Powerpoint

Amit Pundik* AGAINST RACIAL PROFILING†

A police officer sees a suspicious bulge in the pocket of a passing pedestrian and de- liberates whether to stop and search. The pedestrian is also a young, black man, and from past searches and convictions, the police arguably know that such men are much likelier than other people to carry an illegal firearm. Should the police offi- cer be instructed to take this information into account? This article objects to racial profiling because it relies on the following type of inference: from the individual’s membership of a certain racial group, the searcher is invited to infer that the indi- vidual is likelier to exhibit some culpable behaviour. The article shows that such an inference to culpable behaviour requires contradictory presuppositions about the free- dom of the suspected behaviour. On the one hand, racial profiling ought to presup- pose that the individual suspect’s behaviour is unfree because the inference it involves takes the suspect’s behaviour to be determined by his race, age, and gender, none of which is within his control. On the other hand, similarly to criminal trials, search practices ought also to presuppose the exact opposite: that the individual is free to determine his own behaviour. If the suspected behaviour is free, the involved inference to culpable behaviour is not probative of the individual suspect’s beha- viour, so profiling methods which rely on it are useless. And if the suspected beha- viour is unfree, the inference is probative, but the suspect is not culpable and should thus not be put to trial, whatever the profiled search yields.

Keywords: freedom, causation, policing, searches, criminal process

A police officer sees a suspicious bulge in the pocket of a passing pedes- trian and deliberates whether to stop and search. While the bulge is not necessarily proof of carrying an illegal firearm, it is certainly a legitimate indication. The pedestrian is also a young, black man, and from past searches and convictions, the police arguably know that such men are much likelier than other people to carry an illegal firearm. Should the police officer be instructed to take this information into account? The issue is not limited to policing, as some Muslims who are regularly

* Lecturer, Buchmann Faculty of Law, Tel Aviv University, Tel Aviv, Israel † I would like to thank the University of Toronto Law Journal, the anonymous referees, and the editor, David Dyzenhaus, for helping me improve the quality of this article. I am deeply grateful to Ronen Avraham, Yitzhak Benbaji, Antje du Bois-Pedain, François du Bois, Ha- noch Dagan, Benjamin Eidelson, Ariel Porat, Guy Sela, Victor Tadros, and Charlie Webb for their constructive and helpful comments. I also benefited from engaging discussions with participants of workshops at Oxford, Cambridge, St Andrews, and Tel Aviv. The research leading to these results has received funding from the European Community’s Se- venth Framework Programme (FP7/2007–2013) under Grant Agreement 299653.

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

selected for an ‘enhanced security check’ at airports could testify.1 Nor is it necessarily limited to searching for weapons and bombs; if, one day, the tax authority notices that Jews are much likelier to cheat on their tax returns, should more Jews be selected for random investigation? Search practices which are based on racial profiling are deeply contro-

versial. On the one hand, racial profiling has become immensely popu- lar in some police and security forces both inside and outside the United States,2 and this popularity is not without academic support.3 Racial pro- filing is often advocated as a means of maximizing the effectiveness of scarce resources in controlling crime and preventing terrorist attacks.4

On the other hand, some criticize the reliability of profiling methods, drawing attention to difficulties involved in collecting, analyzing, and presenting the data underlying such search practices.5 Others emphasize that racial profiling is offensive even if statistically reliable,6 adding that profiling methods stigmatize not only the profiled individual but also the

1 Kevin R Johnson, ‘Racial Profiling after September 11: The Department of Justice’s 2003 Guidelines’ (2004) 50:1 Loy L Rev 67 at 78; Jack Glaser, Suspect Race: Causes and Consequences of Racial Profiling (Oxford: Oxford University Press, 2014) at 151; Leda Blackwood, ‘Policing Airport Spaces: The Muslim Experience of Scrutiny’ (2015) 9:3 Policing: A Journal of Policy and Practice 255 at 255.

2 The cross-border popularity of these measures has led the United Nations to call for the elimination of ‘racial profiling.’ See Measures to Combat Contemporary Forms of Rac- ism and Racial Discrimination, Xenophobia and Related Intolerance, GA Res 56/267, UN Doc A/56/49 (2002) 31 at para 21.

3 See, most notably, Michael Levin, ‘Responses to Race Differences in Crime’ (1992) 23:1 Journal of Social Philosophy 5. See also Risse and Zeckhauser, who argue that, with some qualifications, ‘the utilitarian argument . . . supports police and security measures that make race a consideration in deciding whom to stop, search, or investi- gate’ and that ‘the use of race in police tactics is neither unfair nor does it violate any moral right.’ Mathias Risse & Richard Zeckhauser, ‘Racial Profiling’ (2004) 32:2 Phi- losophy and Public Affairs 131 at 133 [Risse & Zeckhauser, ‘Racial Profiling’].

4 Nicola Persico, ‘Racial Profiling, Fairness and Effectiveness of Policing’ (2002) 92:5 American Economics Review 1472 at 1472–3. See also the various views described in Samuel R Gross & Debra Livingston, ‘Racial Profiling under Attack’ (2002) 102:5 Colum L Rev 1413.

5 Michael R Smith & Geoffrey P Alpert, ‘Searching for Direction: Courts, Social Science, and the Adjudication of Racial Profiling Claims’ (2002) 19:4 Justice Quarterly 673 at 678 and the sources there; Steven N Durlauf, ‘Racial Profiling as a Public Policy Ques- tion: Efficiency, Equity, and Ambiguity’ (2005) 95:2 American Economics Review 132 at 132–4; Michal Tamir, ‘Public Law as a Whole and Normative Duality: Reclaiming Administrative Insights in Enforcement Review’ (2006) 12:1 Texas Journal on Civil Liberties and Civil Rights 43 at 95.

6 David A Harris, Profiles in Injustice: Why Racial Profiling Cannot Work, revised ed (New York: New Press, 2003).

176 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

entire profiled group, most of which is law-abiding.7 Finally, some have argued that such search practices are counter-effective.8

This article seeks to expose an additional problem with racial profiling, arising from its reliance on the following type of inference: from the indivi- dual’s membership of a certain racial group, the searcher is invited to infer that the individual is likelier to exhibit some culpable behaviour (hereinafter referred to as ‘inference to culpable behaviour’). The article seeks to show that inference to culpable behaviour requires contradictory presuppositions. Unlike existing objections, the argument thus focuses on the epistemic qualities of racial profiling rather than on its social costs. Racial profiling is unsuitable for policing because it is irrational for the criminal justice system to deploy profiling in order to detect a person who engages in criminal activity and then put that person to trial.9 Neverthe- less, since this irrationality has substantial costs under any theory of social costs (for example, by affecting the system’s effectiveness, legitimacy, or public perception), the article is also relevant to any cost-benefit analysis of racial profiling. The crux of the argument is that racial profiling requires contradictory

presuppositions about the freedom of the suspected behaviour. To iden- tify the contradictory presuppositions, the article builds on a general approach that I developed elsewhere,10 according to which legal fact- finding in criminal trials is connected to the issue of free will. This article

7 William M Carter, Jr, ‘A Thirteenth Amendment Framework for Combating Racial Profiling’ (2004) 39:1 Harv CR-CLL Rev 17 at 24–7; Honorable Phyllis W Beck & Patri- cia A Daly, ‘State Constitutional Analysis of Pretext Stops: Racial Profiling and Public Policy Concerns’ (1999) 72:3 Temp L Rev 597 at 617–18.

8 Harcourt, e.g., explains that the overall crime rate within an entire society might increase as a result of profiling, even when profiling succeeds in reducing the crime rate within the profiled group. Bernard E Harcourt, ‘Rethinking Racial Profiling: A Critique of the Economics, Civil Liberties, and Constitutional Literature, and of Crimi- nal Profiling More Generally’ (2004) 71:4 U Chicago L Rev 1275. Another example is Schauer, who voices efficiency and moral concerns about ‘the overuse of race and eth- nicity and the consequent underuse of other relevant factors.’ Frederick Schauer, Pro- files, Probabilities, and Stereotypes (Cambridge, MA: Belknap Press, 2003) at 187 [Schauer, Profiles, Probabilities].

9 Even those who hold that ‘we owe nobody . . . an across-the-board duty to be rational’ are likely to accept that public authorities like the police have a duty to avoid irrational inferences when deciding how to treat suspects. See John Gardner, ‘On the Ground of Her Sex(uality)’ (1998) 18:1 Oxford J Legal Stud 167 at 168. A similar obligation is likely to apply even when the security forces are privatized. See text accompanying note 47 below.

10 Amit Pundik, ‘Freedom and Generalisation,’ online: (2016) Oxford J Legal Stud gqw016 <https://doi.org/10.1093/ojls/gqw016> [Pundik, ‘Freedom and Generalisation’]. For a detailed criticism, see Federico Picinali, ‘Generalisations, Causal Relationships, and Moral Responsibility’ (2016) 20:2 International Journal of Evidence and Proof 121.

