and ML experts, we investigate the dimensions of wise reasoning in the context of ML development. In Study 2, we test the overall model using a sample of 249 data scientists. Our results confirm that to develop effective ML models, data scientists need to not only use ML systems effectively, but also practice wise reasoning in their interactions with domain experts. We discuss the implications of these findings for research and practice.

Significant research explores how developers leverage crowdfunding to attract finance for releasing digital goods. However, researchers seldom study “post-release activities” that are crucial for maintaining and advancing those goods. This article elaborates on the challenging nature of post-release activities for crowdfunding initiatives, asking how developers communicate their post-release plans to effectively prepare backers for possible changes. Using a grounded approach that connects the longitudinal history of fundraising to development to post-release, I examine initiatives that achieved impressive fundraising and development results yet varied significantly in their post-release outcomes. While they consistently signaled post-release plans, the differences are the signals’ costs, backers' reactions, and the post-release activities and outcomes. I present theoretical propositions that (1) developers benefit in the long run by combining high-cost signaling with engaging backers in follow-up conversations about post-release issues and (2) prospective backers can utilize developers' communication to identify their post-release signals. Unlike dominant research findings about signals’ impacts on mobilizing resources during fundraising, the findings emphasize signals’ post-release consequences for stakeholders. While different signaling approaches can enhance short-term performance, they also seed contrasting longer-term outcomes for developers, backers, and the industry. These findings advance knowledge on effective strategies for engaging society to build sustainable digital goods.

Affordance theory provides one of the most prominent lenses through which the socio-technical aspects of a system’s use can be investigated and understood. In this context, the literature has proposed that perceived and actualized affordances may be adjusted over time. Yet, how the adjustment of affordances occurs has not been explained in detail. Thus, in this article, we develop a conceptual model of feedback mechanisms that includes a more explicit description of how affordances are perceived by users, whether actualized and adjusted. With the model, we introduce the central concept of a generative base, out of which affordance perceptions emerge and which can be updated through affordance actualizations. With this base, we integrate mental model (MM) theory to explain better the differing perceptions before and after actor–artifact interactions. Our conceptual model is illustrated and specified using an empirical case of the mental mobile health (mHealth) app “Between The Lines” in Germany. In this study, we conducted 40 interviews in two rounds with 20 teenage app users. Our results show that the users’ perceptions of mHealth affordances become adjusted and hence change over time due to experiencing the actualization process, which may lead to an update of the generative base, including the user’s MM.

Artificial Intelligence (AI) has the potential to significantly impact the educational sector. One application of AI that has increasingly been applied is algorithmic grading. It is within this context that our study takes a focus on trust. While the concept of trust continues to grow in importance among AI researchers and practitioners, an investigation of trust/mistrust in algorithmic grading across multiple levels of analysis has so far been under-researched. In this paper, we argue the need for a model that encompasses the multi-layered nature of trust/mistrust in AI. Drawing on an embedded agency perspective, a model is devised that examines top-down and bottom-up forces that can influence trust/mistrust in algorithmic grading. We illustrate how the model can be applied by drawing on the case of the International Baccalaureate (IB) program in 2020, whereby an algorithm was used to determine student grades. This paper contributes to the AI-trust literature by providing a fresh theoretical lens based on institutional theory to investigate the dynamic and multi-faceted nature of trust/mistrust in algorithmic grading—an area that has seldom been explored, both theoretically and empirically. The study raises important implications for algorithmic design and awareness. Algorithms need to be designed in a transparent, fair, and ultimately a trustworthy manner. While an algorithm typically operates like a black box, whereby the underlying mechanisms are not apparent to those impacted by it, the purpose and an understanding of how the algorithm works should be communicated upfront and in a timely manner.

Requesting personal information in frontline service encounters raises privacy concerns among customers. The proximity contact tracing that occurred during the COVID-19 pandemic provides an intriguing context of information requests. Hospitality venues required contact tracing details from customers, and customer cooperation varied with concerns about privacy. Drawing on gossip theory, we investigate the roles of businesses’ data privacy practices and government support in driving customers’ responses to contact tracing. Our findings show that perceived transparency of a business’s privacy practices has a positive effect on customers’ commitment to the business, while perceived control exerts a negative effect on commitment. These effects are mediated by customers’ information falsification rather than disclosure, because the former is a sensitive behavioral indicator of privacy concerns. The results also reveal the moderating roles of government support. This research contributes to the customer data privacy literature by demonstrating the distinct effects of perceived transparency and control on commitment and revealing the underlying mechanism. Moreover, the research extends the conceptual understanding of privacy practices from online contexts to face-to-face contexts of frontline service. The findings offer implications for the management of customer data privacy.

