1. Introduction

This essay deals with a narrow and meta-ethical claim about discrimination. It argues for a particular thesis: Discrimination law must make a judgment on the value and content of beliefs, it must be, in other words, ‘content-sensitive’. The essay will not lay out a principle for how this judgment is to be made, merely claim that the judgment must be content-sensitive. My argument is conditional on several assumptions, to which I will call attention but not, for reasons of space, defend: I will assume that the reader holds certain intuitive stances on exemplar cases, and that these intuitions are philosophically meaningful. I will also assume that the law is the correct remedy for wrongful discrimination. This may alienate any libertarians in the audience, but this is a sacrifice I am willing to make. It is true that many forms of discrimination, including on the basis of race and gender, should be illegal; a content-sensitive principle is required to adequately defend this proposition.

The central problem is to resolve what kinds of characteristics are wrong to discriminate against. Our cases will deal with private businesses, and when the government can compel them not to discriminate against their clientele. Discrimination, defined narrowly as treating different groups differently, is not always wrong, even when the groups in question track protected characteristics.[1] Nevertheless, it is sometimes wrong; we want to focus on when wrongs of discrimination are of a level or kind that requires government intervention. As shall become clear, the essential aspect of this question is this: What are the groups against which we cannot allow discrimination? Section two will lay out the cases under discussion and articulate a first-pass test for wrongful discrimination. Section three will develop the interesting complications to these tests, and section four will defend content-sensitive tests as the necessary solution to these complications.

2. A Schematic Discrimination Test

To begin, I will lay out some exemplar cases. We will discuss the case of Racist Restaurateur, who wants to deny service to a Black man; Anti-Catholic Barista, who will not serve anyone coming in after mass; and Kosher Butcher, who refuses to slaughter swine. We will also deal with the case of Sign Salesman, who refuses to sell to an odious but nonviolent protest group; and finally Biblical Baker, who refuses on religious grounds to make a wedding cake for a couple. These cases are designed to help tease out different aspects of the discrimination question and show that it cannot be resolved without making adjudications regarding the content of beliefs.

Decisions in the first three cases are straightforward. I assume that it is intuitive and un-controversial that Racist Restaurateur and Anti-Catholic Barista both engage in impermissible discrimination of very similar kinds, in that both refuse to serve members of a particular group just because of their membership. Racist Restaurateur is apparently a holdover from the worst days of the Jim Crow south, and Anti-Catholic Barista is a nearly precise parallel along religious lines. These types of discrimination have occurred in the past, and the historical record supports the efficacy and justice of legal remedy in these cases. The Kosher Butcher is likewise uncomplicated. Because of (perhaps idiosyncratic) interpretation of Jewish dietary law, this imaginary abattoir refuses her services to a prospective client that raises pigs. I assume here that the reader both shares these intuitions and that they consider the use of state power in these cases justified; the conclusions I reach are contingent on these assumptions.

I will try to codify these intuitions into principles, and attempt to make the principles as general, coherent, and common-sense as possible, while still being properly tethered to the intuitions that give rise to them. I will revise the principles as we cover more cases. Abstracting automatically (and naïvely) from the cases just mentioned, the first candidate principle might be:

D1: Refusing service based on group membership should be illegal, but the government cannot force service that violates religious conviction.

The following cases complicate matters. Sign Salesman is wrongfully discriminating under D1: he is refusing service based on group membership. Precisely which group will fit most aptly into this case to create the desired intuition will depend on the individual judging the case, but we might imagine various (nonviolent) groups, such as white supremacists, staunch anti-abortion activists, pro-Israeli settlement protesters, or pro-Boycott, Divest, and Sanction (BDS; a group critical of Israel) activists. The point is to find a situation in which it does not appear to be wrong for Sign Salesman to refuse service to some subset of these groups, and, more to the point, it does not seem the sort of discrimination that ought be outlawed. This invites a revision of D1. The dimension of discrimination (denial of service at a business) is similar to the case of Racist Restaurateur and Anti-Catholic Barista. The relevant change, then, must come in the basis clause, the ‘group membership’ section. We might use the vague:

D2: Refusing service based on some types of group membership should be illegal, but the government cannot force service that violates religious conviction.

The Biblical Baker case will help clarify this vagueness. This case presses at the outer edge of intuitive understanding of discrimination, but by varying the group discriminated against, we may begin to see more clearly. We can imagine a loose spectrum of cases this sort of discrimination, varying as we change the group targeted. On one side of the spectrum are cases that are uncontroversial instances of acceptable discrimination. If Biblical Baker notices that a prospective customer has shaved the corner of his beard (pace Leviticus 21:5), and refuses service on these grounds, we would consider them ‘mad, not bad’ as Cavanagh points out (2002, 156). Away on the other side of the spectrum we could imagine that the targeted group is a racial one. If Biblical Baker is an overzealous Mormon, and it is prior to the 1978 Revelation on Priesthood, Biblical Baker could claim to be discriminating on religious grounds when they refuse to bake for a Black couple. ‘Humbug,’ we say. The religious motivation for discrimination does not excuse it.[2] Where will cases where the couple targeted is gay, or Hindu, or vegetarian fall?

3. Specifying Vulnerable Groups

In order to resolve this case, we will have to ground what has been an abstract principle of discrimination, specifically the ‘some types’ phrase in D2. Anti-Catholic Barista seems to indicate that religious belief outlines one such type of group. Religious belief can, contrariwise, provide justification some types of discrimination, but not others. But at this point we run into a problem: schematically, it is difficult to distinguish between a sincerely held religious belief and any of the various political beliefs that might motivate Sign Salesman to refuse service. What means do we have to distinguish between a Catholic, on the basis of whose sincere belief we cannot discriminate, and, say, a white supremacist?

