Technologies are rapidly evolving to enhance human productivity; however, there is growing potential to deskill or obsolesce the work performed by professionals and managers. Basic laws such as the Fair Labor Standards Act (“FLSA”) and National Labor Relations Act (“NLRA”) — passed in the 1930s — are founded on traditional conceptions of individual discretion and supervision. As artificially intelligent technologies limit the need for having employees direct the work of others or use their expertise, there is growing potential for these employees to fall outside the boundaries of these laws. As a consequence, professional employees who are exempt under the FLSA could make valid claims for overtime pay; managers who are exempt under the NLRA could make valid claims for forming a union and bargaining with employers; or these employees could become gig workers, untethered from employment. More distant possibilities include humanoids, hybridized persons whose minds and bodies are improved for performance with genetic and bionic interventions. Such developments would raise fundamental public policy questions about regulating the competitive effects of artificially-enhanced labor. Technologists should consider these displacing and disruptive effects; and lawmakers should begin to anticipate significant dislocations caused by AI, bionic, and humanoid technologies.
By Michael H. LeRoy[1]
I. INTRODUCTION
Casey Stengel, Sparky Anderson, and Tommy Lasorda applied decades of baseball wisdom to achieve Hall of Fame success. Today, many major league managers rely on data analytics. Algorithms increasingly drive their situational decisions[2] — selecting a pitcher to get a particular out, using an infield shift to match the hitting tendencies of a batter, ordering a batter to take a pitch, and the like.
Baseball is just beginning to substitute artificial intelligence for human judgment to manage teams over a 162-game season. This future is foretold in Norsetter v. Minnesota Twins LLC,[3] a court case from 2021. A 33-year-old executive with an analytics background replaced a 59-year-old baseball scouting coordinator. The court dismissed Norstetter’s age discrimination lawsuit, stating that the court could not “evaluate the merits of the Twins’ decisions to change its scouting philosophy and eliminate Norsetter’s position.”[4] The case reflects a turning point in how teams manage talent.
Will employment laws keep up with AI work? Probably not. Minimum wage and overtime laws, and union organizing laws, are based on outdated assumptions and definitions. Professionals and supervisors use discretion to direct the work of subordinates.[5] They have authority to hire or fire other employees.[6] Generally speaking, they are not owed overtime pay nor a minimum hourly wage. They do not qualify for union representation.
Under certain other conditions, they may be treated as independent contractors.[7] Not only does this classification negate minimum wage requirements and access to unions — it excludes these workers from discrimination laws, worker’s compensation, unemployment insurance, and other protective labor laws.
This baseball scenario suggests the future of gig work for many managers and professional employees. A baseball manager’s expertise will continue to erode as teams search for competitive advantages driven by artificially intelligent technologies. In the near-term, a data analytics guru might advise or even direct the manager in a dugout on a pitch-by-pitch basis. But over time, even the data wonk will obsolesce as teams incorporate cutting edge, artificially intelligent computer programs.
Apart from baseball managers, millions of other managerial and professional employees will find their expertise, education, training, and discretionary judgment infused by artificially intelligent adjuncts to their work.[8] Some will find that AI enhances their work with productivity gains. However, other people will work in tandem with computer programs that usurp their intellectual contribution to their occupation. However, there will be employees whose craft or profession obsolesces to extinction.
Up to now, employment laws in the U.S. have failed to protect low skill workers who labor on app-driven platforms that pay piece rates.[9] This failure of employment laws will likely expand to AI-driven changes that impact a higher skilled segment of the workforce. Broad swaths of professional and managerial employees whose work will be subjected to an increasingly data-governed future will find their jobs are also on a slippery slope toward gigification. No one — perhaps not even an AI program — can predict the scope of this failure of employment laws. Nor does the future doom professions and skilled jobs. But just as gigification of labor has affected workers in various ride-share, courier, and home service occupations, the maturation of AI applications is on a path to put professional and managerial employees at risk for increasing gigification.
