Property against Legality: Takings after Cedar Point

In the American constitutional tradition, a zealous judicial defense of property is closely aligned with the idea of “the rule of law.” Conventional wisdom holds that the Takings Clause of the Fifth Amendment vindicates both property rights and the rule of law by foreclosing arbitrary, lawless state action. But the standard story linking property rights, legality, and a constraint on arbitrary governance is more commonly stipulated than analyzed. This Article uses an apparent sharp break in takings jurisprudence, the United States Supreme Court’s June 2021 decision in Cedar Point Nursery v. Hassid, to closely scrutinize the relationship between legality and property rights. To that end, it offers first a careful analysis of the sharp rupture that Cedar Point makes in takings jurisprudence. Not only is the Court’s result difficult to explain in terms of precedent or traditional legal methods, it also destabilizes a previously settled and reasonably predictable litigation landscape. As a result, it seeds profound uncertainty on the legal ground because it signals a dissolution of the constraining effect formerly realized by standard tools of legal reasoning. There is, further, no obvious way for the Court to restore stability and predictability to the doctrine without drawing new, arbitrary lines. In consequence, takings law will likely abide in confusion, not certainty, for the foreseeable future. Cedar Point’s vindication of property rights hence comes at the paradoxical cost of dramatically increasing the space for decisions unguided by law by one group of officials in the judiciary.

A close reading of Cedar Point invites a more general and abstract analysis of the complex, nuanced relationship between the rule of law and property rights. Drawing on the general jurisprudential theories of H.L.A. Hart and other legal positivists, I use the decision as a launching point for a larger exploration of ways in which the rule of law can be incompletely realized to paradoxical and even socially harmful effect. Placing property at the center of the rule of law, I suggest, can be consistent with, or even an incitement to, serious derogation of the rule of law. Doing so can undermine rule-of-law goals, such as constraining arbitrary rule. This suggests a need to decenter property rights in accounts of the rule of law, and to explore, in more nuanced and grounded fashion, how the practice of judicial review mediates systemic values of legality and predictability. In short, if we value the rule of law, it may in general be appropriate to take a more skeptical, and so more contingent, view of both property as a legal institution, and also the courts as a source of legality and stability.

Introduction

Property and its zealous defense are closely associated with the rule of law in the American constitutional tradition. A “total, arbitrary, and capricious power” is conceived as the enemy of both.1.Kenneth R. Minogue, The Concept of Property and Its Contemporary Significance, in NomosXXII: Property 3, 5 (J. Roland Pennock & John W. Chapman eds., 1980).Show More In Federalist 70, Alexander Hamilton thus termed the constitutional “protection of property against those irregular and high handed combinations, which sometimes interrupt the ordinary course of justice.”2.The Federalist No. 70, at 471 (Alexander Hamilton) (Jacob E. Cooke ed., 1961).Show More Writing just after the Constitution’s ratification, James Madison warned that “just government” and “secure” property are imperiled by “arbitrary restrictions, exemptions, and monopolies [that] deny to part of its citizens that free use of their faculties . . . which not only constitute their property in the general sense of the word; but are the means of acquiring property strictly so called.”3.James Madison, Property, Nat’l Gazette, Mar. 27, 1792, reprinted in 14 The Papers of James Madison 266, 267 (Robert A. Rutland et al. eds., 1983).Show More John Adams agreed. He opined that “[p]roperty must be secured, or liberty cannot exist.”4.Cedar Point Nursery v. Hassid, 141 S. Ct. 2063, 2071 (2021) (quoting John Adams, Discourses on Davila, in 6 The Works of John Adams 223, 280 (Charles Francis Adams ed., 1851)).Show More In a leading constitutional treatise of the early Republic, St. George Tucker also identified the Takings Clause of the Fifth Amendment as a means “to restrain the arbitrary and oppressive mode of obtaining supplies for the army, and other public uses.”5.St. George Tucker, 1 Blackstone’s Commentaries: With Notes of Reference, to the Constitution and Laws, of the Federal Government of the United States; and of the Commonwealth of Virginia, app. at 305–06 (Augustus M. Kelley 1969) (1803); see also Armstrong v. United States, 364 U.S. 40, 49 (1960) (describing the Clause’s aim as preventing the government “from forcing some people alone to bear public burdens which, in all fairness and justice, should be borne by the public as a whole”).Show More The U.S. Supreme Court concurs. Citing Madison, Adams, and others, the Justices have posited that enforcement of the Takings Clause “empowers persons to shape and to plan their own destiny in a world where governments are always eager to do so for them.”6.Murrv.Wisconsin, 137 S. Ct. 1933, 1943 (2017); Cedar Point, 141 S. Ct. at 2071 (quoting Murr, 137 S. Ct. at 1943); see also Sveen v. Melin, 138 S. Ct. 1815, 1827 (2018) (Gorsuch, J., dissenting) (“Federalists like Madison countered that the rule of law permitted ‘property rights and liberty interests [to] be dissolved only by prospective laws of general applicability.’” (citation omitted)). The same theme is found in academic literature. See Richard A. Epstein, Design for Liberty: Private Property, Public Administration, and the Rule of Law 12 (2011) (describing a “close connection” between property’s protection and the rule of law); James W. Ely, Jr., Property Rights and Judicial Activism, 1 Geo. J.L. & Pub. Pol’y 125, 126 (2002) (“The Framers realized that robust protection of the rights of property owners undergirds liberty by diffusing power and protecting individual autonomy from governmental control.”).Show More Judicial protection of property rights serves the rule of law, on this account, by making state action predictable, by restraining arbitrary uses of state power, and by empowering citizens and others to chart their own “destinies” free of government control.

In this way, the idea of the rule of law, albeit not mentioned explicitly in the Constitution, has become “central to our political and rhetorical traditions, possibly even to our sense of national identity.”7.Richard H. Fallon, Jr., “The Rule of Law” as a Concept in Constitutional Discourse, 97 Colum. L. Rev. 1, 3 (1997).Show More Scholars keenly debate what the rule of law—which is also sometimes called the principle of legality8.See, e.g., Lon L. Fuller, The Morality of Law 44 (rev. ed. 1969) (using the phrase “the demands of legality” to capture the rule of law).Show More—requires.9.Jeremy Waldron, Is the Rule of Law an Essentially Contested Concept (in Florida)?, 21 Law & Phil. 137, 140–44 (2002) (surveying disputes over its meaning).Show More But a kernel of common ground is apparent: the rule of law is commonly defined as the law’s clarity, stability, and predictability.10 10.Fuller, supra note 8, at 39 (listing traits of the rule of law, including clarity and the capacity to be followed).Show More These qualities foster “confidence about the legal consequences of their actions.”11 11.Landgraf v. USI Film Prods., 511 U.S. 244, 266 (1994); see also Lawrence B. Solum, Equity and the Rule of Law, in Nomos XXXVI: The Rule of Law 120, 121 (Ian Shapiro ed., 1994) (including “generality, publicity, and regularity” among the rule of law’s features).Show More In contrast, state actions animated by “caprice, passion, bias, [or] prejudice” are all “antithetical to the rule of law.”12 12.TXO Prod. Corp. v. Alliance Res. Corp., 509 U.S. 443, 475 (1993) (O’Connor, J., dissenting). The association of the rule of law with the constraint of official action goes back to a Victorian legal theorist who was one of the early adopters of the term “rule of law.” A.V. Dicey, Introduction to the Study of the Law of the Constitution 181–205 (10th ed. 1959).Show More As such, they cannot be ranked as properly legalistic state action. Such improper—but not, note well, ipso facto illegal—species of official action are impossible if those with official power “exercise their power within a constraining framework of public norms, rather than on the basis of their own preferences, their own ideology, or their own individual sense of right and wrong.”13 13.Jeremy Waldron, The Concept and the Rule of Law, 43 Ga. L. Rev. 1, 6 (2008) [hereinafter Waldron, The Concept and the Rule].Show More Clear, predictable, and stable law binds officials at the same time that it guides citizens. Indeed, the two guidance functions are entwined. An official’s “arbitrary” exercise of power is not just the rule of law’s antipode.14 14.Joseph Raz, The Rule of Law and its Virtue, in The Authority of Law: Essays on Law and Morality 210, 224 (1979) [hereinafter Raz, Rule of Law and its Virtue] (“The law inevitably creates a great danger of arbitrary power—the rule of law is designed to minimize the danger created by the law itself.”).Show More It is also the antithesis of individual liberty.15 15.The kind of liberty protected by the rule of law is, again, contested and plural. See Brian Z. Tamanaha, On the Rule of Law: History, Politics, Theory 34–35 (2004) (distinguishing four kinds of freedom protected by the rule of law). The most relevant here are the freedom from unlicensed (and so unpredictable) legal constraint, and the preservation of a zone of personal autonomy. Id.Show More And so, in this standard narrative, there is a profound complementarity between the rule of law, property rights, and liberty.16 16.Some accounts go further and suggest that it is the propertied who will have the leisure and the resources to defend rule of laws. Minogue, supra note 1, at 8.Show More

