Microtargeting, Dogwhistles, and Deliberative Democracy

‘Dogwhistles’ and microtargeted political advertisements are objects of widespread moral and political concern. With a few notable exceptions in the case of dogwhistles (and none in the case of microtargeting) moral criticism of these speech act types generally focuses on problematic content —that a dogwhistle is, for instance, racist, or a microtargeted advertisement misleading. I argue that these practices are additionally morally wrongful on content-neutral grounds—regardless of their content. My argument proceeds from a deliberative conception of democracy according to which only a vote which follows from an adequate deliberative process confers democratic legitimacy on its results. I claim that both dogwhistles and microtargeting threaten to prevent adequate democratic deliberation from taking place, and therefore that these practices are anti-democratic and, additionally, morally impermissible. I then discuss potential objections and how my argument relates to existing work.


Introduction
Politicians and journalists in the United States and other democracies have voiced concern about microtargeted political advertisements, most notably associated with Facebook, and so-called dogwhistles: political speech acts which communicate a 'surface-level' meaning to some hearers and a 'hidden' meaning to others. With two notable philosophical exceptions, 1 most normative criticism of these two practices focuses on their problematic content, for example that the advertisement or dogwhistle is misleading or racist. 2 I offer a unified and content-neutral account of why these practices are objectionable; dogwhistles and microtargeting are both anti-democratic, and, interestingly, they are anti-democratic for the same reason. But my critique is ultimately moral in nature. I argue that these practices violate norms of political morality-the moral norms that govern how political actors ought to act.
The paper is divided into five sections. Section 2 provides a brief outline of dogwhistles and microtargeting in the political context. Section 3 distinguishes two possible theses concerning the features which make such practices morally wrongful. 3 I call these Content-Based and Content-Neutral. In Section 4, I argue that these practices are wrongful on content-neutral grounds. My argument proceeds from the deliberative conception of democracy associated with, among others, Rawls (1971), Cohen (1989Cohen ( , 2007 and Fishkin (1991Fishkin ( , 1992. According to this view, only a vote which follows from an adequate deliberative process confers democratic legitimacy on its results. Hence, democratic processes which lack an adequate deliberative process are not fully democratic, in the normative sense. Roughly, my argument is that microtargeted political advertisements restrict voters' access to salient political information (i.e., politicians' campaign promises/claims). Dogwhistles likewise deny the democratic body as a whole access to salient political information. This balkanisation of salient political information undermines public scrutiny and society's capacity for democratic deliberation. Therefore, (given a deliberative conception of democracy) these practices are anti-democratic and, I further argue, immoral for political actors to perform-they violate plausible norms of political morality.
Finally, in Section 5, I briefly reply to two possible objections to my account and briefly distinguish it from existing philosophical work.

Dogwhistles and Microtargeting
'Dogwhistles' and 'microtargeting' are relatively new (or newly popular) terms, so it is worth briefly explaining what they mean and their political usages.

Dogwhistles
I will use the term 'dogwhistles' to refer to political dogwhistles, which I understand as speech acts performed by political actors aimed at public uptake, the (intended) function of which is to convey a surface-level meaning to the general audience and another hidden meaning to a subset of listeners. 4 At least, this is how standard 'covert' dogwhistles function (Saul 2018, p. 361-2). 5 Saul argues that while standard dogwhistles may be overt, dogwhistles can also be covert. In other words, the 'hidden' meaning may be recognised by 'insiders' as a hidden meaning in the overt case. In the covert case, the dogwhistle may not be consciously recognised by 'insiders' as a hidden message but still influence them. In terms of Saul's distinction, my analysis certainly extends only to overt dogwhistles. I am less confident it could be extended to covert dogwhistles since it is unclear if political information is being conveyed in the covert case. Saul also argues that dogwhistles can be both intentional and unintentional. I think that my analysis concerns both sides of this distinction.
Let's take an example of a dogwhistle: Martha's family values. Martha, an Australian politician, declares her support for 'family values.' 6 Martha knows two things. The first is that most voters will interpret her as endorsing broadly family-friendly policies, perhaps support for childcare or well-funded public education. The second is that most Evangelical Christians will interpret her as endorsing pro-life and anti-gay marriage positions.
Martha's act of speech exploits the fact that Australian Evangelicals are likely to have heard the phrase 'family values' in churches or in conservative media. In these contexts, the phrase typically denotes pro-life (anti-abortion) and antigay marriage views. Non-Evangelical Australians, lacking the same level of exposure to this usage, are, Lohrey claims (2006, p. 52), more likely to take the phrase more literally, at face value. Linguistically, a dogwhistle can be achieved various ways, but these interesting details are not important for present purposes. 7 Dogwhistles can fail. 8 While the ideal dogwhistle would be undetectable to the general audience and crystal clear to insiders, things do not always go as planned. The hidden message can be too well hidden for it to get picked up on by the insiders. Or the hidden message can be too easily heard by the general audience, potentially resulting in social and political costs for the speaker. When successful, though, dogwhistles allow their users to effectively split their audience in two and to say different things to each part. The political utility of this linguistic trick should be clear. Martha can gain the support of pro-life, anti-gay marriage voters without alienating pro-choice, pro-gay marriage voters.

