Racial policing is one of the most controversial police practices. Proponents typically argue that racial profiling is an important tool if the police is to use its resources efficiently in the interest of deterring and detecting crime. Critics typically argue that it harms, stigmatizes and disrespects minorities. In a recent article, Alexandra Lloyd (2021)  offers a novel epistemic objection to racial profiling. According to this objection, racial profiling typically involves heightened suspicion of those being profiled, which in turns involves blame them, which again in turn requires full, justified belief that the relevant people did something blameworthy. Statistics cannot ground such belief and accordingly police ought not to engage in racial profiling. In this response to Lloyd, I critically examine this novel objection … [please read below the rest of the article].

Image credit: Michael Fleshman via Flickr / Creative Commons

By “racial profiling”, Lloyd means “the use of race or ethnicities, or proxies thereof, by law enforcement officials as a basis for judgment of criminal suspicion” (p. 1; quoted from Glaser 2014, p. 3).[1],[2] Lloyd nicely summarizes her argument against racial profiling as follows:

(1) for blame to be apt, an agent should have a justified belief that the agent they blame is responsible for the bad act; (2) naked statistical evidence is insufficient to justify outright belief in the relevant proposition because such evidence fails to be causally connected to the events in question in the appropriate way; (3) therefore, naked statistical evidence cannot justify blaming; (4) racial profiling subjects individuals to heightened suspicion, which frequently involves blame; (5) therefore, when agents racially profile, they should have a justified belief, not just a high credence, that some individual is worthy of reasonable suspicion; (6) racial profiling relies on naked statistical evidence that is not causally connected to the incident in question and therefore cannot justify an agent’s belief that some individual is worthy of reasonable suspicion. Therefore, I conclude that this is a morally significant epistemic objection to the practice of racial profiling (p. 6; bold added).[3]

This is a novel and significant objection to racial profiling which deserves the attention of theorists addressing the (un)justifiability of racial profiling. It is novel, because previous objections to racial profiling have focused on quite different problematic features of racial profiling, e.g., the way in which it is harmful to members of those groups that are profiled (Eidelson 2015, pp. 173-222; Mogensen 2017), the fact that it stigmatizes members of racially profiled groups (Lever 2005), the way in which racial profiling involves giving its victims a reasonable sense of inferior political status (Hosein 2018), or because of how profiling often fails to pass a test of interpersonal justification (Lippert-Rasmussen 2006; Lippert-Rasmussen 2007, pp. 400-402). Lloyd’s objection focuses on how profiling violates a condition for apt blame and, thus, so she argues, justified suspicion. Lloyd’s objection is significant, in part because it applies even in cases where none of the previous objections in the literature to racial profiling applies.

In her article, Lloyd makes some good points. However, in the end I am not persuaded that her argument captures why we object to racial profiling (when we do). Below I explain why. First, I explain why I think her argument proves too much, as it were. Next, I go through the steps in her argument, which I found less convincing. However, before doing so I want to clarify two matters that might make a difference to how the reader will understand the wider dialectical context of my criticisms below.

First, racial profiling, as we know it, involves a lot of different morally unjustifiable features, e.g., disproportionate attention to specific racial minorities and use of force, abuse and intimidation. I take it is that Lloyd’s ambition is to show that even if racial profiling involved none of these obviously morally objectionable features—call such instances unalloyed racial discrimination (Lippert-Rasmussen 2006, pp. 191-192)—it still would not be justified.

Second, the criticisms below are directed against the argument she offers to fulfill this ambition. They are not intended to justify racial profiling as we know it or even to suggest that, under some circumstances, certain forms of unalloyed racial profiling could be justified. They are intended simply and only to show racial profiling is not morally unjustified, because it requires that the blame it putatively involves is apt and that in turn requires justified full beliefs to the effect that those profiled are worthy of heightened suspicion. Back to Lloyd’s argument.

Does Lloyd’s Argument Show Too Much?

Suppose that Lloyd is right that reasonable suspicion requires the belief (and not just a certain high credence being ascribed to the proposition) that the relevant individuals are worthy of suspicion. If so, racial profiling might be morally problematic. But if so, then other less clearly objectionable practices are problematic too. Consider age and gender profiling in relation to violent crime. Suppose that police has a policy of paying greater attention to young males—irrespective of race—than to 80+ citizens in the interest of reducing violent assaults etc.[4] In my view, such a policy—at least in certain variants—might well be justified. Lloyd’s argument, however, seems easily tweaked to apply to age and gender profiling as well. All we need to do is to substitute:

(4*) age and gender profiling subjects individuals, i.e., young males, to heightened suspicion, which frequently involves blame for (4) and revise (5-7—see note 3) accordingly.