This Article interrogates this standard account of how property rights and the rule of law relate to each other in American constitutional law. Its point of departure is a 2021 Supreme Court opinion about the constitutional status of property rights. That judgment is important in its own right. It changes, potentially quite dramatically, the scope of constitutional protection for real property under the Takings Clause of the Fifth Amendment.17 17.See U.S. Const. amend. V (stating that “private property [shall not] be taken for public use, without just compensation”).Show More Whatever its exact effects on the ground (literally!) may be, it certainly marks a sea change in Fifth Amendment jurisprudence.

Cedar Point Nursery v. Hassid18 18.141 S. Ct. 2063 (2021).Show More held that a 1976 California “take access” regulation permitting union organizers to approach and talk to agricultural workers was a taking requiring compensation under the Fifth Amendment.19 19.Cal. Code Regs. tit. 8, § 20900(e)(3) (2021); see also Cedar Point, 141 S. Ct. at 2069, 2072 (describing regulatory framework for agricultural labor under California law).Show More Part of a larger constitutional transformation sweeping over the separation of powers, the Religion Clauses, and abortion jurisprudence during the Roberts Court, Cedar Point has received less attention than other recent doctrinal convulsions. This neglect is unjustified.

A first contribution of this Article is to explore the possibility that the decision prefigures a dramatic and destabilizing shift in the nature of constitutional property. The legal uncertainty unleashed by that opinion is not likely to abate given the absence of any stable limit on the Court’s apparent reworking of the concept of constitutional property. Of larger theoretical significance, Cedar Point illustrates one way in which property rights and the rule of law can diverge, notwithstanding the standard story, to enlarge the scope for arbitrary state action. Hence, it invites the Article’s second, more theoretical contribution—a nuanced and careful theorization of property’s complex, many-stranded relation to the rule of law.

My analysis begins with the particulars of Cedar Point. The decision’s immediate effect, of course, was to change the terms for the increasingly beleaguered organized labor movement in one of the nation’s most important and fertile agricultural breadbaskets. Its longer-term, more abstract consequence was its implicit invitation for the future reworking of takings doctrine. In particular, the 2021 decision unraveled a central organizing conceit of takings jurisprudence. The latter has stabilized and channeled potential litigants’ expectations for decades. By diminishing the predictive value of precedent respecting property’s boundaries, the Court created uncertainty where property owners and officials previously had benefited from stable expectations.

Prior to Cedar Point, owners and officials got a reasonably clear sense of litigation outcomes by asking whether a state action was an “appropriation” or “regulation” as those words are used in everyday conversation. Under longstanding doctrine, legal challenges to appropriations generally prevailed under a “per se” rule. In contrast, the balancing test applicable to regulations typically, albeit far from inevitably, tilted in favor of the government. Distinguishing appropriations from regulations, moreover, was relatively straightforward when it came to real and chattel property. If the government indefinitely deprived you of the whole or part of the physical thing, you could typically expect to win a takings case.

To be sure, the doctrine elsewhere had other wrinkles. But in the vast majority of cases, these mattered only on 1L property exams and (very occasionally) in appellate litigation. The ensuing doctrine was relatively transparent, even to lay people unburdened by the intellectual pretensions of a legal education.

The Court in Cedar Point did not openly abandon the distinction between appropriations and regulations. More perplexingly, it invoked that distinction while refusing to deploy the ordinary meaning of an “appropriation.” While the verbal formulation of the law remained the same, a key doctrinal term with a clear and predictable lay meaning was replaced with an amorphous category of uncertain and unpredictable application. Legal and ordinary language hence parted company. By diminishing the clarifying force of doctrine, the Court created uncertainty where property owners and officials previously had reasonably stable expectations. This is one way in which Cedar Point revealed a tension between property rights and the clarity, stability, and predictability ambitions of the rule of law. Protecting the first can diminish the latter.

Another tension underlying the majority’s reasoning can be discerned by attending to its methodological choices: the Cedar Point Court ostentatiously relied on dictionaries, ordinary public meaning, and the binding force of precedent. But read closely, each of these argumentative threads unravel. Dictionary definitions were selectively picked; actual lay usage was ignored; and precedent was invoked only via selective quotation—distorting earlier constitutional holdings. In execution, therefore, Cedar Point’s vindication of property rights stood at odds with the application of familiar legal methods central to the rule of law.20 20.See Andrei Marmor, The Rule of Law and Its Limits, 23 Law & Phil. 1, 3 (2004).Show More Nor does the decision promise future clarity or stability.

Worse, Cedar Point did not ask or answer a crucial question: What distinguishes an impermissible “appropriation” from a “regulation” under the Takings Clause? The majority opinion offers hints. But none of these hold promise as a principled basis on which to draw a line between regulation and appropriation. Indeed, the ultimate, long-term effect of Cedar Point may well be to collapse the longstanding distinction between appropriations and regulations—a legal regime where most government action would be evaluated as an appropriation, even if it did not entail a physical invasion by the government. This doctrine’s end-state would dramatically expand judicial discretion, work avulsive change to the authority of state and local governments, and (ironically) foster fresh uncertainty about the resolution of inevitable and pervasive boundary disputes that arise in property law.

There is a bigger principle at stake here too. This close reading of Cedar Point further invites reappraisal of the way in which property and the rule of law have been theorized as working together in American constitutional law: Do they really intertwine as tightly as the standard story holds? A starting point for this analytic enterprise is a distinction drawn by the legal theorists H.L.A. Hart and Meir Dan-Cohen. Hart carved law up into “primary” rules applicable to the citizenry at large, and “secondary” rules that bind officials.21 21.H.L.A. Hart, The Concept of Law 99 (Penelope A. Bulloch & Joseph Raz postscript eds., 2d ed. 1994) [hereinafter Hart, Concept of Law] (“The union of primary and secondary rules is at the centre of a legal system . . . .”).Show More Dan-Cohen, reasoning in a similar vein, distinguished “conduct rules” covering everyone, and “decision rules” directed at officials.22 22.Meir Dan-Cohen, Decision Rules and Conduct Rules: On Acoustic Separation in Criminal Law, 97 Harv. L. Rev. 625, 627 (1984).Show More The law, both Hart and Dan-Cohen thereby insisted, speaks in subtly differently accented voices to the public and to its official custodians.

Cedar Point illustrates the possibility that the rule of law can come apart along this seam. The ordinary subjects of property law remain subject to a body of (somewhat more ambiguous) rules after Cedar Point, even as the Court shrugs off the disciplining constraints imposed by legal method. The rule of law can thus be roughly maintained for primary rules of property, even as it dissipates as a constraint upon officials. As such, we can observe what I call “first-order legality” for rules applicable to private persons, without the “second-order legality” usually experienced by officials. This dichotomy, and the resulting internal fracturing, complicates canonical accounts of the rule of law by showing how it is possible for the qualities of certainty, predictability, and stability to be maintained with respect to one domain of the law, but not another. Consequently, legality can be partial. At worst, it can potentially come to be at war with itself.