Microtargeting
Online political microtargeting (henceforth 'microtargeting') is the practice of collecting personal information about internet users and using it to show them highly targeted political advertisements. Using the collected data, algorithms sort individuals using extremely fine-grained demographic markers, for example: 'targeting fathers aged 35-44 in Texas who frequent gun enthusiast websites.' (Borgesius et al. 2018, p. 83) Political content which is expected to be highly engaging 4 The dogwhistles we generally talk about are political and have objectionable hidden content. While it could be argued that these are essential, constitutive features of dogwhistles, I am assuming what I consider the 'orthodox' account of dogwhistles which claims that even non-political speech-acts with unobjectionable hidden messages can be dogwhistles provided they have the right structure in the ways mentioned. On such an account, even the hidden messages intended for parents in children's' cartoons are dogwhistles (see Witten 2014, p. 2, andSaul 2018, p. 362). Thank you to an anonymous reviewer for helping me clarify this point. 5 An additional complication to this way of demarcating dogwhistles from other speech-act types is that it is it may not always be easy to distinguish 'hidden' content from content which can simply be interpreted in different ways. While I am unable to provide an exploration of these subtilties here, I thank an anonymous reviewer for raising the point.
to members of these demographic slices are then delivered only to that tiny demographic slice.
Targeted advertisements are nothing new, of course. If you are selling snorkels, you would be well advised to advertise them to residents of Hawaii rather than Mongolia. The same goes in the political sphere. A political advertisement might be placed in a newspaper read by members of a particular demographic. It might be specially designed to appeal to that demographic. So, what, if anything, is new or different about microtargeting?
First, microtargeted advertisements are far more precisely targeted than traditional advertisements. According to ProPublica, Facebook has at least 52,235 categories to sort users into demographic slices-and these are merely the categories that they allow marketers to view (Angwin et al. 2016). Secondly, this wealth of categories enables extremely tailored messaging. As I write this, Facebook offers to send highly tailored advertisements to demographic microsegments of as small as 100 people (Hern 2019).
The microtargeting industry makes no secret of microtargeting's benefit: higher engagement per dollar spent. Delivering advertisements to fewer people is cheaper and a highly tailored message influences voters more. Andrew Bleeker, who ran Hillary Clinton's digital advertising in 2016, claims, 'The benefits of digital advertisements is being able to give people information that is relevant to them. There's no ill intent. That's just how it works' (Lapowsky 2017). Besides greater cost-effectiveness, there is less openly discussed benefit of microtargeting: a lack of public accountability for the advertisement's political message. Julia Carrie Wong, writing in The Guardian, discusses this benefit: Any candidate using Facebook can put a campaign message promising one thing in front of one group of voters while simultaneously running an ad with a completely opposite message in front of a different group of voters. The ads themselves are not posted anywhere for the general public to see (this is what's known as "dark advertising"), and chances are, no one will ever be the wiser. (Wong 2018) Radically segmenting one's audience makes it possible to deliver a wide range of highly specific, even contradictory messages, without any one individual encountering messages that are not tailored to them. It is also extremely unlikely that viewers share the advertisements they are shown with their wider community. On the day of the third presidential debate in 2016, the Trump campaign ran what Gary Coby, director of advertising at The Republican National Committee called 'A/B testing on steroids'-175,000 variations of its advertisements to find which were most effective for each demographic slice of the voting public (Lapowsky 2016). Each variation was shown to as few as one hundred Facebook users, with the tens of thousands of other variations remaining invisible to them.