Of course, we would not have to do so if profiling of young males in relation to violent criminal offences were different in terms of whether they frequently involve blame, but I do not see what Lloyd could say in defense of the existence of such a difference between the two. It might be that people who are subjected to racial profiling more often than young men, who are subjected to age and gender profiling, experience their treatment as if they are being blamed. But that in itself is neither here, nor there, because as Lloyd points out people can experience being blamed even when they are not (p. 6). And we may add: (Annoyingly) people can be blamed even if they do not experience being blamed.

Unlike me, some—Lloyd included perhaps—might worry less about the fact that her argument seems transferable to age profiling and for that matter to any other sort of profiling in police work and might be willing to say that this just shows, what they suspected all along, i.e., that profiling whichever the target group is, is morally unjustified.[5] I do not think, however, that this reply will do. First, I am not sure that any kind of profiling, e.g., profiling of perpetrators of hate crimes, is unjustified. Second, there is something specifically objectionable about racial profiling and pointing to a general condition that must be satisfied for subjecting someone to heightened suspicion cannot identify what makes racial profiling so. This motivates examining the different steps in Lloyd’s argument and in the next four sections I will address four potentially problematic steps.

Could High Credence Suffice for Aptness of Blame (1)?

I have some reservations about the first step in Lloyd’s argument, i.e., the claim that “for blame to be apt, an agent should have a justified belief that the agent they blame is responsible for the bad act”. Compare the following two cases:

Belief blames Suspect for having stolen his bicycle. Belief holds the full belief that Beatrice has stolen her bike. The evidence which is available to her is such that the chance of Suspect having done so is 9/10. Her evidence suffices for Belief’s belief being justified.

Credence also blames Suspect for having stolen her bicycle. Credence does not have a full belief that Suspect has stolen her bike. However, Credence reasonably infers from the available evidence, which she has conscientiously collected, that the chance of Suspect having done so is 999/1000.

Suppose Suspect acknowledges these facts; accepts Belief’s blame; but dismisses Credence’s blame on the ground that Credence does not hold the full belief that Suspect has stolen her bike. This differential response on Suspect’s part strikes me as indefensible. Given that Suspect concedes that Credence has a better evidence basis for blaming her for stealing her bike than Belief does, and given that it is the only other relevant difference between the two blamers apart from the difference pertaining to the full belief-credence distinction, and given that Suspect accepts that Belief’s blame is apt, should Suspect not concede that Credence’s blame is apt too?

More generally, I suspect that even if the difference between belief and credence does make a difference for the aptness of blame for a given intermediate level of credence, it is not true that levels of credence cannot be so high that they do not render blame apt even in the absence of a full belief. Credence in the second of my two examples might be aware of the fact that—let us stipulate—her evidence justifies her in forming the full belief that Suspect has stolen her bicycle and, if so, I fail to see why she should resist blaming Suspect for stealing her bike just because she lacks full belief.

As far as I can see Lloyd’s main argument for why high credence does not suffice for aptness of blame is an appeal to Buchak’s observation that “the degree of blame” one should assign to a wrongdoer “is based on severity of the act, not on [one’s] credence that she in fact did it” (p. 3).[6] That observation, however, is consistent with the view that once one’s credence crosses a certain threshold, blame becomes apt (even if greater blame does not become apt, if the credence is increased further).

Naked Statistical Evidence And Justified Belief (2)

I have a related reservation about step (2). As a starter note that it is not true in general that naked statistical evidence is insufficient to justify full belief in a proposition. Suppose I am about to buy a ticket in a lottery with millions of tickets and with only one winning ticket. My belief that my ticket is not the winning ticket is a justified belief, even if it is only based on naked statistical fact that there is a less than in a million chance that my ticket is the winning ticket.[7] I take it that Lloyd and I agree on this. However, she thinks things are different when it comes to “proposition[s] in the domain of blame” (p. 5)—specifically, things are different when it comes to the belief that the agent one blames is responsible for the bad act one blames them for. I am open to the possibility that what is required for a belief to be justified depends on what the content of the belief is. Perhaps beliefs about whether a ticket in a lottery is not a winning ticket requires different or less evidence to be justified than beliefs about whether one’s blame is responsible for the act one blames her for. However, I do not think Lloyd’s arguments suffice to show that this is so.