AGAINST RACIAL PROFILING 177

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

extends the approach by arguing that racial profiling requires contradict- ing presuppositions regarding the freedom of the suspect’s behaviour. On the one hand, racial profiling and the inference to culpable behaviour it involves take the suspect’s behaviour to be determined by his race, age, and gender, none of which is within his control. Racial profiling thus pre- supposes that the suspect’s behaviour is unfree.11 On the other hand, simi- larly to criminal trials, search practices ought to presuppose the exact opposite – namely, that the individual is free to determine his own beha- viour. If the suspected behaviour is in fact free, the involved inference to culpable behaviour is not probative of the individual suspect’s behaviour, rendering profiling methods which rely on it useless. Alternatively, if the suspected behaviour is unfree, the inference is probative, but the suspect is not culpable and should thus not be put to trial as long as criminal trials seek to avoid punishing those who are not culpable. The article is structured in the following way. The second Part ex-

plains why inferring from the suspect’s (racial) group that he is likelier to exhibit some behaviour requires a causal generalization, according to which the property shared by the group members determines their be- haviour.12 The third Part identifies the theories of free will under which behaviour which is so determined is not free and shows that inferring culpable behaviour from group membership is contradictory even when the generalization used is probabilistic. The fourth Part shows that search practices that are aimed at bringing offenders to justice require presup- posing that the individual’s suspected behaviour is culpable and thus free. Deploying racial profiling in such practices and then indicting and convicting those who are caught using these methods is therefore contra- dictory. The final Part identifies the types of search to which the argu- ment of this article applies and also explains why common non-profiled types of search are not subject to the same objection.

11 As for profiling based on variables which are within the suspect’s control, see infra text accompanying note 62.

12 The second and third Parts of this article rehearse the argument I made in Pundik, ‘Freedom and Generalisation,’ supra note 10, and apply them to the issue of racial profiling. Given the complexity of the issues involved (causation, free will, and so on), I chose to repeat the argument itself in full and yet to remove some of the more nuanced qualifications. Readers who are not familiar with this article and are left with some concerns about the claims made here might find replies in there, and readers who are already familiar with the argument might want to skip to the text beginning after note 38 below.

178 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

I Why racial profiling presupposes causation

Inferences from a known to an unknown empirical fact involve a gener- alization about types.13 Even eyewitness testimony still requires the use of generalization. In Twelve Angry Men, a 1957 film that depicts a court- room drama, an old lady who testified to seeing the accused from her bedroom window is discredited once it is discovered that at the relevant time she was not wearing her glasses. Discrediting her testimony relies on a generalization that short-sighted people are not very credible eye- witnesses even when they are sure of their identification. From what is known of short-sighted people in general, we can infer that her specific testimony is not to be trusted. Furthermore, drawing inferences on the credibility of any eyewitness inevitably relies on some generalizations of how credible such identifications are under similar conditions.14 For instance, the identification accuracy of the eyewitness is significantly re- duced when the event includes violence because attention is diverted from the visual characteristics of the actors to the violent acts; once again, assessing the testimony of an eyewitness to a violent event requires taking into account a generalization of how the violence affects the eye- witness’s credibility. It therefore seems that the use of any evidence in fact-finding requires the use of some generalization.15 As Peter Tillers summarizes this point, it is impossible to avoid using generalizations in fact-finding because doing so would mean that ‘the only evidence that is legitimate is only the particular matter or act – the “specific” matter or act – that stands in question. It is not helpful to be told that the only valid evi- dence of what happened is what happened.’16

In some cases, the reference to the generalization is made explic- itly. For example, inferring that Socrates is mortal from our knowledge that human beings are mortal refers explicitly to a generalization about human beings as a type. However, in many cases, the generalization is implicit in the inference. Consider, for example, the inference from the fact that a certain person reacted allergically to a certain cat to the fact that this individual is likely to react allergically to that cat in future. This

13 See e.g. Schauer, Profiles, Probabilities, supra note 8 at 101, who holds that ‘the avoid- ance of generalizations is, with few or no qualifications, simply not possible at all.’

14 For a summary of the troubling findings on the reliability of eyewitnesses, see Paul Ro- berts & Adrian Zuckerman, Criminal Evidence, 2d ed (New York: Oxford University Press, 2010) at 297–301.

15 William Twining, ‘Narrative and Generalizations in Argumentation about Questions of Fact’ (1999) 40:2 S Tex L Rev 351 at 356.

16 Peter Tillers, ‘If Wishes Were Horses: Discursive Comments on Attempts to Prevent In- dividuals from Being Unfairly Burdened by Their Reference Classes’ (2005) 4:1–2 Law, Probability and Risk 33 at 44 (emphasis added).

AGAINST RACIAL PROFILING 179

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

inference is presumably based on our knowledge of an allergy as a condi- tion which appears persistently rather than erratically. This knowledge implies one or more generalizations which could serve as the basis for that inference (though making this inference does not require identify- ing in advance what the implicit generalization is). For example, the gen- eralization implied by the inference could be that the type of people who reacted allergically to cats (or, alternatively, to a specific cat) is likely to continue to react allergically. The important point is that drawing an inference from one empirical fact to another presupposes that there is at least one generalization about types of fact that somehow connects the fact from which the inference begins and the fact with which the infer- ence ends. Without such a presupposition, the inference is invalid be- cause it remains unclear what licenses the move from the first fact to the second. This section argues that inferences drawn from (racial) group member-

ship to culpable behaviour require a causal generalization – namely, a gen- eralization that reflects a causal connection between the type of fact from which the inference begins and the type of fact which the inference seeks to establish. If an inference is based on a non-causal generalization – a mere correlation – it is arguably unlicensed and thus invalid.17 The causal relation can be either direct or through a common cause. Inferring that a smoker is likelier to contract cancer than a non-smoker is based on a causal generalization that smoking is a cause of (lung) cancer. By contrast, inferring that a Coca-Cola drinker is more likely to contract (skin) cancer than a non-drinker involves a causal generalization that reflects a common cause. It is living in a hot country which is the common cause of both Coca-Cola drinking and (skin) cancer. Furthermore, for the inference to be valid, it is not necessary to refer to a specified (direct or indirect) causal generalization. The inference is valid so long as the existence of such a causal generalization is presupposed. That valid inferences must presuppose causal generalizations becomes

clear when considering the opposite stance, according to which some in- ferences could be based on mere correlation. Such a stance would still require that the generalization on which a valid inference is based satis- fies certain conditions or standards, such as statistical significance. It would insist, however, that a correlation between two types of fact can be used to infer an unknown from a known fact, even if there is no causal connection between the types of fact, not even indirectly.

17 This claim is part of the Common Cause Principle. See Hans Reichenbach, The Direc- tion of Time, 2d ed (Berkeley, CA: University of California Press, 1991) at 157–60; Frank Arntzenius, ‘The Common Cause Principle’ (1992) 2 Proceedings of the Bien- nial Meeting of the Philosophy of Science Association 227.

180 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

The difficulty with such a stance is that it renders the rejection of spu- rious correlations more difficult. Spurious correlations are those which do not reflect any actual connection (either causal or not) between the two types of fact, such as the seemingly perfect correlation between the divorce rate in Maine and the per capita consumption of margarine in the United States.18 The lack of any actual connection between these facts entails that this spurious correlation does not hold outside the group of initially observed cases. Hence, it would be a mistake to infer anything about the consumption of margarine from the divorce rate (or vice versa) in a year which is not included in the group of years within which the spurious correlation was identified. Drawing any inference from a spurious correlation to an unobserved case is therefore unli- censed and misleading, whatever the purpose of the inquiry is (be it ob- taining knowledge, providing explanation, making prediction about unobserved cases, and so on). Ensuring that a given correlation is not spurious is important because, unfortunately, spurious correlations seem to be everywhere, as the real-life example above illustrates. Moreover, spurious correlations ought to be present. Since each specific case con- sists of numerous details (most of them, of course, are unimportant), one could go through a vast number of facts until one finds a group in which the identified fact correlates with the fact that one seeks to estab- lish. For example, one might find a correlation between certain human behaviour and the second (or third) letter of the surname of the per- son’s grand aunt. If one accepts that inferences require causal generalizations, one

could apply methods to distinguish between a causal and non-causal con- nection (whatever these methods are)19 in order to identify which gener- alizations are spurious. However, if one denies this, one ought to find how to distinguish between informative and spurious correlations. Note that mere statistical significance will not do because enumerating suffi- ciently large numbers of variables using sufficiently large databases would eventually generate statistically significant, yet spurious, generali- zations. One might hope that such absurd statistically significant correla- tions simply do not exist. Yet this hope relies on the belief that statistically significant correlations need to ‘make sense’ – that is, that it would be possible to explain why this correlation holds, and what would such an explanation be if not causal or causal-like?

18 Tyler Vigen, Spurious Correlations, online: <http://www.tylervigen.com/spurious- correlations>.

19 Various sophisticated methods have been proposed, such as the Markov condition and Bayesian nets. For a detailed introduction, see Jon Williamson, Bayesian Nets and Causality (Oxford: Oxford University Press, 2005).

