In their article, The Invisible Web at Work: Artificial Intelligence and Electronic Surveillance in the Workplace, Professors Bales and Stone argue that illegitimate employer uses of artificial intelligence (“AI”) in the workplace may largely outweigh legitimate uses, creating a potentially problematic, but not necessarily unlawful, encroachment on human workers’ rights. The article is divided into three main sections. First, it comprehensively describes the numerous ways in which employers are utilizing AI to transform traditional managerial prerogatives. Second, it analyzes possible workers’ rights violations, concluding that existing law is unlikely fully to protect those rights. Third, it presents areas for future reform. The article concludes with an ominous observation: “companies are collecting unfathomable quantities of data on workers that will significantly tilt the balance of workplace power in favor of employers at workers’ expense.” (P. 62.)
Section I comprehensively surveys employers’ ubiquitous use of AI to transform traditional managerial prerogatives. The authors note that employers “utilize a dizzying array of electronic mechanisms—including trackers, listening devices, surveillance cameras, metabolism monitors, and wearable technology—to watch their workers, measure their performance, avoid disruption, and identify shirking, theft, or waste.” (P. 4.) While these mechanisms may serve legitimate employer goals, they often allow managers to “observe each worker’s every movement, both inside and outside the workplace, and during and after working hours.” (Id.) Moreover, AI algorithms can transform collected data “into a permanent electronic resume that companies are using to track and assess current workers,” and which “could potentially be shared among companies as workers move around the boundaryless workplace from job to job.” (Id.) This “invisible electronic web threatens to invade worker privacy, deter unionization, enable subtle forms of employer blackballing, exacerbate employment discrimination, render unions ineffective, and obliterate the protections of the labor laws.” (Id.)
Section I is itself divided into three parts: Part A (Pp. 5–9) briefly traces AI’s development in production processes. Innovations such as computerized robotic arms, machine learning, computer vision, AI-amplification of human capability, and voice recognition—initially introduced to the workplace to enhance productivity—are now being turned into surveillance instruments. Part B (Pp. 9–15) introduces the concept of People Analytics or data-driven human resources. The authors explain that the People Analytics field uses AI “to guide HR decisions for many areas, including making hiring decisions, monitoring performance, predicting an individual’s work trajectory, evaluating workers to set compensation, and determining an employee’s likelihood of terminating the employment relationship.” (P. 9.) Part C (Pp. 15–22) traces creative uses of electronic surveillance devices to monitor workers. For example, firms are now using electronic badges not only to access buildings but also to record employee conversations, track employee movements, and monitor employees’ vital signs. (Pp. 17–20.) The Section ends with a discussion of how the development of these technologies can be used to monitor workers’ off-duty activities, which “not only creates the potential for highly intrusive monitoring, but also raises questions about how employers will use the data they collect about employees’ performance, with whom they will share it, and how long they will keep it. AI-enhanced data collection, retention, and analytic capabilities threaten to create a permanent record of employee productivity, activity, and medical and physiological attributes.” (Pp. 20–22.)
Part II, which focuses on potential employer liability, picks up where Part I.C. leaves off. After all, that employers can retrain AI-enhanced data to create worker profiles raises two questions about which every worker should be concerned: How can employers use these data collections to harm workers and does the existing legal framework provide sufficient protection to workers? Like so many questions that have arisen during the technological revolution, the authors show that the law is often woefully inadequate to protect workers’ rights. (Pp. 22–59.)
Part II is divided into four parts, each taking a deep dive into a different area of law. Take employment discrimination. AI can amplify bias, by replicating past employment decisions that generated profitable results for the company but themselves were laced with bias. For example, “a hiring algorithm based on current workplace demographics” in Silicon Valley, which “has long been criticized for its white-male-dominated workplaces … likely will replicate and entrench [those] past hiring practices.” (Pp. 22–23.) While the authors recognize that AI could also reduce bias, it can also augment AI, which may reduce the “salutary effect of AI.” (P. 29.) Part II.B. showcases the legal limitations of workplace privacy laws, providing as one example the use of pre-hire videos to evaluate job candidates. Those interviews can “collect data from an applicant’s own devices or from cookies or other technological tracking devices.” (p. 34.) Part II.C. shows how antitrust laws might be used to sue employers who “use shared employee information amassed through AI and electronic surveillance to set compensation, engage in a no-raiding agreement, or blacklist an employee.” (P. 36.) However, this area of law is currently untested. Finally, there are labor law implications, primarily concerning surveillance, bargaining over privacy and surveillances issues, and representation, all of which are discussed in Parts II.D. and II.E. (Pp. 48–59.) But there are limitations here as well. Although the NLRA has historically protected concerted activity from surveillance, for example, the Trump Board has vastly cut back on those protections in recent cases. (Pp. 53–55.)
Part III establishes a compact agenda for future research and reform. (pp. 59–62.) This section cogently explains that while “gathering and using such data have enormous implications for the application of existing workplace laws,” such activities “are occurring with no legal or regulatory oversight.” The authors opine that “[p]erhaps existing laws will be sufficiently adaptable to respond to these new conditions, but there is significant risk they will not.” Accordingly, the authors identify the need to clarify “the law of disparate impact … to ensure that plaintiffs need to show, in their prima facie case, only that an algorithm as a whole caused a disparate impact; plaintiffs should not be expected to show precisely how the algorithm produced the bias.” (P. 60.) The authors also suggest that Congress amend electronic surveillance laws along the lines of “the European General Data Protection Regulation … as a starting point but augmenting it to specifically address data collection in the employment context.” (P. 61.) Such amendments would give workers greater control over their own personal data collected by their employers. The authors further suggest that the Federal Trade Commission clarify that data collected and shared by several employers about workers are in fact antitrust violations. Finally, the authors suggest the Board affirm the following: electronic surveillance violates Section 7; employers must bargain over such “the existence and scope of electronic monitoring and the use of algorithms in decisions involving discipline, job assignment, promotion, or pay;” and that “the existing duty on employers to provide unions with information necessary for meaningful bargaining and grievance-resolution should be extended to information about an employer’s practices and plans regarding the use of AI in personnel management decisions, and to information about algorithms or data collected by AI that an employer has used in personnel decisions affecting individual grievants.” (Pp. 61–62.)
Were this article limited to Section I, alone, Professors Bales and Stone would have made a significant contribution to work law literature. While descriptive, those accounts are the necessary initial step for understanding the social problem presented—employer data collection jeopardizes workers’ rights. But the article is impactful because it goes beyond the descriptive. For example, Section II explains how technological advances in AI are connected to modern HR decisions. Drawing those connections must have been painstaking and tedious for the authors, yet they read as the most interesting part of the article. Moreover, that Section makes an original contribution because the problems the authors expose have not yet arisen in the caselaw.
In closing, the article’s in-depth analysis of possible workplace violations and conclusion that existing law is likely insufficient to protect workers’ rights expose an uncomfortable truth. Namely, technological progress is currently progressing at such a rapid pace that existing law is unlikely to catch up in time to protect workers. As Professors Bales and Stone explain: “given the blinding pace at which companies currently are collecting data on workers, a legal response may quickly become a moot point. Once sufficient data are collected, it likely will be difficult to put the genie back in the bottle.” (P. 59.) Welcome to the brave new workplace.