Crucial here is the fact that believing that someone is responsible for a bad act is not the same as actually blaming that person. It might be that there is a sense in which to blame someone for a blameworthy action (as opposed to, say: simply pretend that you blame them), you have to believe they are responsible for the blameworthy action. Hence, even if it is was a fact that blaming something on the basis of naked statistical evidence for her having done the blameworthy action in question is inapt blaming, it would not follow that one cannot justifiably believe that she is responsible for the bad (or blameworthy) action in question. The inaptness could result from whatever elements that one needs to add to the belief that the blame is blameworthy for the bad act to get to blame requiring more or other evidence than merely naked statistical evidence (more on this shortly).

Does Heightened Suspicion Frequently Involve Blame? (4)

Let us now turn to Lloyd’s third premise, i.e., (4) in the summary above: “racial profiling subjects individuals to heightened suspicion, which frequently involves blame”. Just to avoid misunderstanding: I do not deny that heightened suspicion sometimes—perhaps even frequently—involves (unreasonable) blame of some sort.[8] However, I suspect that frequently it does not. “Frequently” is a tricky term and it could both be true that racial profiling subjects individuals to heightened suspicion that frequently involves blame, and that racial profiling subjects individuals to heightened suspicion that frequently does not involve blame.[9] Just as it is true that, frequently, it rains in England and frequently it does not.

To see the motivation for the claim that, frequently, heightened suspicion does not involve blame, consider the fact that, typically, racial profiling involves ascribing a higher probability to a member of a certain racialized group having a certain target property, e.g., being involved in criminal activity of some kind, than members of other racialized groups. That, however, is consistent with its being the case that, for any individual member of the group being subjected to “heightened suspicion”, the profiler ascribes a very low probability to that individual having the relevant target property and, by implication, a very high probability to that individual not having the relevant target property. A profiler in an airport might direct heightened suspicion against AMEMSA people and still think the probability of any individual AMEMSA person, whom she profiles, being a terrorist is minuscule.

Take the case of racial profiling in connection with the Washington D.C. sniper—one of the cases discussed in Risse and Zeckhauser’s (2004, p. 137) article to which Lloyd refers. Police thought they were look for a white male, since they are overrepresented in serial killings crime statistics. One thing police did to catch the sniper was to establish road blocks in the vicinity of suspected Washington D.C. sniper crime scenes and check drivers—more thoroughly so if the driver matched their profile of the perpetrator than if the driver was a black male (which, actually, meant that the Washington D.C. sniper, who was a black male, got past one of these road blocks, when he could easily have been arrested with his rifle in his car’s trunk).[10] Plausibly, when police officers stopped any given white male driver on that occasion, they would consider it overwhelmingly unlikely that the driver was the Washington D.C. sniper. Accordingly, it seems plausible to assume that they did not blame the profiled white male driver. Admittedly from the armchair, I conjecture that it is quite uncommon to blame someone for something one thinks it is overwhelmingly unlikely that they have done.[11] Hence, I conjecture that in cases of racial profiling where the probability that those subjected to heightened suspicion is as low as that involved in the case of the Washington D.C. sniper (and perhaps also cases where the probability is higher, but still pretty low), these cases are unlikely to involve the profiler blaming the subject of heightened interest.

Taking a step back, I sense a possible source of confusion in Lloyd’s argument, which is that “heightened suspicion” is a notion that is comparative across groups. Hence, “heightened suspicion” could be directed against an individual on the basis of her group membership compatibly with ascribing a very low credence to that individual having performed any bad act. It seems to me that Lloyd fails to respect this distinction in inferring (5)—“when agents racially profile, they should have a justified belief, not just a high credence, that some individual is worthy of reasonable suspicion” (p. 6)—from (4) which is about “heightened suspicion”.

There is another reason why I think it is dubious whether heightened suspicion frequently involves blame. This reason has to do both with what heightened suspicion is and with what blame is. As already indicated, to blame someone for a bad act is to do something in addition to believing that they did the blameworthy act in question. Philosophers differ on what that additional thing is (Brunning and Millam 2018, p. 145; Coates and Tognazzini 2013, pp. 8-17). Perhaps the most common view is the Strawsonian view that the additional something is negative reactive attitudes, e.g., resentment or anger. However, directing heightened suspicion against someone in relation to a blameworthy act, most naturally is taken to involve simply ascribing a higher probability (relative to some relevant baseline) of their having performed the blameworthy act in question and not to require any extra element like adopting a certain reactive attitude in response to them. In fact, it seems common for people to direct heightened suspicion against someone without harboring any negative reactive attitudes towards them, e.g., because they have no personal stake in whether the wrongdoer is guilty of the wrongdoing. If so, it seems reasonable to believe that frequently in cases involving heightened suspicion, the suspecter does not blame.[12]

Has Lloyd Shown That Profiling As Such Requires Justified Belief?