Two Ways to be Wrong
In this section I distinguish two ways that dogwhistles and microtargeting can be wrongful: based on their content and regardless of their content-the latter being my focus.

Content-Based Wrongfulness
In 1981, former Republican Party strategist Lee Atwater said: You start out in 1954 by saying, 'N****r, n****r, n****r.' By 1968 you can't say 'n****r'-that hurts you, backfires. So you say stuff like, uh, forced busing, states' rights, and all that stuff, and you're getting so abstract. Now, you're talking about cutting taxes, and all these things you're talking about are totally economic things and a by-product of them is, blacks get hurt worse than whites… 'We want to cut this,' is much more abstract than even the busing thing, uh, and a hell of a lot more abstract than 'N****r, n****r.' (Lee Atwater as quoted in Perlstein 2012) Imagine a politician performs the dogwhistle Atwater describes: 'We want to cut this.' If it succeeds, then racist voters understand the politician to have signalled support for racist views, and non-racist voters understand the politician as having signalled support for fiscal conservatism. 9 What is objectionable about the dogwhistle, in this case, is that its hidden content is racist. That is just what we mean when we call a speech act a racist dogwhistle.
Objectionable content also seems to explain what makes microtargeting morally objectionable. In 2019, the UK's Conservative Party purchased a targeted Facebook advertisement which showed a BBC News headline reading '£14 billion pound cash boost for schools.' 10 Fullfact.org, a factchecking organisation, found that the linked BBC article was headlined 'School spending: Multi-billion pound cash boost announced'. The BBC confirmed that their headline 'never referred to the £14 billion.' Moy (2019), writing for Full Fact, rightly criticised this as misleading those who saw the advert into thinking that the BBC had endorsed the claim. In this case, the advertisement is objectionable, at least in part, because of its misleading content.
Both these cases are objectionable in virtue of their content-that they were racist and misleading, respectively. Given the abundance of such cases in the media and the ongoing tendency in the literature to 'focus on content' (Saul 2018, p. 361), one might be tempted think that the wrongfulness of these political speech-act types is fully explained by the objectionableness of their content. Objectionable content is, one might think, a necessary condition for a wrongful dogwhistle or wrongful instance of microtargeting. Call this view Content-Based.
Content-Based has some initial plausibility. It is at least clear that dogwhistles and microtargeting can be wrongful in virtue of their content; if we remove the objectionable content from the examples above, the wrongfulness of the dogwhistle and microtargeted advertisement seems to disappear with it. If the Conservative Party's advertisement had simply shown the genuine BBC headline, then there is no longer anything obviously wrong with it. If Atwater's politician had instead said a dogwhistle whose hidden content was anti-racist, it would no longer be a racist dogwhistle, and it would no longer be wrong on that basis. Furthermore, microtargeting and dogwhistles, as I have considered them, are political speech acts. It may not be immediately obvious how anything besides a speech act's content could make an act of speech wrong.
But Content-Based is false. It claims the only way a speech-act can be wrongful is due to its content. But, as well as the clear moral significant of content, there are other wrong-making features of speech acts to consider. I will turn to these now.

Content-Neutral Wrongfulness
Both dogwhistles and microtargeting are rightful targets of moral criticism in virtue of any objectionable content they might have. Whenever these practices are subject to moral critique it is, with few exceptions in the case of dogwhistles, 11 on these kinds of content-based grounds. I do not wish to dispute any of that. Another question remains unanswered; can these practices additionally be content-neutrally wrongful, that is, morally impermissible regardless of their content. I shall argue that they can be, and, much more interestingly and substantively, that they generally are.
The first thesis I will defend is simply the negation of Content-Based.