AGAINST RACIAL PROFILING 181

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

One might challenge this argument using counter-examples in which an inference from a known to an unknown fact is made without presup- posing a causal connection between the types of fact. For example, if there are ten balls in a jar, of which nine are blue, it might be possible to infer that the probability of a randomly chosen ball’s being blue is 90 per cent without presupposing any causal connection between ‘being in that jar’ and ‘being blue.’ However, even if not all factual inferences require presupposing a

causal connection, the inferences drawn in racial profiling almost always do. Denying an underlying causal connection is easier when the general- ization is extracted from a group of cases to which the instant case at hand belongs. It is important to note that the randomly chosen ball is itself one of the ten balls in the jar. By contrast, generalizations used in racial profiling are almost always extracted from a group of cases which does not include the instant case. For example, the rate of carrying illegal firearms among young black men is inferred from a group of cases in which it was known that the individuals carried illegal firearms. The sus- pect is clearly not a member of this group because the question is whether this particular young black man is carrying an illegal firearm and data on the carrying rate among other young black men are brought to support the conclusion that he is likelier to do so. This difference is important because, if the case at hand is included in

the group of known cases on which the generalization is based, it might be possible to draw some inferences about it without presupposing any- thing about the relation between the types of fact. While such inferences raise a set of difficult problems,20 these differ in kind from those in- volved in drawing inferences from generalizations which do not include the case at hand.21 To infer the probability that a randomly chosen ball is blue from the proportion of blue balls in another jar, it is necessary to presuppose that there is some substantial relation between ‘being in a jar’ and ‘being blue.’ Similarly, inferring that it is likelier that a specific young black man is carrying an illegal firearm from past arrest and con- viction rates among the racial group to which he belongs requires pre- supposing that there is some substantial relation between ‘being a young black man’ and ‘carrying an illegal firearm.’ And, again, if this substan- tial relation is not causal or causal-like, what else could it be?

20 See e.g. Alan Hájek, ‘Fifteen Arguments against Hypothetical Frequentism’ (2009) 70:2 Erkenntnis 211.

21 This issue is related, of course, to the problem of induction, both in David Hume’s original version and particularly in Goodman’s newer version. See Nelson Goodman, Fact, Fiction, and Forecast, 4th ed (Cambridge, MA: Harvard University Press, 1983).

182 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

II Why inferring culpable behaviour from group membership is contradictory

A THE METAPHYSICS OF FREE WILL

One of the main problems regarding free will is the ancient question of whether free will is compatible with determinism: can anyone be free if every given event, including human behaviour, is fully determined by the preceding state of affairs and the laws of nature?22 Understanding the world around us usually involves searching for causes, and this search tends to presuppose that every event is determined by applying causal laws of nature to its antecedent conditions (even if not all of these laws are known, the search for causes presupposes their existence). On the other hand, if human behaviour too is determined by the past, in what sense is it any freer than that of inanimate objects? As Benedict de Spi- noza provocatively suggests, if a falling stone had consciousness, it ‘would believe itself to be completely free, and would think that it contin- ued in motion solely because of its own wish.’23

One kind of response seeks to show that, despite the apparent tension, determinism is compatible with freedom. A simple version of compatibi- lism holds that a person is free whenever she would have done otherwise had she chosen to. Consider, for example, a man who dines in a restaurant and deliberates between ice cream and fruit salad for dessert. His order- ing ice cream may be free even if this choice was determined by his genetic composition, upbringing, and so on because he would have ordered the fruit salad had he chosen to. By contrast, a person who trips over something is not free to stop falling because he would not remain standing even had he chosen to. While there are various compatibilist strategies to explain how freedom is compatible with determinism,24

they all share the view that being determined by causal antecedent fac- tors does not, in and of itself, render human behaviour unfree. By contrast, many people believe that if human behaviour is fully

determined by causal antecedent factors, then it cannot be free.25 In the previous example, if the diner’s choice of ice cream were determined by his genetics or upbringing, then he was not really free to choose the fruit

22 A detailed introduction can be found in Robert Kane, A Contemporary Introduction to Free Will (Oxford: Oxford University Press, 2005) [Kane, Contemporary Introduction].

23 Benedict de Spinoza, ‘Letter LXII (LVIII)’ in MW Dunne, ed, Improvement of the Under- standing: Ethics and Correspondence of Benedict de Spinoza, trans RHM Elwes (New York: Dover Publication, 1901) 395 at 396 (letter sent from Benedict de Spinoza to GH Schaller, October 1674).

24 Kane, Contemporary Introduction, supra note 22, ch 2, 7. 25 But see Eddy Nahmias, Jason Shepard, & Shane Reuter, ‘It’s OK If “My Brain Made

Me Do It”: People’s Intuitions about Free Will and Neuroscientific Prediction’ (2014) 133 Cognition 502 at 503.

AGAINST RACIAL PROFILING 183

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

salad because he is no different from the tripping person. The libertar- ian response embraces this intuition and holds that human behaviour is not always determined by the laws of nature, so determinism must be false.26 Human behaviour can thus be free, and sometimes is, but only when it is not determined by antecedent causal factors; not even by the agent’s genetic composition or upbringing. Furthermore, the agent’s be- haviour must not only be undetermined by antecedent factors, but it must also be under the agent’s control; otherwise it is merely a random event. While compatibilists and libertarians disagree on when one is free,

they tend to agree that one can be culpable only if one acted freely.27

Compatibilists would hold that one can be culpable even if one’s beha- viour was determined by causal antecedent factors, because these factors do not necessarily render one’s behaviour unfree. By contrast, libertar- ians hold that if one’s behaviour were determined by causal antecedent factors, one would be neither free to do, nor culpable for doing, what one did.

B INFERENCE TO CULPABLE BEHAVIOUR UNDER LIBERTARIAN APPROACHES

This article seeks to show that the inference to culpable behaviour drawn in racial profiling requires presupposing a causal factor which renders the suspect’s behaviour unfree. This presupposition is inconsis- tent with the presupposition of free action required for culpability, thereby rendering the inference inconsistent and thus irrational. This subsection focuses on libertarian approaches, under which the applica- tion of the argument is relatively straightforward, and the next subsec- tion outlines when and how a similar argument can be made under compatibilist approaches. Under the libertarian approaches, the argu- ment yields an objection whenever inferring culpable behaviour presup- poses the existence of some causal factor outside the agent’s control. For libertarians, the very existence of such a causal factor suffices to render the behaviour unfree, and this remains true even if the causal factor is unspecified or unknown. Many causal generalizations are probabilistic. For example, the gener-

alization that smoking increases the risk of lung cancer tells us that it is more likely that a smoker will contract cancer than a non-smoker. It does not tell us that every smoker will contract cancer. By contrast, some

26 Randolph Clarke, Libertarian Accounts of Free Will (New York: Oxford University Press, 2003).

27 The argument of this article applies also to theories which hold that free will is not a necessary condition for the attribution of culpability (for example, semi-compatibilist and reactive attitudes theories). See text accompanying note 52 below.

184 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

generalizations seem to be deterministic. For example, the generaliza- tion that human beings who are denied oxygen die can be used to infer with certainty that the victim died if it is known that he had been denied oxygen. While racial profiling tends to involve mainly probabilistic gen- eralizations, they add another layer of complexity. The argument is thus first applied to deterministic generalizations and then, more realistically, to probabilistic generalizations. Starting with a simple example, assume that Richard who is exposed

to a certain radiation develops certain skin marks and an irresistible urge to attack everyone around him. Assume further that every person exposed to this radiation develops these symptoms. When Richard is admitted to hospital, it seems unproblematic to infer from the skin marks that, given the opportunity, he will go berserk and, therefore, should be restrained. However, inferring from these marks that a violent act which had taken place before Richard arrived at the hospital was committed by him (rather than by someone else) for the purpose of con- victing him of violent offences seems problematic. Using this inference for the purpose of conviction is problematic because

it leads to a contradiction. To infer from Richard’s skin marks that he had acted violently, it is necessary to presuppose a causal generalization – either one caused the other, or they both have a common cause. In this example, the radiation caused both Richard’s skin marks and violent behaviour. However, Richard’s acting violently may be culpable only if he acted freely. According to libertarians, people do not act freely when their behaviour is determined by antecedent conditions outside their control. Establishing Richard’s guilt by inferring from his skin marks that it was he who acted vio- lently is hence contradictory; Richard’s behaviour is treated as free and un- free at the same time. Blaming Richard for acting violently, having inferred from the skin

marks that it was he who acted violently, is problematic; such an infer- ence cannot be used without dissolving the person’s culpability. It also explains why the very same inference seems unproblematic in the con- text of medical treatment. While inferring from the skin marks that Ri- chard will act violently in the hospital presupposes that his behaviour is determined and, hence, unfree, this leads to no contradiction. In the medical context, it is not necessary to presuppose that Richard’s violent behaviour will be free. Moving to probabilistic generalizations, there is a common view that

even if being fully determined by causal factors renders the action un- free, the same is not true when the causal influence is only partial. Ac- cording to this view, for example, a person’s eating a certain cake could be free even if his genetic composition causally influenced, but did not fully determine, his eating (for example, by causing him to have ‘a sweet

AGAINST RACIAL PROFILING 185

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

tooth’). The purpose of the following discussion is to challenge this com- mon view by showing that the same argument made above about deter- ministic causal generalizations also applies to probabilistic ones. The way the argument applies to probabilistic generalizations depends

on how the notion of probability is to be understood.28 Consider the fol- lowing probabilistic generalization: dropping a glass from a certain height onto a wooden floor will cause it to break in 50 per cent of cases. While there are various ways to understand how this generalization re- flects the underlying causal relations between dropping a glass and the glass breaking, the one considered to be the most suitable for legal pur- poses by various theorists is the subjective interpretation of probability.29