II. ARTIFICIAL INTELLIGENCE INFUSES PROFESSIONAL AND SUPERVISORY WORK
For centuries, new technology has posed threats to displace labor. William Lee, a hopeful inventor of a mechanical knitting loom who sought a patent from Queen Elizabeth I in 1589, was disappointed when the Queen denied his petition on grounds that the new device would throw hand-knitters out of work.[10] More recently, a comprehensive study of machine learning estimated that in the near-term “most workers in transportation and logistics occupations, together with the bulk of office and administrative support workers” were likely to be displaced by computerized technologies.[11]
AI technologies also affect professional work. This trend includes medical imaging by gastroenterologists, cardiologists, and ophthalmologists.[12] Beyond the work of medical professionals, the work of lawyers is increasingly performed by AI programs.[13] Journalists compete with computer programs to write news stories.[14] AI applications have led to “neurofinance,” a new discipline that explores the intersections of psychology, neuroscience, and finance.[15]
AI technologies are deployed for various reasons. They may improve a professional employee’s productivity and performance. But they have potential to deskill work by substituting machine-learned expertise for human training and experience. They are no less threatening to professional occupations than William Lee’s mechanical knitting machine for hand-knitters.
III. LEGAL EFFECTS OF MANAGING WORK WITH ALGORITHMS
Managers direct and supervise the work of subordinates, whether in baseball or other industries. This implicates how managers are paid. When a manager uses discretion and judgment in making decisions for an employer, federal labor and employment laws are implicated.
The National Labor Relations Act (“NLRA”) excludes certain supervisors from union representation.[16] A person who manages the work of others often has more experience, education, and training than a subordinate. However, by being excluded from collective bargaining, workplace managers likely miss out negotiating with other supervisors for better pay, hours, and benefits.
The Fair Labor Standards Act (“FLSA”) similarly classifies this person as “exempt,” meaning their employers are not required to record their time at work, nor pay minimum wages or overtime.[17] Instead, the law allows for payment of a salary, if it is above a threshold set by the Department of Labor. Currently, that pay is $684 a week ($35,568 a year).[18]
Thus, an employee who is classified under FLSA as an executive, administrative assistant, or professional, and who earns this minimum pay, is not legally owed overtime. To illustrate, an executive secretary who works 60 hours in a week and is paid $684 would earn $11.40 with no overtime under federal wage law. Contrast this with a professional employee, someone with an advanced degree. If their work is so routine that it involves little or no use of professional discretion or judgment— in other words, if their professional label does not match their labor, such as a lawyer who searches documents to match words— they may be able to sue for unpaid overtime.[19] This is the type of work that AI programs can do efficiently.
Enacted in the 1930s, the FLSA and NLRA were built on the idea that managers enjoy high status in their organizations and are paid commensurately. But infusion of AI in the managerial direction of employees may undercut this assumption. Take baseball managers as an example. The average annual salary for a baseball player is $4 million.[20] However, pay for baseball managers — which once kept pace with player salaries — has fallen dramatically.[21] Devaluation of the manager’s pay may reflect the substitution of analytics for his baseball acumen.
In response to eroding pay and control over their work, some managerial and professional employees are seeking union representation. School principals supervise and direct teachers. However, like baseball managers, their ability to manage their work conditions and district policies is shrinking. Now, these front-line managers in Chicago Public Schools are seeking a change in labor laws to allow them to form and join a union.[22]
In March, tech workers at The New York Times voted 404-88 to join a union.[23] More than half of this group consists of product engineers and supervisors. While the newspaper is challenging the outcome of the election, the National Labor Relations Board held the vote because the newspaper did not offer enough evidence to exclude workers with managerial or supervisory functions. In other words, job titles that implied that some employees were excludable as supervisors may not have reflected how little these employees directed the work of colleagues.
Whatever becomes of the union organizing ambitions of Chicago school principals and New York Times tech workers, their efforts show that people who oversee their workplace actually feel a need for a voice in determining their own pay, hours, and working conditions. Their experiences are more blue-collar than white-collar. And even if they don’t get their hands dirty at work, these college educated front-line managers of schools and a prestigious newspaper feel disempowered enough to want a union to speak for them. The role that technologies play in their marginalization is not easy to tease out, but these mid-level professionals may be reacting to computerized work processes and devalued job content, somewhat like baseball managers.