With this bifurcated account of the rule of law in hand, it is possible to interrogate in a more considered way the supposedly monotonic relationships between legality, arbitrary rule, and the ambition of legality. The rule of law, at least in one of its traditions, is often distilled into an image of rigid, impenetrable property rights.23 23.See sources cited supra notes 2–5.Show More These, in turn, are hitched to the aspiration for freedom from arbitrary rule and economic growth. But even as Cedar Point offered an account of property as a cornerstone of the rule of law in precisely these terms, each element of this argument was unraveling. No longer is it clear that centering legality around property minimizes the scope for arbitrary decision making by officials. To the contrary, an account of the rule of law centered around property rights may either increase or decrease the risk of such arbitrary rule without a clear effect on economic growth or social welfare. Legality, when conceptualized in terms of property rights, thus can undermine widely shared normative goals it purports to advance. Their relation is contingent, not necessary.

This Article focuses on the relationship of takings jurisprudence to the rule of law. Existing commentary criticizes Cedar Point’s “hostility to worker power” and “antidemocracy” effects,24 24.Nikolas Bowie, Comment, Antidemocracy, 135 Harv. L. Rev. 160, 163 (2021). For a similar, if more equivocal, suggestion, see Cristina M. Rodríguez, Foreword: Regime Change, 135 Harv. L. Rev. 2, 32 (2021) (“[T]he Supreme Court’s burgeoning jurisprudence . . . has turned to the Bill of Rights, primarily the First Amendment, to limit social welfare and good-government regulation . . . .”); see also Linda Greenhouse, Justice on the Brink: The Death of Ruth Bader Ginsburg, the Rise of Amy Coney Barrett, and Twelve Months That Transformed the Supreme Court 224 (2021) (describing the case as a “potentially transformational development in the law of property rights . . . likely to hobble government land use regulation”). For a contrary view, see Julia D. Mahoney, Cedar Point Nursery and the End of the New Deal Settlement, 11 Brigham-Kanner Prop. Rts. J. 43, 64 (2022) (arguing that “Cedar Point represents an evolution, not a revolution, in the Court’s property rights jurisprudence” while celebrating the Court’s result and analysis). As it will become clear, I respectfully disagree with Professor Mahoney’s conclusions for reasons spelled out at length in this Article.Show More or alternatively defends its “classical liberal” pedigree.25 25.Sam Spiegelman & Gregory C. Sisk, Cedar Point: Lockean Property and the Search for a Lost Liberalism, 2020–21 Cato Sup. Ct. Rev. 165, 178–81 (2021) (talking of “the takings muddle”).Show More In contrast, I explore the Court’s new takings doctrine in relation to the ideal of the rule of law.26 26.Cf. Cedar Point Nursery v. Hassid, 141 S. Ct. 2063, 2071 (2021) (“As John Adams tersely put it, ‘[p]roperty must be secured, or liberty cannot exist.’” (citation omitted)).Show More The property/rule of law connection has been previously explored in a set of lectures by Professor Jeremy Waldron. He, however, trains on the relation of Lockean accounts of property to the rule of law though a political philosophy lens.27 27.Jeremy Waldron, The Rule of Law and the Measure of Property 27 (2012) [hereinafter Waldron, Measure of Property] (doubting that “Locke’s account” of property yields a specific definition of the rule of law).Show More My analysis and conclusions unfold along a different, American-constitutional-law track. Nevertheless, like Waldron, I hope to contribute to larger theoretical debates about the rule of law and its constituent parts.

Part I explores the basic architecture of takings doctrine prior to Cedar Point. Part II then offers a close reading of that decision. I carefully analyze its methodological underpinnings and doctrinal aftermath. Both in its origin and in its reasoning, I demonstrate, the opinion is in sharp tension with legality norms. Part III then broadens the analytic lens to evaluate the role that property plays in understandings of the rule of law. Using Cedar Point as an opening wedge for inquiry, it demonstrates how legality can unravel in ways that foster arbitrary rule and undermine economic growth. Of course, these are precisely the outcomes the rule of law is intended to stave off.

  1.  Kenneth R. Minogue, The Concept of Property and Its Contemporary Significance, in Nomos XXII: Property 3, 5 (J. Roland Pennock & John W. Chapman eds., 1980).
  2. The Federalist No. 70, at 471 (Alexander Hamilton) (Jacob E. Cooke ed., 1961).
  3. James Madison, Property, Nat’l Gazette, Mar. 27, 1792, reprinted in 14 The Papers of James Madison 266, 267 (Robert A. Rutland et al. eds., 1983).
  4.  Cedar Point Nursery v. Hassid, 141 S. Ct. 2063, 2071 (2021) (quoting John Adams, Discourses on Davila, in 6 The Works of John Adams 223, 280 (Charles Francis Adams ed., 1851)).
  5.  St. George Tucker, 1 Blackstone’s Commentaries: With Notes of Reference, to the Constitution and Laws, of the Federal Government of the United States; and of the Commonwealth of Virginia, app. at 305–06 (Augustus M. Kelley 1969) (1803); see also Armstrong v. United States, 364 U.S. 40, 49 (1960) (describing the Clause’s aim as preventing the government “from forcing some people alone to bear public burdens which, in all fairness and justice, should be borne by the public as a whole”).
  6. Murr v. Wisconsin, 137 S. Ct. 1933, 1943 (2017); Cedar Point, 141 S. Ct. at 2071 (quoting Murr, 137 S. Ct. at 1943); see also Sveen v. Melin, 138 S. Ct. 1815, 1827 (2018) (Gorsuch, J., dissenting) (“Federalists like Madison countered that the rule of law permitted ‘property rights and liberty interests [to] be dissolved only by prospective laws of general applicability.’” (citation omitted)). The same theme is found in academic literature. See Richard A. Epstein, Design for Liberty: Private Property, Public Administration, and the Rule of Law 12 (2011) (describing a “close connection” between property’s protection and the rule of law); James W. Ely, Jr., Property Rights and Judicial Activism, 1 Geo. J.L. & Pub. Pol’y 125, 126 (2002) (“The Framers realized that robust protection of the rights of property owners undergirds liberty by diffusing power and protecting individual autonomy from governmental control.”).
  7. Richard H. Fallon, Jr., “The Rule of Law” as a Concept in Constitutional Discourse, 97 Colum. L. Rev. 1, 3 (1997).
  8. See, e.g., Lon L. Fuller, The Morality of Law 44 (rev. ed. 1969) (using the phrase “the demands of legality” to capture the rule of law).
  9. Jeremy Waldron, Is the Rule of Law an Essentially Contested Concept (in Florida)?, 21 Law & Phil. 137, 140–44 (2002) (surveying disputes over its meaning).
  10. Fuller, supra note 8, at 39 (listing traits of the rule of law, including clarity and the capacity to be followed).
  11. Landgraf v. USI Film Prods., 511 U.S. 244, 266 (1994); see also Lawrence B. Solum, Equity and the Rule of Law, in Nomos XXXVI: The Rule of Law 120, 121 (Ian Shapiro ed., 1994) (including “generality, publicity, and regularity” among the rule of law’s features).
  12.  TXO Prod. Corp. v. Alliance Res. Corp., 509 U.S. 443, 475 (1993) (O’Connor, J., dissenting). The association of the rule of law with the constraint of official action goes back to a Victorian legal theorist who was one of the early adopters of the term “rule of law.” A.V. Dicey, Introduction to the Study of the Law of the Constitution 181–205 (10th ed. 1959).
  13. Jeremy Waldron, The Concept and the Rule of Law, 43 Ga. L. Rev. 1, 6 (2008) [hereinafter Waldron, The Concept and the Rule].
  14. Joseph Raz, The Rule of Law and its Virtue, in The Authority of Law: Essays on Law and Morality 210, 224 (1979) [hereinafter Raz, Rule of Law and its Virtue] (“The law inevitably creates a great danger of arbitrary power—the rule of law is designed to minimize the danger created by the law itself.”).
  15. The kind of liberty protected by the rule of law is, again, contested and plural. See Brian Z. Tamanaha, On the Rule of Law: History, Politics, Theory 34–35 (2004) (distinguishing four kinds of freedom protected by the rule of law). The most relevant here are the freedom from unlicensed (and so unpredictable) legal constraint, and the preservation of a zone of personal autonomy. Id.
  16. Some accounts go further and suggest that it is the propertied who will have the leisure and the resources to defend rule of laws. Minogue, supra note 1, at 8.
  17. See U.S. Const. amend. V (stating that “private property [shall not] be taken for public use, without just compensation”).
  18. 141 S. Ct. 2063 (2021).
  19. Cal. Code Regs. tit. 8, § 20900(e)(3) (2021); see also Cedar Point, 141 S. Ct. at 2069, 2072 (describing regulatory framework for agricultural labor under California law).
  20. See Andrei Marmor, The Rule of Law and Its Limits, 23 Law & Phil. 1, 3 (2004).
  21. H.L.A. Hart, The Concept of Law 99 (Penelope A. Bulloch & Joseph Raz postscript eds., 2d ed. 1994) [hereinafter Hart, Concept of Law] (“The union of primary and secondary rules is at the centre of a legal system . . . .”).
  22. Meir Dan-Cohen, Decision Rules and Conduct Rules: On Acoustic Separation in Criminal Law, 97 Harv. L. Rev. 625, 627 (1984).
  23. See sources cited supra notes 2–5.
  24. Nikolas Bowie, Comment, Antidemocracy, 135 Harv. L. Rev. 160, 163 (2021). For a similar, if more equivocal, suggestion, see Cristina M. Rodríguez, Foreword: Regime Change, 135 Harv. L. Rev. 2, 32 (2021) (“[T]he Supreme Court’s burgeoning jurisprudence . . . has turned to the Bill of Rights, primarily the First Amendment, to limit social welfare and good-government regulation . . . .”); see also Linda Greenhouse, Justice on the Brink: The Death of Ruth Bader Ginsburg, the Rise of Amy Coney Barrett, and Twelve Months That Transformed the Supreme Court 224 (2021) (describing the case as a “potentially transformational development in the law of property rights . . . likely to hobble government land use regulation”). For a contrary view, see Julia D. Mahoney, Cedar Point Nursery and the End of the New Deal Settlement, 11 Brigham-Kanner Prop. Rts. J. 43, 64 (2022) (arguing that “Cedar Point represents an evolution, not a revolution, in the Court’s property rights jurisprudence” while celebrating the Court’s result and analysis). As it will become clear, I respectfully disagree with Professor Mahoney’s conclusions for reasons spelled out at length in this Article.
  25. Sam Spiegelman & Gregory C. Sisk, Cedar Point: Lockean Property and the Search for a Lost Liberalism, 2020–21 Cato Sup. Ct. Rev. 165, 178–81 (2021) (talking of “the takings muddle”).
  26. Cf. Cedar Point Nursery v. Hassid, 141 S. Ct. 2063, 2071 (2021) (“As John Adams tersely put it, ‘[p]roperty must be secured, or liberty cannot exist.’” (citation omitted)).
  27. Jeremy Waldron, The Rule of Law and the Measure of Property 27 (2012) [hereinafter Waldron, Measure of Property] (doubting that “Locke’s account” of property yields a specific definition of the rule of law).