(5) is supposed to follow from (3) and (4). But it is unclear that it does. One reason is that (4) merely contends that racial profiling “frequently” involves blame. But then one would suspect that at most (3) and (4) entitles Lloyd to infer:

5*) therefore, when agents racially profile, frequently (i.e., in those cases where profiling does involve blame) they should have a justified belief, not just a high credence, that some individual is worthy of reasonable suspicion.

Or to put this point differently:

(3) and (4) does not entitle Lloyd to infer that profilers should have a justified belief to the effect that the profiled persons are to blame in those cases where profiling does not involve blame and, obviously, her use of “frequently” suggests that she believes that such cases are conceptually possible at least.

Just as we cannot infer from the fact that nothing justifies racism—true, of course—and the fact that racial profiling frequently involves racism—true also, I believe—that racial profiling is never justified. These two propositions do not entitle you to the claim that racial profiling, which does not involve racism—rare or hypothetical (though keep in mind the Washington D.C. sniper case which involved racial profiling of white males, but presumably no anti-white racism) as they might be—is unjustified. Of course, even non-racism-involving forms of racial profiling might be unjustified, but, if so, they are unjustified on non-racism-involving grounds that are not captured by the two claims just mentioned.

Or to put this challenge in yet another way:

Even if we grant Lloyd (3) and (4), at most that enables her to show that there is a certain aptness-of-blame-related objection to a specific kind of racial profiling—blame-involving-racial profiling—not that there is an objection to profiling as such.

Thus, to the extent that we think that racial profiling full stop is morally objectionable, our reason must be different from the one Lloyd offers.

Conclusion

Most or all forms of racial profiling, whether actual forms or unalloyed forms that are cleansed of abuse, hostility, biased statistics and are simply a matter using sound statistics to achieve morally desirable ends, e.g., reducing crime, might be morally unjustified. However, I do not think Lloyd’s argument appealing to aptness conditions for blame succeeds in showing any of this. While I have not shown that an aptness-of-blame-focused objection to racial profiling cannot be tweaked to avoid my objections, there are at least two reasons to remain skeptical of this possibility.

First, as indicated (racial) profiling does not typically involve blame (as opposed to hostility) and it certainly is not a defining feature of (racial) profiling that it does. Second, most people think racial profiling is morally objectionable in ways that other forms of profiling are not. But then we cannot capture what those ways are by focusing on a condition for aptness of blame, which, if it applies to racial profiling, applies to profiling in general. These obstacles, of course, become less relevant if Lloyd’s argumentative ambitions are more modest than I have taken them to be, i.e., if her argument were simply meant to show that there is a specific blame-involving form of racial profiling which is objectionable for a reason, which is not specific to racial profiling but that applies to profiling in general. As indicated above I am not sure the argument would work, but I thought—perhaps mistakenly—that it was meant to show more than that.

Specifically, I note that right after having offered an attempt to rebut a counterexample purportedly showing that heightened suspicion need not involve blame, Lloyd says with little explication that she does “not aim to defend a universal connection between suspicion and blame” and describes her argument as showing “an important epistemic problem with many instances of racial profiling that take place in the real world” (p. 7). So perhaps her objection is meant to show less than I read other parts of her paper to imply and my challenge to (5) does not really address her real view. If so, we disagree less than I thought.

Acknowledgments
This work was funded by the Danish National Research Foundation (DNRF144).

References

Brunning, Luke and Milam, Per-Erik. 2018. “Oppression, Forgiveness, and Ceasing to Blame.” Journal of Ethics and Social Philosophy 14 (2): 143-178.

Coates, D. Justin and Neal Tognazzini, eds. 2013. Blame: Its Nature and Norms. New York: Oxford University Press.

Eidelson, Benjamin. 2015. Discrimination and Disrespect. Oxford: Oxford University Press.

Glaser, Jack. 2014. Suspect Race. New York: Oxford University Press.

Hosein, Adam. 2018. “Racial Profiling and a Reasonable Sense of Inferior Political Status” Journal of Political Philosophy 26 (3): 1-20.