IV. WEARABLE TECHNOLOGIES AND TRANSHUMANISM: A FUTURE FOR HUMANOIDS?
Wearable technologies record biological functions, providing data not only for personal use but for managing work performance. Some pro athletes are outfitted with wearable technologies that enable sophisticated and instant quantification in team sports, including individual and team-tactical behavior.’[24] These highly miniaturized data-capturing devices suggest “the support or eventual substitution of qualitative performance measurements with quantitative methods.”[25]
Futurists have envisioned a transhuman being capable of superior performance.[26] This future is arriving incrementally. For example, a versatile and strong bionic hand, enabled with Bluetooth technology for gripping, is in development.[27] While no bionic human has been created, three examples from sports sketch this future. At the low-end of altered physiology, there are pitchers with Tommy John surgery to repair a torn ulnar collateral ligament inside the elbow. Some evidence shows this surgery has improved a pitcher’s performance.[28] At a higher level of altered physiology, Lia Thomas, a transgender NCAA swimmer for the University of Pennsylvania, has stirred questions over athletic competition. She won a national championship, but cisgender competitors complained that she raced on unfair terms.[29] Oscar Pistorius — the double-amputee track star from South Africa — presents a third example of altered physiology. He made history in the Summer Olympics of 2012 by competing and performing well-enough to be in the mix for a medal.[30] However, competitors complained that he enjoyed an unfair advantage.[31]
Tommy John surgery is not shrouded in controversy, but whether Thomas and Pistorius should have been allowed even to compete for championships is more complicated. These deep and multi-layered controversies could carryover to intellectual endeavors.
Suppose a graduating Ph.D. from an elite school was genetically engineered at conception to excel in STEM fields. She lists her in vitro conceptual history on her resume while applying for a prestigious faculty position. Would her engineered pedigree provide an ancillary advantage? On the other hand, societal unease with altering and selecting human beings at conception could work against her. This hypothetical recontextualizes Francis Galton’s theory of eugenics, which at the turn of the Twentieth Century was embraced by academics and professionals before its fusion with Nazi ideology discredited it.[32]
Is there a future for artificially superior humans? Suppose a biochemical injectable speeds up mathematical reasoning in the human brain. Another injectable levels the range of human emotion for more efficient reasoning. To add one more layer of intrigue, suppose that these mind-altering agents are synched to mimic neural learning in potent AI programs, so that a human and machine learn in tandem with accelerated speed. In sum, imagine a STEM researcher who is engineered as a Spock-like Vulcan symbiotically tethered to a parallel AI technology — a humanoid.
Humanoid labor would accentuate inequality, creating a new upper crust of “haves” who could outcompete professionals and managers who excel today with natural talent. Employment laws are ill-equipped to deal with inequality, notwithstanding their tempering policy goals. Some laws have dealt with work arrangements that are either morally repugnant or unfair. Laws against involuntary servitude, including slavery but also peonage, address morally repugnant labor that, in some contexts such as manual agricultural labor, were defended on efficiency grounds.[33] Child labor laws address the unfairness to adults in competing for wages with an exploited part of the labor force.
Would the creation of a Spock-like knowledge worker lead to a new age of employment law that defined and protected unaltered human labor? Some engineers are striving to define ethical constraints on the use of technologies to reflect humanist values. One possibility, modelled after laws against slavery and child labor, is a blanket prohibition against humanoid labor. However, another possibility is to use GMO food regulations as a model for GMH labor — genetically modified human labor. While the EPA and FDA regulate GMO food safety, perhaps the Department of Labor would regulate types of work that can — and cannot — be performed by humanoids, with the aim of protecting “unmodified” humans from being harmed in labor markets. Or a lighter hand of regulation could require labeling of “humanoid” labor to allow the market to tailor purchasing according to humanist or humanoid values.
Society might take an entirely different approach: Allow humanoid labor to generate so much productivity and wealth that this labor is taxed, with proceeds set aside for a basic income. There are rudiments for such a wealth-spreading approach. Social Security is funded by employment taxes, effectively using payroll contributions of the current workforce to pay benefits to retirees. Unemployment insurance relies on employment taxes to fund short-term benefits for people who have lost their jobs. The Affordable Care Act is another example of a federal law that expands a benefit to provide what some people label a human right — a right to basic health care. Some European nations provide their citizens a basic income, a floor to ensure a minimum living standard and to guard against homelessness and hunger.