A Right to a Human Decision

Recent advances in computational technologies have spurred anxiety about a shift of power from human to machine decision makers. From welfare and employment to bail and other risk assessments, state actors increasingly lean on machine-learning tools to directly allocate goods and coercion among individuals. Machine-learning tools are perceived to be eclipsing, even extinguishing, human agency in ways that compromise important individual interests. An emerging legal response to such worries is to assert a novel right to a human decision. European law embraced the idea in the General Data Protection Regulation. American law, especially in the criminal justice domain, is moving in the same direction. But no jurisdiction has defined with precision what that right entails, furnished a clear justification for its creation, or defined its appropriate domain.

This Article investigates the legal possibilities and normative appeal of a right to a human decision. I begin by sketching its conditions of technological plausibility. This requires the specification of both a feasible domain of machine decisions and the margins along which machine decisions are distinct from human ones. With this technological accounting in hand, I analyze the normative stakes of a right to a human decision. I consider four potential normative justifications: (a) a concern with population-wide accuracy; (b) a grounding in individual subjects’ interests in participation and reason giving; (c) arguments about the insufficiently reasoned or individuated quality of state action; and (d) objections grounded in negative externalities. None of these yields a general justification for a right to a human decision. Instead of being derived from normative first principles, limits to machine decision making are appropriately found in the technical constraints on predictive instruments. Within that domain, concerns about due process, privacy, and discrimination in machine decisions are typically best addressed through a justiciable “right to a well-calibrated machine decision.”

Introduction

Every tectonic technological change—from the first grain domesticated to the first smartphone set abuzz1.For recent treatments of these technological causes of social transformations, see generally James C. Scott, Against the Grain: A Deep History of the Earliest States (2017), and Ravi Agrawal, India Connected: How the Smartphone is Transforming the World’s Largest Democracy (2018).Show More—begets a new society. Among the ensuing birth pangs are novel anxieties about how power is distributed—how it is to be gained, and how it will be lost. A spate of sudden advances in the computational technology known as machine learning has stimulated the most recent rush of inky public anxiety. These new technologies apply complex algorithms,2.An algorithm is simply a “well-defined set of steps for accomplishing a certain goal.” Joshua A. Kroll et al., Accountable Algorithms, 165 U. Pa. L. Rev. 633, 640 n.14 (2017); see also Thomas H. Cormen et al., Introduction to Algorithms 5 (3d ed. 2009) (defining an algorithm as “any well-defined computational procedure that takes some value, or set of values, as input and produces some value, or set of values, as output” (emphasis omitted)). The task of computing, at its atomic level, comprises the execution of serial algorithms. Martin Erwig, Once Upon an Algorithm: How Stories Explain Computing 1–4 (2017).Show More called machine-learning instruments, to vast pools of public and government data so as to execute tasks previously beyond mere human ability.3.Machine learning is a general purpose technology that, in broad terms, encompasses “algorithms and systems that improve their knowledge or performance with experience.” Peter Flach, Machine Learning: The Art and Science of Algorithms that Make Sense of Data 3 (2012); see also Ethem Alpaydin, Introduction to Machine Learning 2–3 (3d ed. 2014) (defining machine learning in similar terms). For the uses of machine learning, see Susan Athey, Beyond Prediction: Using Big Data for Policy Problems, 355 Science 483, 483 (2017) (noting the use of machine learning to solve prediction problems). I discuss the technological scope of the project, and define relevant terms, infra at text accompanying note 111. I will use the terms “algorithmic tools” and “machine learning” interchangeably, even though the class of algorithms is technically much larger.Show More Corporate and state actors increasingly lean on these tools to make “decisions that affect people’s lives and livelihoods—from loan approvals, to recruiting, legal sentencing, and college admissions.”4.Kartik Hosanagar & Vivian Jair, We Need Transparency in Algorithms, But Too Much Can Backfire, Harv. Bus. Rev. (July 23, 2018), https://hbr.org/2018/07/we-need-transparency-in-algorithms-but-too-much-can-backfire [https://perma.cc/7KQ9-QMF3]; accord Cary Coglianese & David Lehr, Regulating by Robot: Administrative Decision Making in the Machine-Learning Era, 105 Geo. L.J. 1147, 1149 (2017).Show More