Content-Neutral (Conceptual)
Political dogwhistles and instances of microtargeting can be wrongful even in the absence of wrongful content.
The case for this thesis is straightforward. We commonly morally evaluate instances of speech based on things other than the speech's content. For instance, we might condemn an utterance for its illocutionary effects-the action that is done with the utterance. Imagine a king orders an unjust invasion of a neighbouring country by uttering 'It is time.' Here, we might criticise the utterance not based on its content, but (also) on the basis of its constituting an immoral order given the king's institutional authority. Most notably though, we also evaluate the perlocutionary effects of speech acts, and other downstream consequences, when determining its ethical permissibility. 12 There's nothing objectionable about the content of 'It is time.' Nonetheless, it is wrong of the king to utter 'It is time' (at least if he is aware that it will constitute ordering an unjust war). We condemn this speech act based on its content-neutral harm-that is, harm which is not grounded in its content. So, Content-Based is false, and its negation, Content-Neutral (Conceptual) is true.
I will ethically evaluate dogwhistles and microtargeting in the same way-content-neutrally. Both the speech acts they constitute and their consequences could, at least potentially, be content-neutrally harmful (that is, harmful even bracketing the objectionableness of their content) and wrongful on that basis. More importantly, I shall argue that they generally are immoral for political actors to perform because they damage democratic deliberation via information balkanisation.

Deliberative Democracy and Content-Neutral Wrongfulness
Content-Neutral (Conceptual), though true, is silent on the more substantial and interesting question of whether dogwhistles and microtargeting actually are morally impermissible. I defend this stronger claim in this section. Here it is stated more precisely:

Content-Neutral (Substantial)
Political dogwhistles and instances of microtargeting are morally impermissible, other things being equal, regardless of their content.

Deliberative Democracy
As stated previously, my argument proceeds from the deliberative conception of democracy. In this subsection, I will briefly explain this. In the following subsection I will explain why dogwhistles and microtargeting violate deliberative democratic norms. An important distinction in the philosophy of democracy is drawn between (1) 'vote-centric' or 'aggregative' and (2) 'talk-centric' or 'deliberative' theories of democracy. 13 According to the former, oft-discussed but almost entirely undefended, democracy is to be understood almost entirely in terms of voting. This conception of democracy has its roots in Rousseau's idea that the results of democratic processes reflects a 'general will'. A free and fair election or referendum is, according to this conception, a necessary and sufficient condition for its result to possess democratic legitimacy-a democratic mandate. The function of voting, according to this view, is to translate, or aggregate, pre-existing voter preferences into binding decisions. On such a view, 'the requirement [on democratic processes] of deliberation can be an invitation to hypocrisy and deceit' (Sunstein 1987(Sunstein , p. 1545. 14 Critics of the vote-centric, aggregative model contend that this conception of democracy cannot explain the bindingness, the normativity, of the democratic mandates which follow from elections and, indeed, from other democratic processes. 15 Nothing in the vote-centric conception of democratic legitimacy requires that voters (and other political actors) get the opportunity to persuade each other, to jointly evaluate the strength of their claims, nor to voice their concerns. Nor does it demand that politicians' claims are held up to scrutiny, nor that minorities' interests are heard out by the majority, nor that claims are made on a reasonable basis rather than on prejudice or ignorance. For all these reasons, its critics regard the vote-centric conception of democracy as impoverished.
Popular elections are not necessarily the exemplar democratic procedure, according to some deliberative democrats, and might ideally be supplemented with citizen-led deliberative bodies such as citizen assemblies. However, deliberative democrats accept that elections are one form of democratic procedure, capable of generating (or failing to generate) democratic mandates. So, for the sake of clarity and simplicity, my discussion will focus on the norms and dynamics which exist around elections (though what I have to say is, I think, largely transferable to the norms and behaviour surrounding other democratic procedures and their preceding deliberative phases).
My aim in this paper is not to defend the deliberative conception of democracy. I will simply assume it. I will assume the deliberative features of well-functioning democracies are legitimacy-generating parts of democratic procedures and that they play some role in making the losers of democratic processes (of e.g., elections) normatively bound to conceding the legitimacy of its result. The losers got their hearing; their concerns were considered and did not win the day. I will assume that, if the voting body is denied enough of these deliberative opportunities, then the results of the vote would not be fully legitimate and binding for losers. I will assume Kymlicka is onto something when he writes that: citizens will accept the legitimacy of collective decisions that go against them, but only if they think their arguments and reasons have been given a fair hearing, and that others have taken seriously what they have to say. But if there is no room for such a fair hearing, then people will question the legitimacy of decisions. (Kymlicka 2002, p. 291) To be precise, the deliberative democrat's claim is not simply that voters will not accept the legitimacy of votes carried out on this basis, which is an empirical, sociological claim. The claim is a normative, political one that they should not, that the results of these votes do not possess democratic legitimacy. Additionally, the deliberative democrat may not think of elections as the most important or most legitimacy-conferring democratic procedure there could be.
So, what requirements does the deliberative democratic theorist place on democratic deliberation? There are competing detailed responses to this question, but we can work with the following generic answer:

Deliberative Democracy. A voting process confers democratic legitimacy on its results only if the voting body is provided sufficient opportunity to engage in public deliberation of an adequate standard.
What is distinctive about the deliberative conception of democracy is not that it offers a completely novel or distinct account of what democratic legitimacy requires. Deliberative democratic theorists merely claim that, regardless of whatever else might be involved in generating democratic legitimacy, it also involves some amount of public 13 I draw this distinction along the lines suggested by Kymlicka (2002, p. 290). 14 Sunstein himself does not endorse the aggregative conception of democracy, although he has some sympathies. 15 For important critiques of the aggregative model and defences of the deliberative conception of democracy, see Rawls (1971), Cohen (1989Cohen ( , 2007, and Fishkin (1991Fishkin ( , 1992. deliberation (or perhaps merely a sufficient opportunity for such deliberation) prior to voting taking place. 16

Democratic Deliberation and Publicity
Deliberative democrats list differing requirements on legitimacy-conferring democratic deliberation. My focus will be on a much-included and little-discussed item on those lists: publicity. Publicity, in this context, simply means the extent to which salient political information is accessible to the public as a whole. 17 A piece of information enjoys greater publicity if it is easily accessible to more people, and lesser publicity the greater the barriers to access.
It has long been argued that a certain degree of publicity is a fundamental requirement of democracy, or at least of a good democracy. In 1840, Thomas Carlyle wrote of the essential democratic role of publicity, enabled by the 'fourth estate' of the Reporters' Gallery in Parliament. This pillar of democracy meant that 'the parliamentary debate goes on now, everywhere and at all times, in a far more comprehensive way, out of Parliament altogether' (Carlyle 1993, Lecture 5). More recently, deliberative democrats have identified even closer conceptual ties between democratic legitimacy and publicity. Amy Gutmann and Dennis Thompson claim that publicity is one of the 'chief standards regulating the conditions of deliberation' (Gutmann and Thomson 2009, p. 133) while Parkinson describes publicity as 'the essence of deliberative democracy' because it is the: procedural foundation, the means by which information is brought into a deliberative moment and by which claims and counterclaims are weighed and sorted; and it is its ethical foundation, the yardstick by which one judges the rightness or wrongness of political action, and one of the means by which the powerful are restrained. (Parkinson 2006, pp. 99-100) In other words, the publicity of relevant information is a requirement for democratic deliberation, scrutiny, and accountability. In Habermas's words, publicity has a dual democratic function of 'monitoring' and 'feeding' democratic deliberation (Habermas 1996, p. 171). For all these reasons, the deliberative democrat holds that without sufficient publicity, adequate democratic deliberation is impossible. Without adequate publicity, there can be no public scrutiny: no chance for correction or elimination of repugnant or self-interested claims from the debate, no chance for opposing views to be 'heard out.'

Information Balkanisation and Publicity
The simplest way to violate publicity is for political actors to keep salient political information secret: out of the hands of the people. But I want to look at a more underhand method, a way of violating publicity by preventing the free flow of information. I will call this underhand mechanism information balkanisation. Imagine the following society.