In sum, there is no shortage of employment law models to accommodate society’s dual aims of maximizing productivity — even with humanoid labor — while safeguarding the rest of the labor force from a competitive disadvantage.
V. CONCLUSION
The advent of the Internet in the late 20th Century was greeted with naïve anticipation. There was a broad consensus to leave this technology unregulated to foster innovation. Hindsight shows that a light regulatory hand unleashed one nightmare after another: the proliferation of hate speech and re-emergence of armed extremist groups extolling the virtues of free speech; the subversion of American democracy with Russian interference in 2016 and a nearly successful attempt to halt a transfer of presidential power in 2021; the rise of deadly disinformation about COVID-19 vaccines and the apocalyptic potential of QAnon; and more.
Do artificially intelligent enhancements to human labor pose similarly massive dislocations for the institution of employment? Only time will tell. However, it is important to realize that protective employment and labor laws were not enacted until long-after abuses had taken a great toll on the nation. Slavery endured more than 75 years after the United States Constitution was ratified in 1787, until ratification of the Thirteenth Amendment.[34] Peonage lasted well into the early 1900s.[35] Child labor was tolerated until the FLSA was enacted in 1938.[36] Racial segregation in the workplace was common until enactment of Title VII of the 1964 Civil Rights Act.[37] LGBTQ discrimination was not forcefully rejected until the Supreme Court ruled in 2020 that the nation’s sex discrimination law protects these sometimes vulnerable people.[38] Disability discrimination was the norm until the Americans with Disabilities Act was passed in 1990.[39]
This much is clear: Before new employment laws and regulations can be imagined with clarity, there is still time for engineers, technologists, technology companies, and humanists to design work-enhancements that mitigate the possibilities of dehumanizing labor. The more thought that goes into defining humanistic AI systems, the less need there will be for employment laws that mitigate harmful technologies.
[1] Professor, School of Labor and Employment Relations, and College of Law, University of Illinois at Urbana-Champaign.
>[2] Kaan Koseler & Matthew Stephan, Machine Learning Applications in Baseball: A Systematic Literature Review, 31 Applied Artificial Intelligence 745 (2018) (baseball is well suited for machine learning technologies, and innovations include the PITCHf/x system3, which tracks large amounts of data for each pitched ball).
[3] 2021 WL 5173764, at *3 (quoting lower court).
[4] Id. at *3 (quoting lower court).
[5] U.S. Department of Labor, Wage and Hour Division, Fact Sheet #17A: Exemption for Executive, Administrative, Professional, Computer & Outside Sales Employees Under the Fair Labor Standards Act (FLSA).
[6] Id.
[7] Independent Contractor Status Under the Fair Labor Standards Act (FLSA): Withdrawal, A Rule by the Wage and Hour Division, 86 Fed. Reg. 24303 (May 6, 2021), withdrawing the Trump administration’s more expansive definition of independent contractor in favor of a five-part test that emphasizes whether an entity controls the work of an individual, and whether the individual has an opportunity for profit or loss in their labor. Id. at 24306.
[8] Jeremias Adams-Prassl, What If Your Boss Was an Algorithm? Economic Incentives, Legal Challenges, and the Rise of Artificial Intelligence at Work, 41 Comp. Lab. L. Pol’y 123 (2019), at 132 (some software providers offer programs that “support and potentially automate management decision-making across all dimensions of work, including the full socioeconomic spectrum of workplaces”).
[9] Michael H. LeRoy, Misclassification under the Fair Labor Standards Act: Court Rulings and Erosion of the Employment Relationships, 2017 Univ. Chi. Legal Forum 327, 344-45 (2017) (“Before gig work is celebrated as the wave of the future, there are serious questions to answer about ensuring living wages for workers and obligating gig companies to bear societal costs associated with work that currently burden employers.”). Also see Cynthia Estlund, What Should We Do After Work? Automation and Employment Law, 128 Yale L.J. 254, 295 (2018) (“Especially at the bottom of the labor market, raising the floor on wages, benefits, and working conditions strengthens the business case for automation of technically automatable jobs.”).