As a result, many people feel a loss of control over key life decisions.5.Shoshana Zuboff, Big Other: Surveillance Capitalism and the Prospects of an Information Civilization, 30 J. Info. Tech. 75, 75 (2015) (describing a “new form of information capitalism [that] aims to predict and modify human behavior as a means to produce revenue and market control”).Show More Machines, they fear, resolve questions of critical importance on grounds that are beyond individuals’ ken or control.6.See, e.g., Rachel Courtland, The Bias Detectives, 558 Nature 357, 357 (2018) (documenting concerns among the public that algorithmic risk scores for detecting child abuse fail to account for an “effort . . . to turn [a] life around”).Show More Many individuals experience a loss of elementary human agency and a corresponding vulnerability to an inhuman and inhumane machine logic. For some, “the very idea of an algorithmic system making an important decision on the basis of past data seem[s] unfair.”7.Reuben Binns et al., ‘It’s Reducing a Human Being to a Percentage’; Perceptions of Justice in Algorithmic Decisions, 2018 CHI Conf. on Hum. Factors Computing Systems 9 (emphasis omitted).Show More Machines, it is said, want fatally for “empathy.”8.Virginia Eubanks, Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor 168 (2017).Show More For others, machine decisions seem dangerously inscrutable, non-transparent, and so hazardously unpredictable.9.Will Knight, The Dark Secret at the Heart of AI, MIT Tech. Rev. (Apr. 11, 2017), https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/ [https://perma.cc/L94L-LYTJ] (“The computers that run those services have programmed themselves, and they have done it in ways we cannot understand. Even the engineers who build these apps cannot fully explain their behavior.”).Show More Worse, governments and companies wield these tools freely to taxonomize their populations, predict individual behavior, and even manipulate behavior and preferences in ways that give them a new advantage over the human subjects of algorithmic classification.10 10.For consideration of these issues, see Mariano-Florentino Cuéllar & Aziz Z. Huq, Economies of Surveillance, 133 Harv. L. Rev. 1280 (2020), and Mariano-Florentino Cuéllar & Aziz Z. Huq, Privacy’s Political Economy and the State of Machine Learning: An Essay in Honor of Stephen J. Schulhofer, N.Y.U. Ann. Surv. Am. L. (forthcoming 2020).Show More Even the basic terms of political choice seem compromised.11 11.See, e.g., Daniel Kreiss & Shannon C. McGregor, Technology Firms Shape Political Communication: The Work of Microsoft, Facebook, Twitter, and Google with Campaigns During the 2016 U.S. Presidential Cycle, 35 Pol. Comm. 155, 156–57 (2018) (describing the role of technology firms in shaping campaigns).Show More At the same time that machine learning is poised to recalibrate the ordinary forms of interaction between citizen and government (or big tech), advances in robotics as well as machine learning appear to be about to displace huge tranches of both blue-collar and white-collar labor markets.12 12.For what has become the standard view, see Larry Elliott, Robots Will Take Our Jobs. We’d Better Plan Now, Before It’s Too Late, Guardian (Feb. 1, 2018, 1:00 AM), https://www.theguardian.com/commentisfree/2018/feb/01/robots-take-our-jobs-amazon-go-seattle [https://perma.cc/2CFP-3JJV]. For a more nuanced account, see Martin Ford, Rise of the Robots: Technology and the Threat of a Jobless Future 282–83 (2015).Show More A fearful future looms, one characterized by massive economic dislocation, wherein people have lost control of many central life choices, and basic consumer and political preferences are no longer really one’s own.

This Article is about one nascent and still inchoate legal response to these fears: the possibility that an individual being assigned a benefit or a coercive intervention has a right to a human decision rather than a decision reached by a purely automated process (a “machine decision”). European law has embraced the idea. American law, especially in the criminal justice domain, is flirting with it.13 13.See infra text accompanying notes 70–73.Show More My aim in this Article is to test this burgeoning proposal, to investigate its relationship with technological possibilities, and to ascertain whether it is a cogent response to growing distributional, political, and epistemic anxieties. My focus is not on the form of such a right—statutory, constitutional, or treaty-based—or how it is implemented—say, in terms of liability or property rule protection—but more simply on what might ab initio justify its creation.

To motivate this inquiry, consider some of the anxieties unfurling already in public debate: A nursing union, for instance, launched a campaign urging patients to demand human medical judgments rather than technological assessment.14 14.‘When It Matters Most, Insist on a Registered Nurse,’ Nat’l Nurses United, https://www.­nationalnursesunited.org/insist-registered-nurse [https://perma.cc/MB66-XTXW] (last visited Jan. 19, 2020).Show More And a majority of patients surveyed in a 2018 Accenture survey preferred treatment by a doctor in person to virtual care.15 15.Accenture Consulting, 2018 Consumer Survey on Digital Health: US Results 9 (2018), https://www.accenture.com/_acnmedia/PDF-71/Accenture-Health-2018-Consumer-Survey-Digital-Health.pdf#zoom=50 [https://perma.cc/TU5F-9J82].Show More When California proposed replacing money bail with a “risk-based pretrial assessment” tool, a state court judge warned that “[t]echnology cannot replace the depth of judicial knowledge, experience, and expertise in law enforcement that prosecutors and defendants’ attorneys possess.”16 16.Quentin L. Kopp, Replacing Judges with Computers Is Risky, Harv. L. Rev. Blog (Feb. 20, 2018), https://blog.harvardlawreview.org/replacing-judges-with-computers-is-risky/ [https://perma.cc/WS5S-ARVF]. On the current state of affairs, see California Set to Greatly Expand Controversial Pretrial Risk Assessments, Filter (Aug. 7, 2019), https://filtermag.org/­california-slated-to-greatly-expand-controversial-pretrial-risk-assessments/ [https://perma.cc­/2FNX-U3C9].Show More In 2018, the City of Flint, Michigan, discontinued the use of a highly effective machine-learning tool designed to identify defective water pipes, reverting under community pressure to human decision making with a far lower hit rate for detecting defective pipes.17 17.Alexis C. Madrigal, How a Feel-Good AI Story Went Wrong in Flint, Atlantic (Jan. 3, 2019), https://www.theatlantic.com/technology/archive/2019/01/how-machine-learning-fou­nd-flints-lead-pipes/578692/ [https://perma.cc/V8VA-F22W].Show More Finally, and perhaps most powerfully, consider the worry congealed in an anecdote told by data scientist Cathy O’Neil: An Arkansas woman named Catherine Taylor is denied federal housing assistance because she fails an automated, “webcrawling[,] data-gathering” background check.18 18.Cathy O’Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy 152–53 (2016).Show More It is only when “one conscientious human being” takes the trouble to look into the quality of this machine result that it is discovered that Taylor has been red-flagged in error.19 19.Id. at 153.Show More O’Neil’s plainly troubling anecdote powerfully captures the fear that machines will be unfair, incomprehensive, or incompatible with the flexing of elementary human agency: it provides a sharp spur to the inquiry that follows.