Lever, Annabelle. 2005. “Why Racial Profiling Is Hard to Justify.” Philosophy and Public Affairs 33 (1): 94-110.

Lippert-Rasmussen, Kasper. 2007. “Nothing Personal: On Statistical Discrimination.” Journal of Political Philosophy 15 (4): 385-403.

Lippert-Rasmussen, Kasper. 2006. “Racial Profiling versus Community.” Journal of Applied Philosophy 23 (2): 191-205.

Lloyd, Alexandra. 2021. “An Epistemic Objection to Racial Profiling.” Social Epistemology 1-9. doi: 10.1080/02691728.2021.1943560.

Mogensen, Andreas. 2017. “Racial Profiling and Cumulative Injustice”, Philosophy and Phenomenological Research 98 (2): 452-477.

Risse, Mathias and Richard Zeckahauser. 2004.“Racial Profiling.” Philosophy and Public Affairs 32.2:  131-170.

[1] Given the accidental confusion caused by crisscrossing the numbering scheme (1-6) employed by Alexandra Lloyd in her summarized argument (on page 6 of the article and quoted by Lippert-Rasmussen), the page numbering (1-9) of the online version of Lloyd’s article, and Kasper Lippert-Rasmussen’s extensive references to Lloyd’s numbered elements in his reply, references to page numbers in citations will be preceded with a ‘p’ or, as appropriate, ‘pp’. In addition, I have bolded the elements of Lloyd’s argument to which Lippert-Rasmussen refers in the text.

[2] All unattributed references are to Alexandra Lloyd “An Epistemic Objection to Racial Profiling.” Social Epistemology 1-9. doi: 10.1080/02691728.2021.1943560.

[3] With the risk of being pedantic, I note that I think that “(7)” should be inserted before there last “Therefore” in this quotation and that the last sentence in the quote should probably be reformulated to say “Therefore, agents should not racially profile”. The summarized argument needs to be cleaned up in a number of other ways too, which, however, are irrelevant for present purposes.

[4] In fact, it seems that, implausibly, an investigator who based on forensic evidence and testimony from witnesses directs heightened suspicion against a particular suspect, though without forming the full belief that the suspect committed the crime, does something morally objectionable on Lloyd’s account.

[5] Lloyd might add that she is not aiming to show that racial profiling is unjustified, all things considered. Rather, she is simply identifying a pro tanto reason against it, which sometimes might be defeated by countervailing reasons. My sense is that she does not want to go that route, but I might well be mistaken.

[6] I guess one could say the same thing about full belief and from that one cannot infer either that the suitable degree of blame depends on belief.

[7] The evidence in question here is statistical in Lloyd’s sense, i.e., I have the same evidence available to me whether I buy the winning ticket or I don’t (p. 5).

[8] It is a bit unclear which object of blame Lloyd has in mind. In many cases of racial profiling (unlike the case of the Washington D.C. sniper which I discuss below), profilers do not have any particular criminal act in mind and, thus, it is hard to believe that they blame profiled persons for any specific crime. As Lloyd points out: racial profiling is typically pre-emptive (p. 2). Of course, profilers might blame members of profiled groups for, say, societal ills in general. If they do, we are moving into the territory of what I would call alloyed racial profiling, i.e., the territory of morally objectionable racial profiling which however, is morally objectionable not in virtue of its definitorial properties but in virtue of properties that in a conceptual (though perhaps not in a psychological) sense are only contingently associated with racial profiling.

[9] I take it that even if it is consistent with (4) that racial profiling subjects individuals to heightened suspicion, which frequently does not involve blame, this claim goes against Lloyd’s overall argument.

[10] This is an unusual case of racial profiling in the sense that, in a US context, African American males are the prime targets of racial profiling. However, Lloyd’s argument pertains to racial profiling in general, so the Washington D.C. sniper cases is relevant in our context.

[11] This is not to imply that our blaming dispositions are perfectly rational, e.g., I suspect that one common bias is to adjust our assessment of probabilities in such a way that they rationalize our blaming those whom we blame.

[12] Admittedly, Lloyd briefly suggests that treating racial minority people differently on the basis of heightened suspicion “involves blame” (p. 7). However, on a normal understanding of blame, I can take precautions against someone on the basis of a heightened suspicion, even if I in no way blame that person.

Categories: Critical Replies

Tags: Alexandra Lloyd, belief vs. credence, blame, Kasper Lippert-Rasmussen, policing, racial profiling, statistical discrimination