[10] Daron Acemoglu & James Robinson, Why Nations Fail 182-83 (2012), quoting Queen Elizabeth I: “Thou aimest high, Master Lee. Consider thou what the invention could do to my poor subjects. It would assuredly bring to them ruin by depriving them of employment, thus making them beggars.”
[11] Carl Benedikt Frey & Michael A. Osborne, The Future of Employment: How Susceptible Are Jobs to Computerisation? 114 Technological Forecasting and Social Change 254 (2017), at 265.
[12] Artificial Intelligence Cuts Miss Rate for Colonoscopies, Physicians Weekly (May 2, 2022) (research published in Gastroenterology shows that artificial intelligence reduces the miss rate of colorectal neoplasia nearly in half); Louise Flintoft, Is AI the Future of Healthcare? Med-Tech (April 29, 2022) (AI imaging technology can also identify heart structure and function issues with 40% more accuracy than the human eye); and Rose McNulty, AI-Based Anomaly Detection Holds Promise in Screening for Retinal Diseases, AJMC (Jan. 16, 2022) (research on AI technologies for retinal screening show promising results).
[13] Caroline Hill, Deloitte Insight: Over 100,000 Legal Roles To Be Automated, Legal Insider (March 16, 2016), available in https://legaltechnology.com/2016/03/16/deloitte-insight-over-100000-legal-roles-to-be-automated/; and Steve Lohr, A.I. Is Doing Legal Work. But It Won’t Replace Lawyers, Yet., N.Y. Times (March 19, 2017) (reporting that the McKinsey Global Institute estimates that 23 percent of a lawyer’s job can be automated).
[14] Steven Johnson, A.I. Is Mastering Language. Should We Trust What It Says? N.T. Times (April 15, 2022) (GPT-3 and other neural nets can now write original prose).
[15] Oleksandr Melnychenko, Is Artificial Intelligence Ready to Assess an Enterprise’s Financial Security?, 13 J. Risk & Financial Manag. 191 (2020).
[16] National Labor Relations Act, 29 U.S.C. §164.
[17] U.S. Department of Labor, Wage and Hour Division, Fact Sheet #17A: Exemption for Executive, Administrative, Professional, Computer & Outside Sales Employees Under the Fair Labor Standards Act (FLSA).
[18] Id.
[19] Lola v. Skadden, Arps, Slate, Meagher & Flom LLP, No. 14-3845-cv, 2015 WL 4476828, at *2 (2d Cir. 2015) ruled that an attorney sufficiently alleged he did not engage in the practice of law, and therefore could state a claim against his employer for not paying overtime under the Fair Labor Standards Act. The work he performed — looking at documents for search terms; marking those documents in predetermined categories; and drawing black boxes to redact text — are functions that an AI program could probably do.
[20] James Wagner, Play Ball! Lockout Ends as M.L.B. and Union Strike a Deal, N.Y. Times (March 10, 2022).
[21] Bob Nightengale, MLB Power Shift Has Managers’ Salaries in Free Fall, USA Today (Aug. 27, 2018) (Mike Scioscia’s ten-year contract paying $50 million, and Joe Torre’s contract paying $7.5 million a year, had given way by 2018 to a labor market where 21 out of 30 major league managers earned $1.5 million or less).
[22] Rich Miller, Chicago School Principals Revive Unionization Bill, Push for Higher Pay, CapitolFax.com (Feb. 17, 2022), available in https://capitolfax.com/2022/02/17/chicago-school-principals-revive-unionization-bill-push-for-higher-pay/.
[23] Daniel Wiessner, New York Times Tech Workers Vote to Join Union, Reuters (March 4, 2022).
[24] Jonas Lutz, et al., Wearables for Integrative Performance and Tactic Analyses: Opportunities, Challenges, and Future Directions, Int. 17 J. Environ. Res. Public Health 59, 61 (2020).
[25] Id. at 78.