The most important formulation of a right to a human decision to date is found in European law. In April 2016, the European Parliament enacted a new regime of data protection in the form of a General Data Protection Regulation (GDPR).20 20.Regulation 2016/679, of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data, and Repealing Directive 95/46/EC (General Data Protection Regulation), 2016 O.J. (L 119) (EU) [hereinafter GDPR]; see also Christina Tikkinen-Piri, Anna Rohunen & Jouni Markkula, EU General Data Protection Regulation: Changes and Implications for Personal Data Collecting Companies, 34 Computer L. & Security Rev. 134, 134–35 (2018) (documenting the enactment process of the GDPR).Show More Unlike the legal regime it superseded,21 21.See Directive 95/46, of the European Parliament and of the Council of 24 October 1995 on the Protection of Individuals with Regard to the Processing of Personal Data and on the Free Movement of Such Data, art. 1, 1995 O.J. (L 281) (EC) [hereinafter Directive 95/46].Show More the GDPR as implemented in May 2018 is legally mandatory even in the absence of implementing legislation by member states of the European Union (EU).22 22.Bryce Goodman & Seth Flaxman, European Union Regulations on Algorithmic Decision Making and a “Right to Explanation,” AI Mag., Fall 2017, at 51–52 (explaining the difference between a non-binding directive and a legally binding regulation under European law).Show More Hence, it can be directly enforced in court through hefty financial penalties.23 23.Id. at 52.Show More Article 22 of the GDPR endows a natural individual with “the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her.”24 24.GDPR, supra note 20, arts. 4(1), 22(1) (inter alia, defining “data subject”).Show More That right covers private and some (but not all) state entities.25 25.See id. art. 4(7)–(8) (defining “controller” and “processor” as key scope terms). The Regulation, however, does not apply to criminal and security investigations. Id. art. 2(2)(d).Show More On its face, it fashions an opt-out of quite general scope from automated decision making.26 26.As I explain below, this is not the only provision of the GDPR that can be interpreted to create a right to a human decision. See infra text accompanying notes 53–58.Show More

The GDPR also has extraterritorial effect.27 27.GDPR, supra note 20, art. 3.Show More It reaches platforms, such as Google and Facebook, that offer services within the EU.28 28.There is sharp divergence in the scholarship over the GDPR’s extraterritorial scope, which ranges from the measured, see Griffin Drake, Note, Navigating the Atlantic: Understanding EU Data Privacy Compliance Amidst a Sea of Uncertainty, 91 S. Cal. L. Rev. 163, 166 (2017) (documenting new legal risks to American companies pursuant to the GDPR), to the alarmist, see Mira Burri, The Governance of Data and Data Flows in Trade Agreements: The Pitfalls of Legal Adaptation, 51 U.C. Davis L. Rev. 65, 92 (2017) (“The GDPR is, in many senses, excessively burdensome and with sizeable extraterritorial effects.”).Show More And American law is also making tentative moves toward a similar right to a human decision. In 2016, for example, the Wisconsin Supreme Court held that an algorithmically generated risk score “may not be considered as the determinative factor in deciding whether the offender can be supervised safely and effectively in the community” as a matter of due process.29 29.State v. Loomis, 881 N.W.2d 749, 760 (Wis. 2016).Show More That decision precludes full automation of bail determinations. There must be a human judge in the loop. The Wisconsin court’s holding is unlikely to prove unique. State deployment of machine learning has, more generally, elicited sharp complaints sounding in procedural justice and fairness terms.30 30.See, e.g., Julia Angwin, Jeff Larson, Surya Mattu & Lauren Kirchner, Machine Bias: There’s Software Used Across the Country to Predict Future Criminals. And It’s Biased Against Blacks, ProPublica 2 (May 23, 2016), https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing [https://perma.cc/Q9ZU-VY6J] (criticizing machine-learning instruments in the criminal justice context).Show More Further, the Sixth Amendment’s right to a jury trial has to date principally been deployed to resist judicial factfinding.31 31.See, e.g., Apprendi v. New Jersey, 530 U.S. 466, 477 (2000) (explaining that the Fifth and Sixth Amendments “indisputably entitle a criminal defendant to a jury determination that [he] is guilty of every element of the crime with which he is charged, beyond a reasonable doubt” (alteration in original) (internal quotation marks omitted) (quoting United States v. Gaudin, 515 U.S. 506, 510 (1995))).Show More But there is no conceptual reason why the Sixth Amendment could not be invoked to preclude at least some forms of algorithmically generated inputs to criminal sentencing. Indeed, it would seem to follow a fortiori that a right precluding a jury’s substitution with a judge would also block its displacement by a mere machine.

In this Article, I start by situating a right to a human decision in its contemporary technological milieu. I can thereby specify the feasible domain of machine decisions. I suggest this comprises decisions taken at high volume in which sufficient historical data exists to generate effective predictions. Importantly, this excludes many matters presently resolved through civil or criminal trials but sweeps in welfare determinations, hiring decisions, and predictive judgments in the criminal justice contexts of bail and sentencing. Second, I examine the margins along which machine decisions are distinct from human ones. My focus is on a group of related technologies known as machine learning. This is the form of artificial intelligence diffusing most rapidly today.32 32.See infra text accompanying note 88 (defining machine learning). I am not alone in this focus. Legal scholars are paying increasing attention to new algorithmic technologies. For leading examples, see Kate Crawford & Jason Schultz, Big Data and Due Process: Toward a Framework to Redress Predictive Privacy Harms, 55 B.C. L. Rev. 93, 109 (2014) (arguing for “procedural data due process [to] regulate the fairness of Big Data’s analytical processes with regard to how they use personal data (or metadata . . . )”); Andrew Guthrie Ferguson, Big Data and Predictive Reasonable Suspicion, 163 U. Pa. L. Rev. 327, 383–84 (2015) (discussing the possible use of algorithmic prediction in determining “reasonable suspicion” in criminal law); Kroll et al., supra note 2, at 636–37; Michael L. Rich, Machine Learning, Automated Suspicion Algorithms, and the Fourth Amendment, 164 U. Pa. L. Rev. 871, 929 (2016) (developing a “framework” for integrating machine-learning technologies into Fourth Amendment analysis).Show More A right to a human decision cannot be defined or evaluated without some sense of the technical differences between human decision making and decisions reached by these machine-learning technologies. Indeed, careful analysis of how machine learning is designed and implemented reveals that the distinctions between human and machine decisions are less crisp than might first appear. Claims about a right to human decision, I suggest, are better understood to turn on the timing, and not the sheer fact, of such involvement.

With this technical foundation in hand, I evaluate the right to a human decision in relation to four normative ends it might plausibly be understood to further. A first possibility turns on overall accuracy worries. My second line of analysis takes up the interests of an individual exposed to a machine decision. The most pertinent of these interests hinge upon an individual’s participation in decision making and her opportunity to offer reasons. A third analytic salient tracks ways that a machine instrument might be intrinsically objectionable because it uses a deficient decisional protocol. I focus here on worries about the absence of individualized consideration and a machine’s failure to offer reasoned judgments. Finally, I consider dynamic, system-level effects (i.e., negative spillovers), in particular in relation to social power. None of these arguments ultimately provides sure ground for a legal right to a human decision.

Rather, I suggest that the limits of machine decision making be plotted based on its technical constraints. Machines should not be used when there is no tractable parameter amenable to prediction. For example, if there is no good parameter that tracks job performance, then machine evaluation of those employees should be abandoned. Nor should they be used when decision making entails ethical or otherwise morally charged judgments. Most important, I suggest that machine decisions should be subject to a right to a well-calibrated machine decision that folds in due process, privacy, and equality values.33 33.A forthcoming companion piece develops a more detailed account of how this right would be vindicated in practice through a mix of litigation and regulation. See Aziz Z. Huq, Constitutional Rights in the Machine Learning State, 105 Cornell L. Rev. (forthcoming 2020).Show More This is a better response than a right to a human decision to the many instruments now implemented by the government that are highly flawed.34 34.For a catalog, see Meredith Whittaker et al., AI Now Inst., AI Now Report 2018, at 18–22 (2018), https://ainowinstitute.org/AI_Now_2018_Report.pdf [https://perma.cc/2BCG-M4­54].Show More