According to the subjective interpretation,30 while the generalization which describes our knowledge of the world is probabilistic, it imper- fectly reflects a reality which may be deterministic. Had we known every- thing that can possibly be known about the world, it might have been possible to infer with certainty about any specific glass whether dropping it would cause it to break. However, we do not know all the relevant facts. All that we know is that if the glass is dropped, it will break on half of such occasions. The variable (or set of variables) which determines when a glass breaks is unknown to us. This is the reason why the general- ization is probabilistic even if the underlying reality is deterministic. Probabilistic generalizations reflect the limited state of our knowledge rather than the true nature of the world. Consider the following variation on the previous example. Assume

that Stephen is exposed to another type of radiation, one which always causes certain skin marks, but causes an irresistible urge to attack others when the opportunity arises only in 80 per cent of cases. Stephen be- longs to one of two possible subgroups. One possibility is that he belongs to the subgroup of people who possess an extra unknown variable which, together with the radiation, determines that he will go berserk. The other possibility is that he belongs to the subgroup of people who

28 Donald Gillies, Philosophical Theories of Probability (London: Routledge, 2000). 29 For criminal law, see Larry Alexander & Kimberly Kessler Ferzan, with the collabora-

tion of Stephen Morse, Crime and Culpability: A Theory of Criminal Law (New York: Cam- bridge University Press, 2009) ch 2 at 31. For tort law, see Stephen Perry, ‘Risk, Harm, and Responsibility’ in David G Owen, ed, Philosophical Foundations of Tort Law (Oxford: Clarendon Press, 1995) 321 at 333–5. For health and safety regulation, see Matthew D Adler, ‘Against “Individual Risk”: A Sympathetic Critique of Risk Assessment’ (2005) 153:4 U Pa L Rev 1121 at 1247.

30 Inference to culpable behaviour is objectionable also under the objective interpreta- tion of probability. See Pundik, ‘Freedom and Generalisation,’ supra note 10, subsec- tion 3.D.

186 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

do not possess the extra variable, in which case the exposure to the radi- ation will not cause him to go berserk. If Stephen possesses the extra variable, supporting his conviction by

inferring from his skin marks that he was likely (by 80 per cent) to have acted violently is problematic. Similarly to deterministic generalizations, such an inference leads to a contradiction. His behaviour is taken to be both free (in order to be culpable) and unfree (together with another unknown variable, his violent behaviour was determined by the radia- tion). To avoid the contradiction, either the evidence of skin marks has to be accepted as probative of the violent act’s having been committed by Stephen, in which case Stephen is not culpable, or not probative, in which case it should be ignored. If Stephen does not possess the extra variable, inferring from his skin marks that he was likely (by 80 per cent) to have acted violently is mistaken and, hence, misleading. If he belongs to the subgroup of people who were not caused to act violently by the radiation, then the probability that he acted violently is the same as the probability that he would act violently even without the exposure to the radiation. Inferring from the skin marks that he is more likely to have acted violently than he would have been had he not had these marks is therefore mistaken. Inferring culpable behaviour using such probabilistic generalizations is either contradictory, because it requires inconsistent presuppositions, or misleading, because it is mistaken and yet the generalizations are presented as informative. This objection also applies to the inference used in racial profiling. As ar-

gued in the second section of this article, inferring that an individual sus- pect is likelier to carry an illegal firearm from his being a young black man requires a causal generalization which connects the type of fact ‘being a young black man’ and the type of fact ‘carrying an illegal firearm.’ The generalization is probabilistic since clearly not all young black men carry illegal firearms. To be probative, it suffices that the inference concludes that a suspect is likelier to carry an illegal firearm if he is a young black man. However, inferring that the suspect is likelier to carry an illegal fire- arm culpably requires contradictory presuppositions. Either the individual suspect belongs to the subgroup of young black men who possess an extra unknown variable that, together with their race, age, and gender, deter- mined their behaviour, thereby rendering it unfree and non-culpable, or the individual suspect belongs to the subgroup of young black men who do not possess the extra variable, thereby rendering the inference mislead- ing since the probability that the suspect is carrying an illegal firearm is the same as it would have been had he not been a young black man. Notably, the argument that the inference to culpable behaviour used in

racial profiling is objectionable since it is based on a causal generalization between ‘being a young black man’ and ‘carrying an illegal firearm’ goes

AGAINST RACIAL PROFILING 187

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

significantly further than existing arguments against racial profiling. Mathias Risse and Richard Zeckhauser, for example, accept that profiling may cause certain harm to minorities: the feeling of resentment, the sense of hurt, and the loss of trust in the police.31 However, they hold that these harms are parasitical on pre-existing social injustices suffered by these minorities. In a world without racism, ‘using race for investigative pur- poses would not be considered offensive and would not trigger resent- ment, hurt, or loss of trust in law enforcement.’32 According to their view, if profiling targets a group which has not been an historical victim of racist attitudes and practices, it will not cause these harms. By contrast, the argument of this paper applies irrespective of the

identity of the target group and its being a victim of social injustice. It shows that the inference involved in racial profiling is rational only if based on a causal relation between group membership and criminal be- haviour – namely, there is a property shared by the group members which causes them to commit crime. This has nothing to do with how the group was treated in the past, so removing other pre-existing injustices would not rectify the profiling’s irrationality. Similarly, the argument ap- plies even when racial profiling targets privileged groups. For instance, profiling white middle-aged men for tax evasion is objectionable because the inference is rational only if being a white middle-aged man causes them to commit tax offences. One might retort that this analysis stands in contrast to a common

intuitive view of criminal responsibility. While the analysis implies that the agent is either fully determined (when belonging to one subgroup) or entirely unaffected (when belonging to the other subgroup), the practices of assigning criminal responsibility often seem to assume that an agent can be partially causally influenced. The agent is treated as caus- ally influenced by some factor, but only to some degree, leaving him with a less-than-maximal extent of freedom. For example, a paedophile’s sentence might be mitigated by the fact that he was a victim of molesta- tion in his childhood. According to this view, the mitigation acknowl- edges that his childhood experience has casually influenced the way he currently acts, yet left him sufficiently free and, hence, responsible for molesting other children. The difficulty with this view of criminal responsibility is that it fails to

account for the conviction stage of the trial, which seeks a binary out- come: the accused is either guilty of the alleged crime or not. Finding him guilty requires that he is culpable of committing the crime, which,

31 Risse & Zeckhauser, ‘Racial Profiling,’ supra note 3 at 144. 32 Ibid at 146.

188 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

in turn, requires that he acts freely.33 Free action is a precondition of criminal responsibility,34 and when undermined by a defence such as insanity or duress, the accused is found not guilty rather than less guilty. One way to explain away the intuitive force of this view of criminal

responsibility is to note that, while the question of guilt is binary, the re- sponses to conviction are typically scalar. The punishment could include a longer or shorter period of imprisonment or a heftier or slighter fine. It is at the sentencing stage that the paedophile’s childhood experience is taken into consideration. However, there could be various explana- tions for why this experience serves to mitigate the appropriate punish- ment which do not refer to a partial causal influence. To mention just a few alternatives, there would be the increased effect that punishment would have on him as a result of his experience, his vulnerability to becoming a victim again during imprisonment, or maybe even the attempt to compensate him for his bad luck. Whatever the justification may be, it need not rely on a causal generaliza-

tion, according to which his childhood experience causally influenced him to commit the alleged offence. If such a causal generalization is used at the sentencing stage, it becomes difficult to explain why the prosecution should not be allowed to admit the very same evidence at the conviction stage to support its allegation that the accused has committed the offence.35

After all, if the accused’s personal background (be it his childhood expe- rience, socio-economic conditions, education, and so on) causally influ- enced him, it means that he is likelier to have committed the alleged crime, rendering his background probative evidence, which should not be ignored at the conviction stage. While exploring the justification for such mitigation lies outside the scope of this article, the important point is that taking into account the paedophile’s childhood background at the senten- cing stage need not be based on his being less free when molesting the children he did. This point is in line with the fact that freedom is taken by almost every theorist of free will as a binary concept.36

33 For theories which hold that free will is not a necessary condition for the attribution of culpability (for example, semi-compatibilist and reactive attitudes theories), see text accompanying note 52 below.

34 This precondition does not amount to presupposing retributivism since it applies to any theory of punishment which seeks to avoid punishing those who are not culpable of their (unfree) actions. See note 55 below.

35 Note that this issue differs from the question of admissibility of previous offences of paedophilia. For that issue, see Pundik, ‘Freedom and Generalisation,’ supra note 10 at n 84 and the accompanying text.

36 Nomy Arpaly seems to be a notable exception. See Nomy Arpaly, Merit, Meaning, and Human Bondage: An Essay on Free Will (Princeton, NJ: Princeton University Press, 2006).

AGAINST RACIAL PROFILING 189

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

Therefore, the above analysis of probabilistic causal generalizations does not stand in contrast to our sentencing practices. On the contrary, the view of partial causal influences stands in contrast to our binary prac- tices of conviction. Proponents of such a view thus need to explain how freedom and criminal responsibility work under their view.

