Knowledge that Transforms

Machine learning (ML) tools reduce the costs of performing repetitive, time-consuming tasks yet run the risk of introducing systematic unfairness into organizational processes. Automated approaches to achieving fairness often fail in complex situations, leading some researchers to suggest that human augmentation of ML tools is necessary. However, our current understanding of human–ML augmentation remains limited. In this paper, we argue that the Information Systems (IS) discipline needs a more sophisticated view of and research into human–ML augmentation. We introduce a typology of augmentation for fairness consisting of four quadrants: reactive oversight, proactive oversight, informed reliance, and supervised reliance. We identify significant intersections with previous IS research and distinct managerial approaches to fairness for each quadrant. Several potential research questions emerge from fundamental differences between ML tools trained on data and traditional IS built with code. IS researchers may discover that the differences of ML tools undermine some of the fundamental assumptions upon which classic IS theories and concepts rest. ML may require massive rethinking of significant portions of the corpus of IS research in light of these differences, representing an exciting frontier for research into human–ML augmentation in the years ahead that IS researchers should embrace. 1

Organizational decision-makers need to evaluate AI tools in light of increasing claims that such tools outperform human experts. Yet, measuring the quality of knowledge work is challenging, raising the question of how to evaluate AI performance in such contexts. We investigate this question through a field study of a major U.S. hospital, observing how managers evaluated five different machine-learning (ML) based AI tools. Each tool reported high performance according to standard AI accuracy measures, which were based on ground truth labels provided by qualified experts. Trying these tools out in practice, however, revealed that none of them met expectations. Searching for explanations, managers began confronting the high uncertainty of experts’ know-what knowledge captured in ground truth labels used to train and validate ML models. In practice, experts address this uncertainty by drawing on rich know-how practices, which were not incorporated into these ML-based tools. Discovering the disconnect between AI’s know-what and experts’ know-how enabled managers to better understand the risks and benefits of each tool. This study shows dangers of treating ground truth labels used in ML models objectively when the underlying knowledge is uncertain. We outline implications of our study for developing, training, and evaluating AI for knowledge work.

The remarkable connectivity and embeddedness of digital technologies enable innovations undertaken by a broad set of actors, often beyond organizational and industry boundaries, whose relationships mimic those of interdependent species in a natural ecosystem. These digital innovation ecosystems, if successful, can spawn countless innovations of substantial social and economic value, but are complex and prone to often surprising failure. Aiming to understand ecosystems as a new organizational form for digital innovations, I develop a theory that addresses an underexplored but important question: In a digital innovation ecosystem, how are the efforts of autonomous parties integrated into a coherent whole and what role do digital technologies play in this integration? By synthesizing ecological and information perspectives, this information ecology theory identifies several key functions that digital technologies serve in providing the information needed to support the interactions and tasks for innovation in ecosystems of varying scales. This theory contributes to digital innovation research new insights on managing part–whole relations, the role of digital technologies in innovation, and multilevel interactions in and across digital innovation ecosystems. The theory can also inspire the development of next-generation information systems for ecosystems as a new organizational form.

With the rapidly evolving permeation of digital technologies into everyday human life, we are witnessing an era of personal data digitalization. Personal data digitalization refers to the sociotechnical encounters associated with the digitization of personal data for use in digital technologies. Personal data digitalization is being applied to central attributes of human life—health, cognition, and emotion—with the purported aim of helping individuals live longer, healthier lives endowed with the requisite cognition and emotion for responding to life situations and other people in a manner that enables human flourishing. A concern taking hold in manifold fields ranging from IT, bioethics, and law, to philosophy and religion is that as personal data digitalization permeates ever more areas of human existence, humans risk becoming artifacts of technology production. This concern brings to center stage the very notion of what it means to be human, a notion encapsulated in the term human dignity, which broadly refers to the recognition that human beings possess intrinsic value and, as such, are endowed with certain rights and should be treated with respect. In this paper, we identify, describe, and transform what we know about personal data digitalization into a higher order theoretical structure around the concept of human dignity. The result of our analysis is the CARE (claims, affronts, response, equilibrium) theory of dignity amid personal data digitalization, a theory that explains the relationship of personal data digitalization to human dignity. Building upon the CARE theory as a foundation, researchers in a variety of IS research streams could develop mid-range theories for empirical testing or could use the CARE theory as an overarching lens for interpreting emerging IS phenomena. Practitioners and government agencies can also use the CARE theory to understand the opportunities and risks of personal data digitalization and to develop policies and systems that respect the dignity of employees and citizens.

Information systems (IS) use, the dominant theoretical paradigm for explaining how users apply IS artifacts toward goal attainment, gives primacy to human agency in the user–IS artifact relationship. Models and theorizing in the IS use research stream tend to treat the IS artifact as a passive tool; lacking in the ability to initiate action and accept rights and responsibilities for achieving optimal outcomes under uncertainty. We argue that a new generation of “agentic” IS artifacts requires revisiting the human agency primacy assumption. Agentic IS artifacts are no longer passive tools waiting to be used, are no longer always subordinate to the human agent, and can now assume responsibility for tasks with ambiguous requirements and for seeking optimal outcomes under uncertainty. To move our theorizing forward, we introduce delegation, based on agent interaction theories, as a foundational and powerful lens through which to understand and explain the human– agentic IS artifact relationship. While delegation has always been central to human–IS artifact interactions, it has yet to be explicitly recognized in IS use theorizing. We explicitly theorize IS delegation by developing an IS delegation theoretical framework. This framework provides a scaffolding which can guide future IS delegation theorizing and focuses on the human–agentic IS artifact dyad as the elemental unit of analysis. The framework specifically reveals the importance of agent attributes relevant to delegation (endowments, preferences, and roles) as well as foundational mechanisms of delegation (appraisal, distribution, and coordination). Guidelines are proposed to demonstrate how this theoretical framework can be applied toward generation of testable models. We conclude by outlining a roadmap for mobilizing future research.