It is in fact very difficult to draw such a line. Both types of beliefs can change, but as this fact does not eliminate the possibility for confessional discrimination, it seems it cannot eliminate the possibility of political discrimination. We can reasonably suppose both types of beliefs are sincerely held, central to their holder’s identity, and date back to early childhood. Both might mark the holder out as a member of a distinguishable community, and both may be grounded in philosophical or theological ideas and have only secondary political or social content. Both characteristics are in fact protected under international law, such as the International Covenant on Civil and Political Rights (Article 26). In fact, we could make these beliefs alike in every aspect not having to do with their actual content, and still maintain the intuitive schism that casts Anti-Catholic Barista as a wrongful discriminator and Sign Salesman as a rightful one. Which group should be protected from discrimination?

There has been some discussion of this point in the literature. Scanlon argues that discrimination against a group rises to a significant wrong it is part of a pattern of beliefs “so widely held in a society that members of that group are denied access to important goods and opportunities” (2008, 73-74). Fiss claims that discrimination against Black people in the United States is so odious because of their social, economic, and political disenfranchisement, making them a “specially disadvantaged group” in a way that justifies legal action against discrimination (1976, 151-155). Anderson and Girgis propose a sophisticated set of tests: wrongful discrimination in need of remedy must impose “a) material and/or b) social harms c) which the law can best cure” (2017, 179). These types of tests do not hinge on the content of belief. The authors do not claim, for instance, that discrimination is wrong only if is against those that deny the conclusions of the Council of Trent. Rather, these tests focus on the context of a group within society.

But these tests do not match important intuitions. Compare Anti-Catholic Barista to Sign Salesman, and imagine that the group in need of protest signage is a widely socially and economically maligned ‘race-realist’ movement, that (always peacefully) argues for the deportation of all non-White non-Christians from the US on the basis that they lack moral worth. On the tests above, Anti-Catholic Barista would be doing no wrong, as Catholics are not (any longer in Barista’s country, we can assume) the subject of any substantive discrimination. Sign Salesman, however, is denying service to a group we may consider is widely condemned and whose members face significant disadvantage, and so runs afoul of the tests above. The intuitions I assume, however, cut precisely the opposite way.

4. The Need for Content-Sensitivity

There are three ways to resolve this situation. We could revise our principle in in order to 1) recast Anti-Catholic Barista as permissible, or so as to 2) make Sign Salesman impermissible. Neither is, I think, palatable, as both cut firmly against what this essay assumes are common intuitive stances. The right choice, I believe, is to 3) develop a principle that distinguishes between beliefs on their content, rather than context, and dropping the implicit assumption present in many liberal accounts that makes government officially agnostic on what constitutes the good life.[3] This would let us coherently follow intuitive judgements in the cases mentioned.

What we need is the ‘best public morality’ that Arneson proposes (2013), a set of common moral rules and judgments that are action guiding for both individuals and government. This would of course be controversial. But it would also get satisfyingly at what seems to be the real germ of dispute on the Barista — Salesman spectrum: some deeply held beliefs ought to be protected from discrimination, and some should not, a distinction that hinges, in a significant way, on the content of beliefs. Some beliefs are not compatible with just societies. Our final D3 then, might run, bluntly:

D3: Refusing service based on group membership or defensible belief should be illegal, but the government cannot force service that violates religious conviction.

Obviously, the phrase ‘defensible belief’ is argumentatively load-bearing, not to mention normatively fraught. As above, this paper is not concerned with the prickly question of what, precisely, this might entail; this is an issue that must be decided by robust public debate within each society, as an essential question of democratic justice. The broad shape of whatever sub-principle determines ‘defensible belief’ is clear: it must thread the needles of the examples made here, protecting the customers of Racist Restaurateur, Anti-Catholic Barista, and resolve the muddled cases of the Biblical Baker, while also protecting Sign Salesman and Kosher Butcher in their denial of customers. I suspect these rules will only be satisfactorily developed as part of a value theory that is both deep and broad, what Dworkin would call a ‘hedgehog theory’ (2011). This concludes the defence of the claim that picking out cases of wrongful belief-based discrimination requires judgments about the contents of beliefs.


This piece would not be any good without the close attention of Dr Liam Kofi Bright, Sarah Quinn, and two anonymous PPE student editors, for whose work I am very grateful.


[1] E.g., offering free screening for sickle cell disease to Black but not Native American people would be permissible, due to higher incidence of the disease in the former group. 

[2] Note that we don’t need to interrogate either the degree of religious conviction or the coherence of the religious reasoning in this (or other) cases.

[3] Anderson and Girgis seem to agree, at least implicitly, in their use of ‘unfair’ to qualify harms of wrongful discrimination. They do not develop the point fully, however (2017, 180).


Arneson, R. 2013. Discrimination, Disparate Impact, and Theories of Justice. In Philosophical Foundations of Discrimination Law, ed. D. Hellman and S. R. Moreau 2013, 87–111. Oxford: Oxford University Press.

Cavanagh, M. 2002. Against Equality of Opportunity. Oxford: Clarendon Press.

Corvino, J., R. T. Anderson, and S. Girgis. 2017. Debating Religious Liberty and Discrimination. Oxford: Oxford University Press.

Dworkin, R. 2011. Justice For Hedgehogs. Cambridge, MA: Harvard University Press.

Fiss, O. M. 1976. Groups and the Equal Protection Clause. Philosophy and Public Affairs 5:2: 107-177.

International Covenant on Civil and Political Rights, G.A. res. 2200A (XXI), 21 U.N. GAOR Supp. (No. 16) at 52, U.N. Doc. A/6316 (1966), 999 U.N.T.S. 171, entered into force Mar. 23, 1976.

Scanlon, T. 2008. Moral Dimensions, Cambridge, MA: Harvard University Press.