C INFERENCE TO CULPABLE BEHAVIOUR UNDER COMPATIBILIST

APPROACHES

This subsection outlines how the argument could also work under some compatibilist approaches to free will. Whether a certain general- ization is objectionable under a compatibilist approach depends upon the exact presupposition which the use of the generalization requires. Compatibilists distinguish between free and unfree behaviour, but, unlike libertarians, they hold that behaviour can be free even if it was determined by causal factors outside the agent’s control. Instead, they offer alternative criteria to distinguish between free and unfree beha- viour (for example, whether the agent would have acted differently had he wanted to). Whatever the compatibilist criteria, they regard be- haviour as unfree once certain causal factors are present. For example, a person whose hand was coercively pressed against a button clearly did not press the button freely because his behaviour was caused by physical coercion. What these causal factors are and how they explain why the compatibilist criteria for freedom are unsatisfied varies ac- cording to the specific compatibilist theory. But the important point for the purpose of this article is that the argument is still applicable under a compatibilist approach. A generalization is objectionable if its use presupposes a causal factor that belongs to the group of causal fac- tors that render behaviour unfree according to the respective compati- bilist criteria. Consider again the deterministic generalization according to which

every person exposed to certain radiation develops certain skin marks and an irresistible urge to attack everyone around them. Is it objection- able to convict Richard of a violent offence by inferring from his skin marks that it was he (rather than someone else) who acted violently? Ac- cording to compatibilists, the answer depends on whether any of the facts presupposed by the inference (for example, that Richard was ex- posed to this radiation) belongs to the group of causal factors which ren- der Richard’s behaviour unfree. For example, some compatibilists might argue that the radiation rendered Richard unfree if its effect was such that Richard would not avoid acting violently even had he wanted to. Other compatibilists might adhere to Harry Frankfurt’s theory, accord- ing to which a person acts freely only if he had a (second-order) desire

190 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

to have the (first-order) desire to act as he did.37 If the radiation caused Richard to have a desire to act violently even though he did not have a second-order desire to have the desire to act violently, then being ex- posed to this radiation belongs to the causal factors which render Ri- chard’s behaviour unfree. A similar point can be made about probabilistic generalizations. Even

if the radiation causes an irresistible urge to attack others only in 80 per cent of cases, the generalization is still objectionable if the radiation be- longs to the group of causal factors which render Stephen’s behaviour unfree. The inference is either misleading or inconsistent with holding Stephen culpable for his violent actions. Compatibilists would hence need to examine the question of whether a generalization is objection- able through the lens of their specific criteria for freedom.38 Notably, the same points made above about partial causal influences and their inability to explain the binary practices of conviction would also apply under a compatibilist approach to free will. Whether being determined by one’s race, age, and gender renders

one’s behaviour unfree is a complicated question under compatibilist theories, a question on which there is surprisingly little scholarship.39

Unlike libertarians, for whom the very fact that one’s behaviour is deter- mined by one’s race, age, and gender suffices to render it unfree (because none of these is under one’s control), a compatibilist would reach a similar conclusion only if race, age, and gender somehow inter- fere with the conditions which her specific compatibilist theory identifies as necessary for freedom. As a result, answering the above question re- quires first identifying the exact causal connection between ‘being a young black man’ and ‘carrying an illegal firearm’ and then assessing whether and how it affects the freedom of young black men under a given compatibilist theory. Some compatibilists could easily reach the conclusion that carrying an

illegal firearm is unfree if it was caused by the carrier’s being a young black man. Assume, for the sake of argument, that young black men are more likely to carry an illegal firearm because they are subject to a con- stant threat of assault or to immense peer pressure. Some compatibilists might conclude that such psychological or social factors, if strong

37 Harry G Frankfurt, ‘Freedom of the Will and the Concept of a Person’ (1971) 68:1 Journal of Philosophy 5.

38 A similar analysis can be made regarding theories which hold that free will is not a necessary condition of the attribution of culpability. See Pundik, ‘Freedom and Gener- alisation,’ supra note 10 at n 67.

39 A notable exception is Levin’s work discussed below. Michael Levin, Why Race Matters: Race Differences and What They Mean (Westport, CT: Praeger, 1997) [Levin, Why Race Matters].

AGAINST RACIAL PROFILING 191

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

enough, might render the carrying unfree. For example, if they only caused the individual to have a first-order desire to carry an illegal fire- arm even though he has a second-order desire not to have such a dan- gerous desire. This example illustrates how compatibilist reasoning could reach the conclusion that these causes render his behaviour un- free, though other compatibilists could reach a similar conclusion by ex- posing a different causal connection or identifying another condition of freedom which is violated. The important point is that for compatibilists who reach this conclusion the argument still holds. Inferring that a sus- pect is more likely to culpably carry an illegal firearm from his being a young black man still requires contradictory presuppositions. Other compatibilists might reach the opposite conclusion and insist

that the carrying was free even if it was determined by the carrier’s race, age, and gender. According to these compatibilists (hereinafter referred to as ‘broad compatibilists’), the range of causal factors which are com- patible with freedom is broad enough to include, in some circumstances, race, gender, and age, and thus criminal behaviour determined by these factors may nevertheless be free and culpable. As a result, for broad com- patibilists, racial profiling is not necessarily irrational in the sense that it does not involve the contradiction highlighted in this article. A notable example (if not the only example) for a proponent of racial

profiling who explicitly adopts a broad compatibilist view is Michael Levin. According to Levin’s troubling view, racial profiling is probative because the cause of the increased crime rate among blacks is probably genetic, mediated by lower intelligence, impulsive temperament, or higher serum testosterone.40 Yet while such genetic influences make blacks more prone to crime, they do not render them unfree.41 While this view seems highly offensive,42 this article does not seek to explain

40 Ibid at 316 and 317 respectively: ‘race differences in IQ and temperament, variables which significantly affect criminal behavior, are significantly genetic in origin’ and ‘one mediating mechanism may be the race difference . . . in serum testosterone, known to facilitate aggression.’

41 Ibid at 320: ‘that we are free when we do what we choose to do, although our choices are caused by unchosen genes, preserves the freedom of individuals whose unchosen genetic aggressiveness leads them to [choose] lawbreaking.’ This is only part of Le- vin’s view since he then proceeds to argue that freedom is a matter of degree (this view is discussed in the text following note 32 above and rejected for being incompati- ble with the binary notion of conviction).

42 Levin concludes that ‘blacks have less free will than whites’ (ibid at 322). However, ‘di- minished responsibility does not imply greater leniency’ (at 323). Rather, Levin calls for harsher treatment of blacks (for example, in addition to heavier punishment, he also suggests ‘swifter administration of punishment to blacks, along with stricter limits on appeals’ (at 325)).

192 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

why it is so, nor does it require accepting that this view is indeed offen- sive.43

However, it is important to emphasize how difficult it is for broad com- patibilists to accept racial profiling as probative of criminal behaviour without slipping into an acceptance of Levin’s claims or others that are similarly troubling. One way to do so would be to retain the claim that the carrying is caused by genetic composition while replacing Levin’s offensive mediating factors (for example, lower intelligence) with others, which are supposedly less offensive. The problem is that such a claim still implies that young black men are prone to criminal activity by their nature – namely, by their genetic composition. Another way would be to replace or supplement the genetic composition with environmental fac- tors. For example, they might refer to the same psychological and social factors mentioned above (a constant threat of assault or immense peer pressure) but insist that these factors do not render the behaviour un- free according to their compatibilist theory (for example, arguably he would not have become a criminal had he wanted to). Be that as it may, such a view still implies that young black men are prone to criminal activ- ity by their nurture – namely, by the environmental factors to which they are exposed, which are not within their control.44 Nature or nurture, racial profiling is probative only if there is a causal connection between being a young black man and being a criminal. Proponents of racial pro- filing who find Levin’s view of young black men’s criminality unpalatable need to explain how racial profiling can be based on a causal relation that is both probative and not offensive.45

43 One might argue that the view that a young black man’s criminal behaviour is caused by his nature or nurture is not necessarily offensive because, in some contexts, the view that one’s nature or nurture causes one to act criminally is considered to be pro- gressive rather than offensive (for example, imposing a more lenient punishment on a drug addict on the basis that his parents’ drug consumption caused his addiction). Note, however, that in contexts such as this one, the view that his criminal behaviour was caused by his nature or nurture serves as a reason to advantage him. By contrast, in the context of racial profiling, the view that the young black man’s criminal beha- viour is caused by his nature or nurture serves as a reason to burden him. Even if advan- taging someone on the basis that his nature or nurture causes him to act criminally is progressive rather than offensive (but see the conclusion of this article), there is little progressiveness in imposing burdens on someone on that basis.

44 For profiling based on variables which are within the suspect’s control, see text accom- panying note 62 below.

45 Herein, it is only suggested that broad compatibilism is no different from Levin’s view without assessing the merits of this view. The next section argues, inter alia, that it is difficult for broad compatibilists to explain why evidence of this causal connection, which is taken by the proponents of racial profiling as sufficiently informative to assist in searching, is nevertheless inadmissible in criminal trials. See the text accompanying note 50 below.

AGAINST RACIAL PROFILING 193

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

III Why searching ought to presuppose freedom?