Despite the importance of information technology (IT) innovation in today’s digitalized world, little research attention has been paid to examining how firms can incentivize IT innovation. To fill this gap, the current study investigates the impact of managerial incentives provided to chief executive officers (CEOs) on IT innovation, measured by the number of IT patents. In particular, we examine the role of risk-taking incentives provided to CEOs, captured by the sensitivity of CEO wealth to stock return volatility (i.e., Vega). Vega can motivate CEOs to engage in risky IT innovation projects by aligning their wealth with firm-specific risk. In so doing, we focus on how CEOs’ IT-related human capital (i.e., IT education and IT experience) moderates the relationship between Vega and IT innovation. Our empirical analyses reveal that a higher Vega encourages CEOs to support more IT innovation; more importantly, the impact of Vega on the amount of IT patents is stronger for firms with CEOs who have higher levels of IT education and IT experience. Our study contributes to research and practice by conceptualizing a CEO’s IT-related human capital and validating its moderating role in the relationship between risk-taking incentives provided to the CEO and the amount of IT innovation.

Online labor platforms (OLPs) can use algorithms along two dimensions: matching and control. While previous research has paid considerable attention to how OLPs optimize matching and accommodate market needs, OLPs can also employ algorithms to monitor and tightly control platform work. In this paper, we examine the nature of platform work on OLPs, and the role of algorithmic management in organizing how such work is conducted. Using a qualitative study of Uber drivers’ perceptions, supplemented by interviews with Uber executives and engineers, we present a grounded theory that captures the algorithmic management of work on OLPs. In the context of both algorithmic matching and algorithmic control, platform workers experience tensions relating to work execution, compensation, and belonging. We show that these tensions trigger market-like and organization-like response behaviors by platform workers. Our research contributes to the emerging literature on OLPs.

We analyze how advice from an AI affects complementarities between humans and AI, in particular what humans know that an AI does not know: “unique human knowledge.” In a multi-method study consisting of an analytical model, experimental studies, and a simulation study, our main finding is that human choices converge toward similar responses improving individual accuracy. However, as overall individual accuracy of the group of humans improves, the individual unique human knowledge decreases. Based on this finding, we claim that humans interacting with AI behave like “Borgs,” that is, cyborg creatures with strong individual performance but no human individuality. We argue that the loss of unique human knowledge may lead to several undesirable outcomes in a host of human–AI decision environments. We demonstrate this harmful impact on the “wisdom of crowds.” Simulation results based on our experimental data suggest that groups of humans interacting with AI are far less effective as compared to human groups without AI assistance. We suggest mitigation techniques to create environments that can provide the best of both worlds (e.g., by personalizing AI advice). We show that such interventions perform well individually as well as in wisdom of crowds settings.

Occupations are increasingly embedded with and affected by digital technologies. These technologies both enable and threaten occupational identity and create two important tensions: they make the persistence of an occupation possible while also potentially rendering it obsolete, and they magnify both the similarity and distinctiveness of occupations with regard to other occupations. Based on the critical case study of an online community dedicated to data science, we investigate longitudinally how data scientists address the two tensions of occupational identity associated with digital technologies and reach transient syntheses in terms of “optimal distinctiveness” and “persistent extinction.” We propose that identity work associated with digital technologies follows a composite life-cycle and dialectical process. We explain that people constantly need to adjust and redefine their occupational identity, i.e., how they define who they are and what they do. We contribute to scholarship on digital technologies and identity work by illuminating how people deal in an ongoing manner with digital technologies that simultaneously enable and threaten their occupational identity.

The Internet of Things (IoT) is increasingly transforming the way we work, live, and travel. IoT devices collect, store, analyze, and act upon a continuous stream of data as a by-product of everyday use. However, IoT devices need unrestricted data access to fully function. As such, they invade users’ virtual and physical space and raise far-reaching privacy challenges that are unlike those examined in other contexts. As advanced IoT devices, connected cars offer a unique setting to review and extend established theory and evidence on privacy and data sharing. Employing a sequential mixed methods design, we conducted an interview study (n=120), a survey study (n=333), and a field experiment (n=324) among car drivers to develop and validate a contextualized model of individuals’ data sharing decisions. Our findings from the three studies highlight the interplay between virtual and physical risks in shaping drivers’ privacy concerns and data sharing decisions—with information privacy and data security emerging as discrete yet closely interrelated concepts. Our findings also highlight the importance of psychological ownership, conceptualized as drivers’ feelings of possession toward their driving data, as an important addition to established privacy calculus models of data sharing. This novel perspective explains why individuals are reluctant to share even low-sensitivity data that do not raise privacy concerns. The psychological ownership perspective has implications for designing incentives for data-enabled services in ways that augment drivers’ self-efficacy and psychological ownership and thereby encourage them to share driving data. These insights help reconcile a fundamental tension among IoT users—how to avail the benefits of data-enabled IoT devices while reducing the psychological costs associated with the sharing of personal data.