Understanding how the application of big data analytics (BDA) generates business value is a persistent challenge in information systems (IS) research. Improving understanding of how BDA realizes business value requires unpacking theories to study the phenomenon. This study unpacks the task-technology fit (TTF) theory toward generating new and improved insights into the business value of BDA. Extant studies on TTF have mainly focused on traditional IT which is different from digital technologies like BDA that are malleable and dynamic. While TTF has primarily focused on how the technology meets task requirements, this study contends that tasks can also be structured to fit the functionality of technology. This study proposes a 2 × 2 matrix framework to explain how BDA and tasks interact. The framework indicates how the reconfigurability of tasks and the editability of BDA impact the fit between tasks and BDA. Future research should explore how the fit between tasks and BDA changes over time.

Microtask gig workers (MGWs) rely on digital platforms to arrange work agreements with requesters to complete well-defined microtasks. Many MGWs use an electronic network of practice (ENP) to facilitate information sharing about desirable and undesirable microtasks. This study uses social capital theory to theorize how social capital’s dimensions – structural, cognitive, and relational – shape the development of uncertainty-reducing and individualized-skill benefits. Based on survey data from 436 Amazon Mechanical Turk (MTurk) workers, the findings demonstrate that unique social capital dimensions affect specific ENP benefits. Understanding the communication style of an ENP (i.e., cognitive social capital) positively influences the uncertainty-reducing benefits of microtask information quality (MIQ) related to MTurk work. Combined with expectations of reciprocity and trust in ENP members (i.e., relational social capital), MIQ shapes microtask opportunity recognition (MOR), whereby individual MGWs identify opportunities to complete financially beneficial microtasks. The present study demonstrates that contextual factors, based on the coopetive nature of microtask ENPs, affect the interrelated structure of social capital theory and its underlying dimensions. Lastly, post hoc findings demonstrate the influence of MOR on MGWs’ financial performance, challenging previously held assumptions about the role of MIQ within the microtask literature.

Digital entrepreneurship is one of the most important and impactful forms of entrepreneurship. However, the majority of digital start-ups fail to scale, despite their potential to achieve growth. Moreover, what constitutes growth is significantly different for start-ups as compared with established firms: they have limited financial performance, and their growth cannot be accurately measured by metrics such as customers, revenues, profits, and turnover at the early stages of the new venture creation process or before the start-up has reached sustainability. Therefore, the first objective of this research is to contribute to the definition of growth by examining less tangible dimensions of growth. Moreover, the second objective is to identify the characteristics of early-stage digital start-ups experiencing growth. Based on a mixed-methods approach, which combines an adapted Delphi study, a questionnaire-based survey, and a comparative case study, our results suggest that growing digital start-ups exhibit a set of characteristics that include: an agile culture combining clan with adhocracy; the ability to nurture their absorptive, innovative, and adaptive capabilities effectively; and a human capital with adequate entrepreneurial skills, emotional attachment to and fitness with the start-up.

There is an exponential growth of the use of AI applications in organisations. Due to the machine learning capability of artificial intelligence (AI) applications, it is critical that such systems are used continuously in order to generate rich use data that allow them to learn, evolve and mature into a better fit for their user and organisational context. This research focuses on the actual use of conversational AI, in particular AI chatbot, as one type of workplace AI application to answer the research question: how do employees experience the use of an AI chatbot in their day-to-day work? Through a qualitative case study of a large international organisation and by performing an inductive analysis, the research uncovers the different ways in which users appropriate the AI chatbot and identifies two key dimensions that determine their type of use: the dominant mode of interaction and the understanding of the AI chatbot technology. Based on these dimensions, a taxonomy of users is presented, which classifies users of AI chatbots into four types: early quitters, pragmatics, progressives, and persistents. The findings contribute to the understanding of how conversational AI, particularly AI chatbots, is used in organisations and pave the way for further research in this regard. The implications for practice are also discussed.