As shown in the second section of this article, racial profiling requires a causal generalization and, as shown in the third section, this generaliza- tion is inconsistent with various theories of free will. Using racial profil- ing in search practices thus requires presupposing that the suspected behaviour is unfree. Yet, in many contexts, there is no difficulty in pre- supposing that given behaviour is unfree. Returning to the radiation example, restraining Richard after inferring from the skin marks that he will act violently leads to no contradiction because in the medical con- text it is not necessary to presuppose that his violent behaviour will be free. The next step in the argument is thus to establish that searching ought to presuppose that the suspect acts freely. Consider a young black man put to trial for carrying an illegal firearm

where the fact that he was carrying is disputed – the weapon was found on the floor, and it is unclear whether it belonged to him or to another person, who happens to be white. If inferring that a suspect is likelier to carry an illegal firearm from his being a young black man is probative, why should not the same inference be used to support the conviction of the black accused? The question is not merely whether this inference suffices on its own to prove that the accused carried an illegal firearm but, rather, whether evidence on higher crime rates among young black men should be admissible in criminal trials, for such evidence is not used (nor should it be used) to determine the individual’s guilt to begin with. In previous work, I have argued that inference to culpable behaviour

should not be used in criminal trials because it requires contradictory presuppositions: either the evidence is probative of the case at hand, in which case the accused is unfree and thus not culpable, or the evidence is not probative, in which case the evidence is irrelevant and inadmissi- ble.46 Conviction in a criminal trial treats the accused as free to deter- mine his own behaviour, and since searching is an integral part of the criminal process, it should also be committed to the same presupposi- tion. As a result, neither the police nor the court can use contradictory presuppositions about the very same question of fact : whether the accu- sed’s behaviour was free or determined by causes outside of his control. By analogy, consider a criminal case in which the court determines that a certain material X is not toxic. Surely, the police should not continue enforcing the prohibition on possession of toxic materials as if X were

46 See Pundik, ‘Freedom and Generalisation,’ supra note 10. For the advantages of this analysis over existing attempts to explain what is wrong with using statistical evidence in court, see ibid, ss 4, 5.

194 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

toxic. While the legal implications of possessing a toxic material might differ along the criminal process, the same material should not be taken to be toxic in one part and non-toxic in another. This article argues that the same rationale applies to the question of

whether the accused acted freely in a specific event (which will be referred to as ‘the objection from the unity of process’). Either both the court and the police should treat his behaviour in the specific event as free, rendering the police’s use of racial profiling not probative, or both should treat this behaviour as determined by the suspect’s profile, ren- dering the court’s attribution of culpability unjustified. What the crimi- nal process should not do is first presuppose that the suspect’s behaviour in the specific event is unfree when searching him because of his profile, and then presuppose that it was free when indicting and convicting him. Deploying racial profiling thus renders the criminal process’s treatment of the accused irrational because the presuppositions made at each stage about the very same event are contradictory. Notably, the objection from the unity of process is applicable not only

when the searches are conducted by an organ of the state (for example, the police) but also when conducted by a private company (for exam- ple, airport security). Firstly, such a private company may still perform ‘functions of a public nature’47 and, hence, be considered a ‘hybrid’ public authority, subject to requirements similar to those imposed on the police.48 Secondly, and more fundamentally, as long as the private company participates in (the enforcement part of) the criminal process, it does not matter whether the search is conducted by a private company or a public authority. Consider, for example, a criminal case in which the court determines that a certain drug which was considered by the police to be illegal is not so. Surely, the state should not be allowed to bypass the court judgment by outsourcing the enforcement to a pri- vate company. Furthermore, any private company which takes part in en- forcing the prohibition against possessing illegal drugs (for example,

47 See Human Rights Act 1998 (UK), s 6(3)(b). 48 YL v Birmingham City Council and Others, [2007] UKHL 27 at para 63. Baroness Hale

holds that ‘it is common ground that “functions of a public nature” include the exer- cise of the regulatory or coercive powers of the state,’ specifying private prisons and private psychiatric hospitals as paradigmatic examples of bodies which perform such functions. Private security companies seem likely to belong to the same category. See Rowbottom, who argues that a private security company ‘could be subject to the HRA as a public authority if it has been “accredited” by the police and can exercise some police powers that are not available to the ordinary citizen.’ Jacob Rowbottom, Democ- racy Distorted: Wealth, Influence and Democratic Politics (New York: Cambridge University Press, 2010) at 153. See also Jody Freeman, ‘Extending Public Law Norms through Pri- vatization’ (2003) 116:5 Harv L Rev 1285 at 1345–7.

AGAINST RACIAL PROFILING 195

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

private laboratories or private prisons) should also be prohibited from continuing to treat this drug as illegal. Similarly, if the police ought to presuppose that the suspect acts freely, the same requirement also ap- plies to private companies which conduct security searches at airports. The objection from the unity of process is available even to those who

hold that certain behaviour should not be criminalized because it in- volves no culpability. For example, some hold that carrying a small quan- tity of cannabis is not culpable because there is nothing wrong with a competent adult using light drugs.49 On such a stance, using racial pro- filing when searching for light drugs for personal consumption is still objectionable even though this stance implies that the carrier is not cul- pable, and, hence, there is no need to presuppose that she acts freely. This is because, on this stance, not only do profiled searches for light drugs mistakenly take the carrying of these substances to be culpable, but they also inconsistently treat the carrying as being both culpable and thus free and determined by the suspect’s profile and thus unfree. The suspect should not be put to trial at all and, certainly, not on the basis of contradictory presuppositions. Proponents of racial profiling could respond to the objection from

the unity of process in various ways. Broad compatibilists could argue that referring to the issue of free will is unhelpful because there is no contradiction in presupposing that one’s behaviour is both free and determined by one’s race, age, and gender. Hence, even if the criminal process must presuppose that the accused’s behaviour is free, this free- dom is compatible with his behaviour’s being determined by his race, age, and gender. Alternatively, one could object to the comparison between searching and convicting by suggesting that the question of why the inference to culpable behaviour is excluded from criminal trials could be answered without any reference to the issue of free will. The lit- erature contains various attempts to justify why the inference to culpable behaviour should be excluded from criminal trials, many of which seek to identify an epistemic feature in the inference to culpable behaviour which makes its use in criminal trials objectionable. The criminal process thus does not need to presuppose that the accused’s behaviour is free in

49 Mark AR Kleiman & Aaron J Saiger, ‘Drug Legalization: The Importance of Asking the Right Question’ (1990) 18:3 Hofstra L Rev 527 at 533–4 (evaluating Szasz’s liber- tarian position). Thomas Szasz, Ceremonial Chemistry: The Ritual Persecution of Drugs, Ad- dicts, and Pushers, revised ed (Holmes Beach, FL: Learning Publications, 1985). Such a stance should be distinguished from other views which may accept that the behaviour is culpable but insist that it should nevertheless be left unpunished, either for political reasons (for example, the appropriate limits of state intervention via criminal law) or practical considerations (for example, efficient allocation of scarce enforcement re- sources).

196 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

a way which is incompatible with his behaviour’s being determined by his race, age, and gender. According to both responses, the objection from the unity of process fails because the question of free will is not treated differently in searching and convicting. The challenge for both responses would be to explain why the very

same inference to culpable behaviour, which they deem helpful enough to be used in searching, is not even admissible in criminal trials. While each of the epistemic accounts suffers from its own problems,50 they all share some common deficiencies.51 More importantly, why should the very same inference which is condemned as epistemically objectionable nevertheless be used for searching purposes? If the inference suffers from some epistemic deficiency, this deficiency arises not only in the context of convicting but also in that of searching. It is therefore difficult to see how either response could defend racial profiling without conced- ing that the inference to culpable behaviour should be used in criminal trials as well. One might argue that the fact that so many jurisdictions conduct pro-

filing and convicting simultaneously shows that these activities cannot be contradictory. If they were truly contradictory, it would be impossible to engage in both of them simultaneously, and they would look something like going to Mexico and staying home at the same time. At the end of the day, one either went to Mexico or stayed at home. However, this description of profiling and convicting ignores what the criminal process aims to achieve. In the criminal process, the police seek to single out those who are likelier to have acted criminally, and the court seeks to sin- gle out those who are culpable. Deploying profiling at the searching stage contributes to the criminal process’s effort to single out those who act criminally only if the suspect’s profile is probative of his acting crimi- nally. At the same time, convicting at the trial stage contributes to the criminal process’s effort to single out those who are culpable only if the accused’s behaviour was free. As a result, when the criminal process de- ploys profiling, it surely engages simultaneously in profiling and convict- ing, but it fails in either profiling usefully or convicting justifiably. At most, one of these goals has been achieved. That it might remain unknown which of these two was achieved does not mean that it is possible to achieve both simultaneously, just like it is impossible to go to Mexico and stay home.

50 For a critical survey of various epistemic accounts, see Amit Pundik, ‘What Is Wrong with Statistical Evidence? The Attempts to Establish an Epistemic Deficiency’ (2008) 27:4 CJQ 461.