My analysis here focuses on state action that imposes benefits or coercion on individuals—and not on either private action or a broader array of state action—for three reasons. First, salient U.S. legal frameworks, unlike the GDPR’s coverage, are largely (although not exclusively) trained on state action. Accordingly, a focus on state action makes sense in terms of explaining and evaluating the current U.S. regulatory landscape. Second, the range of private uses of algorithmic tools is vast and heterogenous. Algorithms are now deployed in private activities ranging from Google’s PageRank instrument,35 35.See, e.g., David Segal, The Dirty Little Secrets of Search: Why One Retailer Kept Popping Up as No. 1, N.Y. Times, Feb. 13, 2011, at BU1.Show More to “fintech” applied to generate new revenue streams,36 36.See Falguni Desai, The Age of Artificial Intelligence in Fintech, Forbes (June 30, 2016, 10:42 PM), http://www.forbes.com/sites/falgunidesai/2016/06/30/the-age-of-artificial-intelli­gence-in-fintech [https://perma.cc/DG8N-8NVS] (describing how fintech firms use artificial intelligence to improve investment strategies and analyze consumer financial activity).Show More to medical instruments used to calculate stroke risk,37 37.See, e.g., Benjamin Letham, Cynthia Rudin, Tyler H. McCormick & David Madigan, Interpretable Classifiers Using Rules and Bayesian Analysis: Building a Better Stroke Prediction Model, 9 Annals Applied Stat. 1350, 1350 (2015).Show More to engineers’ identification of new stable inorganic compounds.38 38.See, e.g., Paul Raccuglia et al., Machine-Learning-Assisted Materials Discovery Using Failed Experiments, 533 Nature 73, 73 (2016) (identifying new vanadium compounds).Show More Algorithmic tools are also embedded within new applications, such as voice recognition software, translation software, and visual recognition systems.39 39.Yann LeCun et al., Deep Learning, 521 Nature 436, 438–41 (2015).Show More In contrast, the state is to date an unimaginative user of machine learning, with a relatively constrained domain of deployments.40 40.See infra text accompanying notes 117–21 (describing state uses of machine learning).Show More This makes for a more straightforward analysis. Third, where the state does use algorithmic tools, it often results directly or indirectly in deprivations of liberty, freedom of movement, bodily integrity, or basic income. These normatively freighted machine decisions present arguably the most compelling circumstances for adopting a right to a human decision and so are a useful focus of normative inquiry.

The Article proceeds in three steps. Part I catalogs ways in which law has crafted, or could craft, a right to a human decision. This taxonomical enterprise demonstrates that such a right is far from fanciful. Part II defines the class of computational tools to be considered, explores the manner in which such instruments can be used, and teases out how they are (or are not) distinct from human decisions. Doing so helps illuminate the plausible forms of a right to a human decision. Part III then turns to the potential normative foundations of such a right. It provides a careful taxonomy of those grounds. It then shows why they all fall short. Finally, a brief conclusion inverts the Article’s analytic lens to gesture at the possibility that a right to a well-calibrated machine decision can be imagined, and even defended, on more persuasive terms than a right to a human decision.