[26] For an overview of the transhuman movement and its opponents, see Stephen Lilley, Transhumanism and Society: The Social Debate Over Human Enhancement 2 (2013) (“transhumanists” advocate all human enhancements, while “conservationists” seek to preserve a time-honored conception of human life, available in https://digitalcommons.sacredheart.edu/cgi/viewcontent.cgi?article=1006&context=sociol_fac. Also see Daniel Lyons, Ray Kurzweil Wants to Be a Robot, Newsweek (May 16, 2009) (“Ray Kurzweil’s wildest dream is to be turned into a cyborg—a flesh-and-blood human enhanced with tiny, embedded computers, a man-machine hybrid with billions of microscopic nanobots coursing through his bloodstream.”).
[27] The Bionic Man was Science Fiction; The Bionic Hand Is Not, Mind Matters News (September 29, 2021), available in https://mindmatters.ai/2021/09/the-bionic-man-was-science-fiction-the-bionic-hand-is-not/.
[28] Scott Jenkins, Does Tommy John Surgery Give MLB Pitchers an Advantage? Sportscasting (July 29, 2020), at https://www.sportscasting.com/does-tommy-john-surgery-give-mlb-pitchers-an-advantage/, captures the debate surrounding the effects of this surgery, citing studies of improved performance by MLB pitchers and no effects. Also see Alva Noë, Is It Fair For Baseball To Reject Drugs But Embrace Surgery? NPR (July 26, 2013) at https://www.npr.org/sections/13.7/2013/07/25/205513618/is-it-fair-for-baseball-to-reject-drugs-but-embrace-surgery.
[29] Alan Blinder, Lia Thomas Wins an N.C.A.A. Swimming Title, N.Y. Times (March 17, 2022) (also reporting that a hurdler, CeCe Telfer, became the first transgender athlete to capture an N.C.A.A. championship). Nancy Hogshead-Makar, winner of three Olympic gold medals in swimming in the 1980s, joined dozens of other women swimmers who protested to the university that Thomas had “an unfair advantage over competition in the women’s category.” Id.
[30] Bill Chappell, Oscar Pistorius Makes Olympic History In 400 Meters, And Moves On To Semifinal, NPR (Aug. 4, 2012) (double-amputee runner finished second in a heat with five contestants).
[31] Promising New Developments in AI Prostheses Raise Stark Questions, Mind Matters News (March 22, 2022), available in https://mindmatters.ai/2022/03/promising-new-developments-in-ai-prostheses-raise-stark-questions/, also reporting on AI advances that improve the interface between a prosthesis and rest of an amputee’s limb).
[32] Daniel Wikler, Can We Learn from Eugenics? 25 J. of Med. Ethics 183, 184-185 (1999).
[33] See Robert Fogel & Stanley L. Engerman, Time on the Cross: The Economics of American Negro Slavery (1974), using data to argue that slaveowners were more rational and tempered than previously supposed because they had financial incentives to treat their slaves as productive assets.
[34] Nat’l Archives, 13th Amendment to the U.S. Constitution: Abolition of Slavery (1865) (passed by Congress on January 31, 1865, and ratified on December 6, 1865).
[35] Pollock v. Williams, 322 U.S. 4 (1944) (Florida criminal fraud statute violated the Thirteenth Amendment and the Anti-Peonage Act of 1867). More generally, see William Wirt Howe, The Peonage Cases, 4 Colum. L. Rev. (279) (1904).
[36] Fair Labor Standards Act of 1938, June 25, 1938, ch. 676, § 12, 52 Stat. 1067, codified at 29 U.S. Code § 212 (2018).
[37] Civil Rights Act of 1964 § 7, Pub. L. 88–352, title VII, § 701, July 2, 1964, 78 Stat. 253; codified at 42 U.S.C. § 2000e et seq (1964).
[38] Bostock v. Clayton County, __ U.S. __, 140 S.Ct. 1731 (2020).
[39] Americans with Disabilities Act of 1990 (ADA), Pub. L. 101–336, § 2, July 26, 1990, 104 Stat. 328; codified at 42 U.S.C. §§ 12101-12213 (2018).