51 Amit Pundik, ‘The Epistemology of Statistical Evidence’ (2011) 15:2 International Journal of Evidence and Proof 117.

AGAINST RACIAL PROFILING 197

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

While the argument so far has assumed that certain behaviour is either free or unfree, the law sometimes treats the same person as one thing for one purpose and as the opposite for another. For instance, a person may be considered ‘married’ for the purpose of tax law but not for the purpose of family law. One could argue that, similarly, the person’s beha- viour could be considered unfree for one purpose (searching) but free for another (convicting). However, even if another legal status, such as ‘cohabitation,’ yields the same legal rights and duties as ‘marriage,’ it does not entail that the law answers the same question (whether a cer- tain couple is married) with opposite answers. And even if it did, ‘mar- ried’ describes a social construct, and, hence, it is relatively easy to accept that it is context sensitive – that the question of marital status may depend on the purpose for which the question is asked. By contrast, ‘free’ describes a property of an action, and most, if not all, theories of free will identify an action as free or unfree irrespective of any social con- vention or belief about this action’s being free or not. Some might hold that even if ‘freedom’ is context insensitive, the attri-

bution of criminal responsibility is a social practice which is discon- nected from the question of whether the action is free.52 They could draw support from Peter Strawson’s claim, according to which the prac- tice of attributing moral responsibility does not require settling the meta- physical problem of free will,53 concluding that if this is true for moral practices, it is all the more so true for legal practices. However, the dis- tinction between justified and unjustified attribution of criminal respon- sibility, which any theory of criminal responsibility needs to make, is likely to rely on (or to bring through the backdoor) notions very similar to ‘freedom’ and ‘control.’54 For how else would a theory of criminal law explain why the conviction of a mentally ill person, or for a coerced action, is unjustified? A more plausible stance would be to accept that the attribution of

criminal responsibility requires presupposing freedom but to hold that

52 Cane, for example, takes legal responsibility to be disconnected from the issue of free will and suggests that ‘responsibility might be a heterogeneous, context-specific prac- tice and concept.’ Peter Cane, Responsibility in Law and Morality (Oxford: Hart Publish- ing, 2002) at 25; see also 66.

53 Peter Strawson, ‘Freedom and Resentment’ (1962) 48 Proceedings of the British Academy 1.

54 The ‘criminal law theorist must . . . distinguish between actions for which the defen- dant is responsible and actions which he is not responsible for. If determinism is true, the idea that such an action does not have a cause cannot do the job. On what grounds, then, are we to make such a distinction? That problem is central to any account of criminal responsilibity.’ Victor Tadros, Criminal Responsibility (Oxford: Oxford University Press, 2005) at 69.

198 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

searching is mainly conducted for the purpose of prevention (for exam- ple, of deadly incidents involving illegal firearms), for which the freedom of the suspect’s behaviour is irrelevant. While searching and convicting are both parts of the same process, each has a different purpose. How- ever, although prevention is important to policing, it is hard to deny that another central goal of policing is to bring offenders to justice. To see this, consider the following hypothetical legal rule, according to which any illegal weapon found in a profiled search is inadmissible as evidence in court. Had such a rule existed, this search would likely have been con- ducted very differently (by using other, non-profiled methods). This shows that even though preventing the use of illegal firearms is impor- tant, bringing offenders who carry them to justice is an indispensable goal of existing searching practices and that giving up this goal is likely to change the practice itself. As long as the criminal process seeks to avoid punishing those who are

not culpable for their (unfree) actions,55 searching should not be based on the presupposition that the suspect’s actions are unfree. Sure, it might be suggested that some types of search are, or should be, con- ducted for the purpose of prevention alone (for example, searching pa- tients hospitalized in a psychiatric secure unit for sharp objects), and these searches do not need to presuppose that the suspect acts freely.56

However, such a move in fact accepts the argument of this article; it con- cedes that racial profiling could be used only at the price of not prose- cuting any offender who was caught by a profiled search.

55 This constraint does not imply retributivism, namely that punishment is inflicted because it is deserved. It is argued here that any theory of punishment which consid- ers culpability as a constraint on other legitimate goals of punishment should refrain from profiling and convicting simultaneously. One notable example of such a theory is Hart’s mixed theory of punishment, which accepts the retributivist constraint ac- cording to which ‘only those who have broken the law – and voluntarily broken it – may be punished’ while rejecting retributivism as the ‘General Justifying Aim of the system.’ see HLA Hart, Punishment and Responsibility, 2nd ed (Oxford: Oxford Univer- sity Press, 2008) at 9.

56 One might argue that imposing administrative fines does not require presupposing that the person acts freely, hence profiling in such a context is not subject to the objection of this article (for example, it is arguably permissible to search mainly young black men for failing to carry their driver’s licence). Whether this is true depends, of course, on the substantive theory underlying administrative fines and on whether the law is willing to impose fines even on those who are not culpable.

AGAINST RACIAL PROFILING 199

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

IV Which types of search are objectionable?

The discussion so far might have created the wrong impression that the argument of this article objects to any search whatsoever. For example, searching a suspect for the murder weapon after noticing that he has bloodstains on his clothing involves an inference from the bloodstains to his acting culpably. If even such a search is impossible to conduct with- out jeopardizing the justifiability of convicting the suspect of murder, then the argument renders policing impossible and, hence, must be flawed. However, the argument of this article does not object to this inference (nor to many others). The simplest instance of a search that is unobjectionable is the basic airport security check conducted on all pas- sengers. This search is not based on profiling at all; it does not require inferring that some passengers are more likely than others to carry ex- plosives, and, hence, it does not require presupposing that some passen- gers act unfreely.57 Another example is ruling out suspects who have had no opportunity to commit the crime. If there are two suspects of a mur- der by stabbing, and one of them is completely paralyzed, focusing on the other person is not objectionable. Paralyzed people are outside the group of potential suspects to begin with because they are incapable of committing the crime. As argued above, devoting police effort to investi- gating young black men is rational only if based on a causal generaliza- tion that being a young black man causes members of this group to commit crime. By contrast, ruling out the paralyzed and focusing on the able-bodied is not based on the causal generalization that being healthy (or being able to move) causes people who are not paralyzed to commit murder. Being able to commit the crime is a necessary condition of com- mitting the crime, not its cause – just like the presence of the victim at the crime scene.58

More importantly, the objection does not apply to inferences such as that drawn from the bloodstains because of the direction of the causal connection underlying such inferences – his act of murder was not caused by his having bloodstains but, rather, the other way around. As a result, this search cannot be objected to on the grounds that it presup- poses a cause which renders his act of murder unfree. Whatever were the causes of his act of murder, the bloodstains have nothing to do with them. More generally, the argument of this article objects to an

57 By contrast, choosing passengers for enhanced security checks according to their pro- file is of course objectionable.

58 Hence, this article assumes that being a cause is not reducible merely to being a neces- sary condition (though it does not assume that being a necessary condition is not part of what being a cause is).

200 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

inference to culpable behaviour only if the direction of the causal con- nection is from the property the person shares with the other people to whom the generalization applies to the culpable behaviour. It is only then that the behaviour might have been caused by an antecedent which rendered it unfree. The argument does not object to inferring from the suspect’s having

bloodstains on his clothing that it is likelier that it was he who killed the victim. The inference requires a generalization along the lines of ‘people who have bloodstains on their clothing are likelier to be involved in the murder than those who do not.’ Unlike racial profiling, it is possi- ble to draw this inference without presupposing that there is a shared property which caused the group members to commit the crime. Rather, it is their culpable behaviour which causes them to have the property they share; committing the murder caused them to have bloodstains on their clothing. While the underlying generalization is still causal, draw- ing the inference does not require presupposing anything about the causes of the suspect’s culpable behaviour, only about its effects. Thus, there is no contradiction between such a causal generalization and freedom because there is no tension with the requirement that the agent’s culpable behaviour was not determined by (certain) antecedent causes.59

For similar reasons, the argument does not object to many other infer- ences commonly used by the police. For example, if an eyewitness pro- vides the police with a facial composite, there is no problem with the police searching for a suspect who matches it. This is true even if the only thing that the eyewitness says is that the perpetrator was black. As in the bloodstains example, nothing is presupposed in it about the causes of the suspect’s culpable behaviour. In particular, it is not presup- posed that the suspect’s being black caused him to commit the crime. Rather, it is the suspect’s culpable behaviour which caused the eyewit- ness to see a black perpetrator. Similarly, searching for a suspect whose DNA profile matches a sample found at the crime scene is not objection- able. Rather than presupposing that his culpable behaviour was caused by his genetic composition, the inference is based on his culpable beha- viour’s causing his genetic composition to be found at the crime scene.60

These examples also illustrate the importance of identifying the exact

59 This example also illustrates an important difference between the argument of this article and existing attempts to give an account of what is wrong with using statistical evidence in court. While some accounts focus on what caused the creation of the evidence (for instance, Judith Jarvis Thomson, ‘Liability and Individualized Evidence’ (1986) 49:3 Law & Contemp Probs, 1999), this article focuses on what caused the individual’s behaviour.

60 For a more detailed explanation why DNA evidence is not subject to the objection, see Pundik, ‘Freedom and Generalisation,’ supra note 10, s 5.B.