  1. * Frank and Bernice J. Greenberg Professor of Law, University of Chicago Law School. Thanks to Faith Laken for terrific research aid. Thanks to Tony Casey, David Driesen, Lauryn Gouldin, Daniel Hemel, Darryl Li, Anup Malani, Richard McAdams, Eric Posner, Julie Roin, Lior Strahilevitz, Rebecca Wexler, and Annette Zimmermann for thoughtful conversation. Workshop participants at the University of Chicago, Stanford Law School, the University of Houston, William and Mary Law School, and Syracuse University School of Law also provided thoughtful feedback. I am grateful to Christiana Zgourides, Erin Brown, and the other law review editors for their careful work on this Article. All errors are mine, not the machine’s.
  2. For recent treatments of these technological causes of social transformations, see generally James C. Scott, Against the Grain: A Deep History of the Earliest States (2017), and Ravi Agrawal, India Connected: How the Smartphone is Transforming the World’s Largest Democracy (2018).
  3. An algorithm is simply a “well-defined set of steps for accomplishing a certain goal.” Joshua A. Kroll et al., Accountable Algorithms, 165 U. Pa. L. Rev. 633, 640 n.14 (2017); see also Thomas H. Cormen et al., Introduction to Algorithms 5 (3d ed. 2009) (defining an algorithm as “any well-defined computational procedure that takes some value, or set of values, as input and produces some value, or set of values, as output” (emphasis omitted)). The task of computing, at its atomic level, comprises the execution of serial algorithms. Martin Erwig, Once Upon an Algorithm: How Stories Explain Computing 1–4 (2017).
  4. Machine learning is a general purpose technology that, in broad terms, encompasses “algorithms and systems that improve their knowledge or performance with experience.” Peter Flach, Machine Learning: The Art and Science of Algorithms that Make Sense of Data 3 (2012); see also Ethem Alpaydin, Introduction to Machine Learning 2–3 (3d ed. 2014) (defining machine learning in similar terms). For the uses of machine learning, see Susan Athey, Beyond Prediction: Using Big Data for Policy Problems, 355 Science 483, 483 (2017) (noting the use of machine learning to solve prediction problems). I discuss the technological scope of the project, and define relevant terms, infra at text accompanying note 111. I will use the terms “algorithmic tools” and “machine learning” interchangeably, even though the class of algorithms is technically much larger.
  5. Kartik Hosanagar & Vivian Jair, We Need Transparency in Algorithms, But Too Much Can Backfire, Harv. Bus. Rev. (July 23, 2018), https://hbr.org/2018/07/we-need-transparency-in-algorithms-but-too-much-can-backfire [https://perma.cc/7KQ9-QMF3]; accord Cary Coglianese & David Lehr, Regulating by Robot: Administrative Decision Making in the Machine-Learning Era, 105 Geo. L.J. 1147, 1149 (2017).
  6. Shoshana Zuboff, Big Other: Surveillance Capitalism and the Prospects of an Information Civilization, 30 J. Info. Tech. 75, 75 (2015) (describing a “new form of information capitalism [that] aims to predict and modify human behavior as a means to produce revenue and market control”).
  7. See, e.g., Rachel Courtland, The Bias Detectives, 558 Nature 357, 357 (2018) (documenting concerns among the public that algorithmic risk scores for detecting child abuse fail to account for an “effort . . . to turn [a] life around”).
  8. Reuben Binns et al., ‘It’s Reducing a Human Being to a Percentage’; Perceptions of Justice in Algorithmic Decisions, 2018 CHI Conf. on Hum. Factors Computing Systems 9 (emphasis omitted).
  9. Virginia Eubanks, Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor 168 (2017).
  10. Will Knight, The Dark Secret at the Heart of AI, MIT Tech. Rev. (Apr. 11, 2017), https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/ [https://perma.cc/L94L-LYTJ] (“The computers that run those services have programmed themselves, and they have done it in ways we cannot understand. Even the engineers who build these apps cannot fully explain their behavior.”).
  11. For consideration of these issues, see Mariano-Florentino Cuéllar & Aziz Z. Huq, Economies of Surveillance, 133 Harv. L. Rev. 1280 (2020), and Mariano-Florentino Cuéllar & Aziz Z. Huq, Privacy’s Political Economy and the State of Machine Learning: An Essay in Honor of Stephen J. Schulhofer, N.Y.U. Ann. Surv. Am. L. (forthcoming 2020).
  12. See, e.g., Daniel Kreiss & Shannon C. McGregor, Technology Firms Shape Political Communication: The Work of Microsoft, Facebook, Twitter, and Google with Campaigns During the 2016 U.S. Presidential Cycle, 35 Pol. Comm. 155, 156–57 (2018) (describing the role of technology firms in shaping campaigns).
  13. For what has become the standard view, see Larry Elliott, Robots Will Take Our Jobs. We’d Better Plan Now, Before It’s Too Late, Guardian (Feb. 1, 2018, 1:00 AM), https://www.theguardian.com/commentisfree/2018/feb/01/robots-take-our-jobs-amazon-go-seattle [https://perma.cc/2CFP-3JJV]. For a more nuanced account, see Martin Ford, Rise of the Robots: Technology and the Threat of a Jobless Future 282–83 (2015).
  14. See infra text accompanying notes 70–73.
  15. ‘When It Matters Most, Insist on a Registered Nurse,’ Nat’l Nurses United, https://www.­nationalnursesunited.org/insist-registered-nurse [https://perma.cc/MB66-XTXW] (last visited Jan. 19, 2020).
  16. Accenture Consulting, 2018 Consumer Survey on Digital Health: US Results 9 (2018), https://www.accenture.com/_acnmedia/PDF-71/Accenture-Health-2018-Consumer-Survey-Digital-Health.pdf#zoom=50 [https://perma.cc/TU5F-9J82].
  17. Quentin L. Kopp, Replacing Judges with Computers Is Risky, Harv. L. Rev. Blog
    (Feb. 20, 2018), https://blog.harvardlawreview.org/replacing-judges-with-computers-is-risky/ [https://perma.cc/WS5S-ARVF]. On the current state of affairs, see California Set to Greatly Expand Controversial Pretrial Risk Assessments, Filter (Aug. 7, 2019), https://filtermag.org/­california-slated-to-greatly-expand-controversial-pretrial-risk-assessments/ [https://perma.cc­/2FNX-U3C9].
  18. Alexis C. Madrigal, How a Feel-Good AI Story Went Wrong in Flint, Atlantic (Jan. 3, 2019), https://www.theatlantic.com/technology/archive/2019/01/how-machine-learning-fou­nd-flints-lead-pipes/578692/ [https://perma.cc/V8VA-F22W].
  19. Cathy O’Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy 152–53 (2016).
  20. Id. at 153.
  21. Regulation 2016/679, of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data, and Repealing Directive 95/46/EC (General Data Protection Regulation), 2016 O.J. (L 119) (EU) [hereinafter GDPR]; see also Christina Tikkinen-Piri, Anna Rohunen & Jouni Markkula, EU General Data Protection Regulation: Changes and Implications for Personal Data Collecting Companies, 34 Computer L. & Security Rev. 134, 134–35 (2018) (documenting the enactment process of the GDPR).
  22. See Directive 95/46, of the European Parliament and of the Council of 24 October 1995 on the Protection of Individuals with Regard to the Processing of Personal Data and on the Free Movement of Such Data, art. 1, 1995 O.J. (L 281) (EC) [hereinafter Directive 95/46].
  23. Bryce Goodman & Seth Flaxman, European Union Regulations on Algorithmic Decision Making and a “Right to Explanation,” AI Mag., Fall 2017, at 51–52 (explaining the difference between a non-binding directive and a legally binding regulation under European law).
  24. Id. at 52.
  25. GDPR, supra note 20, arts. 4(1), 22(1) (inter alia, defining “data subject”).
  26. See id. art. 4(7)–(8) (defining “controller” and “processor” as key scope terms). The Regulation, however, does not apply to criminal and security investigations. Id. art. 2(2)(d).
  27. As I explain below, this is not the only provision of the GDPR that can be interpreted to create a right to a human decision. See infra text accompanying notes 53–58.
  28. GDPR, supra note 20, art. 3.
  29. There is sharp divergence in the scholarship over the GDPR’s extraterritorial scope, which ranges from the measured, see Griffin Drake, Note, Navigating the Atlantic: Understanding EU Data Privacy Compliance Amidst a Sea of Uncertainty, 91 S. Cal. L. Rev. 163, 166 (2017) (documenting new legal risks to American companies pursuant to the GDPR), to the alarmist, see Mira Burri, The Governance of Data and Data Flows in Trade Agreements: The Pitfalls of Legal Adaptation, 51 U.C. Davis L. Rev. 65, 92 (2017) (“The GDPR is, in many senses, excessively burdensome and with sizeable extraterritorial effects.”).
  30. State v. Loomis, 881 N.W.2d 749, 760 (Wis. 2016).
  31. See, e.g., Julia Angwin, Jeff Larson, Surya Mattu & Lauren Kirchner, Machine Bias: There’s Software Used Across the Country to Predict Future Criminals. And It’s Biased Against Blacks, ProPublica 2 (May 23, 2016), https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing [https://perma.cc/Q9ZU-VY6J] (criticizing machine-learning instruments in the criminal justice context).
  32. See, e.g., Apprendi v. New Jersey, 530 U.S. 466, 477 (2000) (explaining that the Fifth and Sixth Amendments “indisputably entitle a criminal defendant to a jury determination that [he] is guilty of every element of the crime with which he is charged, beyond a reasonable doubt” (alteration in original) (internal quotation marks omitted) (quoting United States v. Gaudin, 515 U.S. 506, 510 (1995))).
  33. See infra text accompanying note 88 (defining machine learning). I am not alone in this focus. Legal scholars are paying increasing attention to new algorithmic technologies. For leading examples, see Kate Crawford & Jason Schultz, Big Data and Due Process: Toward a Framework to Redress Predictive Privacy Harms, 55 B.C. L. Rev. 93, 109 (2014) (arguing for “procedural data due process [to] regulate the fairness of Big Data’s analytical processes with regard to how they use personal data (or metadata . . . )”); Andrew Guthrie Ferguson, Big Data and Predictive Reasonable Suspicion, 163 U. Pa. L. Rev. 327, 383–84 (2015) (discussing the possible use of algorithmic prediction in determining “reasonable suspicion” in criminal law); Kroll et al., supra note 2, at 636–37; Michael L. Rich, Machine Learning, Automated Suspicion Algorithms, and the Fourth Amendment, 164 U. Pa. L. Rev. 871, 929 (2016) (developing a “framework” for integrating machine-learning technologies into Fourth Amendment analysis).
  34. A forthcoming companion piece develops a more detailed account of how this right would be vindicated in practice through a mix of litigation and regulation. See Aziz Z. Huq, Constitutional Rights in the Machine Learning State, 105 Cornell L. Rev. (forthcoming 2020).
  35. For a catalog, see Meredith Whittaker et al., AI Now Inst., AI Now Report 2018, at 18–22 (2018), https://ainowinstitute.org/AI_Now_2018_Report.pdf [https://perma.cc/2BCG-M4­54].
  36. See, e.g., David Segal, The Dirty Little Secrets of Search: Why One Retailer Kept Popping Up as No. 1, N.Y. Times, Feb. 13, 2011, at BU1.
  37. See Falguni Desai, The Age of Artificial Intelligence in Fintech, Forbes (June 30, 2016, 10:42 PM), http://www.forbes.com/sites/falgunidesai/2016/06/30/the-age-of-artificial-intelli­gence-in-fintech [https://perma.cc/DG8N-8NVS] (describing how fintech firms use artificial intelligence to improve investment strategies and analyze consumer financial activity).
  38. See, e.g., Benjamin Letham, Cynthia Rudin, Tyler H. McCormick & David Madigan, Interpretable Classifiers Using Rules and Bayesian Analysis: Building a Better Stroke Prediction Model, 9 Annals Applied Stat. 1350, 1350 (2015).
  39. See, e.g., Paul Raccuglia et al., Machine-Learning-Assisted Materials Discovery Using Failed Experiments, 533 Nature 73, 73 (2016) (identifying new vanadium compounds).
  40. Yann LeCun et al., Deep Learning, 521 Nature 436, 438–41 (2015).
  41. See infra text accompanying notes 117–21 (describing state uses of machine learning).

Standing for the Structural Constitution

Who speaks in federal court for the structural principles of the federal Constitution? Under familiar practice—endorsed by the Supreme Court in its 2011 decision Bond v. United States—it is not solely the institutions empowered directly by federalism or the separation of powers but also individual litigants who can raise structural constitutional objections. Such individual standing for the structural constitution is unusual because, in effect, it enables a species of third-party standing elsewhere condemned by the Court.  This Article analyzes individual standing for the structural constitution from both doctrinal and political economy perspectives. Such individual standing, I contend, conflicts with Article III’s larger ambition to exclude from federal court those controversies with excessive externalities. Consideration of structural litigation’s political economy further shows that enlarging the pool of litigants by allowing individual as well as institutions to sue is unlikely to yield closer conformity to the Constitution given interest group dynamics. As an alternative to the current regime, the Article specifies a straitened regime of narrow justiciability that is more harmonious with Article III goals and more likely to secure fidelity to the structural constitution.