AGAINST RACIAL PROFILING 201

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

generalization involved, evaluating its empirical basis, and considering whether using it for policing conflicts with other presuppositions made in the criminal process. That said, the variety of searching practices to which the argument of

this article objects is still vast. In addition to practices in which the use of racial profiling is transparent and blunt, the objection also applies to those in which the use of profiling is subtler.61 One category which is objectionable is search via proxy. Some might insist that in some search- ing practices the searched-for behaviour does not have to be free because it is only used as a proxy for other, more serious offences. For example, searching for cannabis for personal consumption could be conducted in order to catch those carriers who also carry cocaine or ille- gal firearms and put them to trial for these offences. If carrying cannabis for personal consumption is truly a proxy, and carriers are not put to trial even when caught with cannabis, the search practice need not pre- suppose that the carrying is undertaken freely and culpably. In such a case, it could be argued that inferring from an individual’s profile that he is likelier to carry cannabis may be probative and unproblematic. However, the question is why carrying cannabis makes the suspect

more likely also to carry cocaine or illegal firearms. As argued in second section of this article, the inference is probative only if there is a causal connection between carrying cannabis and carrying cocaine or illegal firearms. As a result, even if the carrying of cannabis is truly a proxy, the argument of this article still applies to such practices because the contra- diction is merely pushed to the next step in the inferential chain, infer- ring from his carrying cannabis that he is likelier to commit other, more serious offences. This article has argued that the suspect’s profile is pro- bative of his carrying cannabis only if the carrying is undertaken unfreely. Not putting to trial carriers who are caught with cannabis alone impli- citly accepts this. But if the carrying is unfree, it becomes irrational to infer from it that he is likelier to commit the other, more serious of- fences freely and culpably. Such an inference is irrational because there is no difference between inferring that he is likelier to commit those crimes freely and culpably from his carrying and inferring it from his racial profile. In both cases, it is inferred that the agent is likelier to act culpably from a property beyond his control, be it either his racial pro- file or his carrying light drugs (unfreely). Thus, the use of racial profiling in police searches for cannabis is objectionable even if carrying it is merely a proxy for other, more serious offences.

61 For a detailed description of the development of existing racial profiling methods, see Bernard E Harcourt, Against Prediction: Profiling, Policing, and Punishing in an Actuarial Age (Chicago: University of Chicago Press, 2007) chs 2–3.

202 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

The type of profiling which is most complicated to analyze is that based on the suspect’s previous behaviour. Consider, for example, the stopping and searching of a young black man because of his saggin’ pants.62 In these cases, it is unclear whether the search involves an objectionable inference, like that involved in the above example of carrying cannabis, or an unobjectionable inference, like that involved in searching someone with bloodstains. On the one hand, some could argue that profiling based on sagging pants is merely racial profiling in disguise since almost every- one with sagging pants is a young black man, thus, sagging is probative of criminal behaviour only if it reflects some underlying causal connection between being a young black man and being a criminal. In profiling based on sagging pants, being a young black man is thus used as a proxy for criminal behaviour, similarly to how carrying cannabis is used as a proxy for committing other, more serious offences. On the other hand, others could argue that there is no need to presuppose that the criminal behaviour was caused by the suspect’s being a young black man because sagging is within the agent’s control. Whether such a profiling is subject to the argument of this article de-

pends on the type of inference and generalization used. Recall that, as in any kind of profiling, this inference requires presupposing a causal gener- alization between sagging and culpable behaviour.63 Such a causal connec- tion is likely to involve a common cause responsible for both the sagging and the culpable behaviour (the causal connection is unlikely to be direct because it is difficult to see how sagging would cause the wearer to start carrying an illegal firearm or how the acquisition or possession of the ille- gal firearm would cause the offender to start wearing sagging pants). One possible explanation for how sagging pants could be indicative of

carrying an illegal firearm is that both have a common environmental cause. For example, perhaps strong cultural influence or intense peer pressure led to the individual’s wearing sagging pants and to criminal be- haviour. If true, such an explanation immediately raises the question of whether such influence and pressure render both the suspect’s sagging and criminal behaviour unfree (a question on which libertarians and some compatibilists are likely to disagree).64

62 This is the fashion popular among young black men of wearing one’s pants very low to imitate prison inmates whose belts are confiscated. See Lee D Baker, ‘Saggin’ and Braggin’’ in Alisse Waterston & Maria D Vesperi, eds, Anthropology of the Shelf: Anthro- pologists on Writing (Oxford: Wiley-Blackwell, 2009) 46 at 47.

63 See supra Part II. 64 Another possible explanation, which raises a similar concern, is that these men freely

choose to join a certain group and adopt these variables, and being members of that group renders them likelier to act culpably. The same concern about the culpable be- haviour’s being unfree appears here because it is difficult to explain how becoming

AGAINST RACIAL PROFILING 203

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

Alternatively, one might insist that the common cause is not environ- mental but, rather, something arguably more internal to the suspect himself – perhaps a certain character trait or propensity causes these men both to adopt these variables and to act culpably. For example, if young black men sag their pants in order to imitate prison inmates,65

their looking up to prison inmates shows that they have little respect for the law, which, in turn, shows that they have a certain character trait or propensity to crime that makes them more likely to carry illegal weap- ons. Among the many difficulties to which such an explanation is sub- ject,66 one is particularly relevant to this article. If the suspected behaviour is determined by the suspect’s bad character or propensity to crime, it is unclear whether the behaviour is still free (a question which is yet again controversial to libertarians and compatibilists). Whatever the common cause, proponents of racial profiling based on

sagging pants should address the question of whether similar evidence of sagging pants should be admitted to support the conviction of a cer- tain wearer of sagging pants of carrying an illegal firearm. Interestingly, such general references to cultural influence, character traits, and pro- pensity are usually impermissible in court,67 so advocates of this explana- tion would have to explain why such references, if so helpful in determining whom to search, should not be used in criminal trials as well. While these are preliminary remarks, they illustrate that the use of racial profiling is strongly connected to the issue of freedom, and, as this article has argued, the presuppositions used in racial profiling might be inconsistent with other presuppositions about human freedom which are fundamental to the criminal process.68

V Conclusion

While this article has focused on the inconsistency between the presup- positions involved in using racial profiling in search practices, perhaps it

members of the group made them likelier to act culpably without referring to some cultural influence or peer pressure.

65 Angelica M Sinopole, ‘“No Saggy Pants”: A Review of the First Amendment Issues Pre- sented by the State’s Regulation of Fashion in Public Streets’ (2008) 113:1 Penn St L Rev 329 at 331.

66 Mike Redmayne, Character in the Criminal Trial (Oxford: Oxford University Press, 2015) at 35–41.

67 Jonathan D Kurland, ‘Character as a Process in Judgment and Decision-Making and Its Implications for the Character Evidence Prohibition in Anglo-American Law’ (2014) 38 Law & Psychol Rev 135 at 136–7.

68 For more on these questions, see e.g. George Sher, In Praise of Blame (New York: Oxford University Press, 2006).

204 UNIVERSITY OF TORONTO LAW JOURNAL

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

could also shed light on other potential tensions which the use of racial profiling might generate. In particular, some tensions arise if the focus is shifted from the presuppositions which the use of a certain generaliza- tion requires to the consequences, implications, impressions, and so on which the use of that generalization creates. For example, a common objection to racial profiling is based on the stigmata these practices cre- ate, by implicating all group members in the reprehensible behaviour of a few.69 Focusing on Muslims in the airport security case is commonly ob- jected to on the basis that it stigmatizes all Muslims, almost all of whom are law-abiding citizens who pose no security threat to a flight. While the argument of this article is different, this common objection could be understood as somewhat analogous. The inference is deemed objection- able because it is perceived by some (such as the passengers themselves, the security officers, or even the general public) as implying that all Mus- lims, or at least those who are selected for the ‘enhanced security check,’ are criminals. Furthermore, perhaps these practices should be objected to not

(only) because they stigmatize Muslims as terrorists (‘bad agents’) but (also) because they stigmatize them as people who, like children, are not fully responsible (‘non-agents’).70 Under libertarian and non-broad compatibilist theories of freedom, instead of being treated as free agents, they are treated as people whose behaviour is determined by a causal antecedent factor which they all share and that renders their behaviour unfree. This treatment is particularly troubling when the group is disad- vantaged because it regards members of this group as being inferior to members of privileged groups. Be that as it may, it should be emphasized that the argument of this

article does not hinge on stigmata and perceptions, nor does it focus on the consequences of racial profiling. The problem highlighted in this article is internal to the aspiration of the search practice used to effec- tively identify those suspects who are in fact culpable. Using racial profil- ing to identify those who have engaged in a crime and then putting them to trial is irrational because it requires one to presuppose a fact which contradicts another presupposed fact.

69 RA Lenhardt, ‘Understanding the Mark: Race, Stigma, and Equality in Context’ (2004) 79:3 NYU L Rev 803.

70 Levin’s theory is a paradigmatic example of such an approach. Levin argues that ‘on average whites are more autonomous and responsible for their actions than are blacks (and Asians more autonomous than whites).’ Levin, Why Race Matters, supra note 39 at 322.

AGAINST RACIAL PROFILING 205

(Spring 2017) UTLJ © UNIVERSITY OF TORONTO PRESS DOI: 10.3138/UTLJ.3883

Copyright of University of Toronto Law Journal is the property of University of Toronto Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.