Silent Democracy.
A nation seemingly has all the ordinary features of a democracy. There are regular elections without fraud, elected representatives, a universal franchise, etc. But this society has taken the polite norm against discussing politics and enshrined it into law. This law is enforced by flawless monitoring and harsh punishment. As a result, no one discusses any politically relevant information that could foreseeably influence how someone else might vote. The one exception to the law is this. Candidates running for office are permitted to send one leaflet to each voter before each election, making their case. The candidates are, naturally, forbidden from seeing or discussing one another's leaflets-doing so would be considered extremely rude, not to mention illegal. Given that so much rides on these leaflets, the candidates make use of large databases and are able to create a customised, individual message for each voter based on what they believe the voter will care most about.
What kind of a political information ecosystem would a society like Silent Democracy generate? What implications would there be for the democratic legitimacy of an ensuing democratic process, such as an election? A stark deficit of publicity is predictable in such a society. As such, there would be no incentive for political actors to provide voters with coherent or consistent messaging. Political information would be restricted from flowing, by law, beyond the confines of individual voters' minds. It would be prevented from spilling into the light of more general public awareness.
Even stipulating that the political messages in Silent Democracy's leaflets are not false or misleading, politicians are still carefully curating the information they provide (i.e. choosing not to inform each citizen completely so they cannot make an optimal choice about how they will vote and what information they will share with others). Even this seems to problematically distort the flow of political information. 18 Even in this more benign version of Silent Democracy in which there is no false or misleading messaging, citizens are unable to cooperate, to help each other develop their understandings and deliberative stances in the light of information they receive from one another. They cannot escape or supplement their limited pools of information. They are unable to update their beliefs and preferences in light of what they learn from their fellow citizens' responses to politicians' claims, proposals, and policies. In an ordinary democracy, our democratic deliberation is profoundly socially extended, in that much of the deliberation we rely on occurs in the minds of others and is shared with us via communication. 19 In Silent Democracy, individual citizens lose the ability to leverage this extended deliberative capacity, and hence their deliberative capacities are correspondingly diminished. In sum, the laws of Silent Democracy seem likely to result in information balkanisation and this is problematic, from the deliberative democrat's perspective, in three ways. First, the laws of Silent Democracy strongly incentivise inconsistent messaging. Second, they prevent citizens from operating with adequate information bases. Third, they prevent citizens from jointly transforming their beliefs and preferences in the light of what they can learn from others.
In the context of an election, the information ecosystem in Silent Democracy would likely result in something like the following.

Balkanised
Voters. An election takes place in which each voter has access to completely non-overlapping sets of information. Peter the Politician tells each voter about a different set of policies, endorses a different set of values, etc. Anna has access to information set 1. Bob has access to information set 2, which has no overlap with 1. Clara has access to information set 3, and so on. The voters are allowed and even encouraged to deliberate privately, but they are forbidden from communicating with one another to share their information. Peter wins the election based on the (different) information he gives to each voter.
Does Peter winning this way generate a democratic mandate for Peter's rule and policies? My intuition is that it does not. 20 Even if each voter is free to deliberate about all the information that they have access to, there is still something important missing from this democratic process. The radical process of information segmentation in Balkanised Voters prevents sufficient publicity of salient information, resulting in an election result which intuitively fails to confer democratic legitimacy on its results. This is particularly clear if the politician has made inconsistent or mutually unrealisable promises. If Peter has promised to both increase and decrease spending on policing, he has no democratic mandate to do either. But even if Peter has avoided committing to mutually exclusive goals, voters could rightfully complain that none of them voted for his policies as a package since their understandings of the Peter's proposals were so highly curated and they lacked opportunities for information sharing and deliberating together. For all these reasons, for deliberation to confer legitimacy on the results of democratic processes, salient information must be sufficiently public.
What would 'sufficient publicity' look like? Even in the ideal case, it is unrealistic to demand that every single voter has access to all salient political information. But there will, in the good case, be enough overlaps and sufficiently open lines of communication such that all salient political information is somewhat accessible to the public as a whole. The picture I have in mind is something like this. Anna may have access to information sets 1-2, Bob to information sets 2-3, and Clara to information sets 3-4. Open, public discussion of the information that is shared between at least two voters will then make it more accessible to the public. This process is both complicated and aided in real life by the existence of various public forms of media, including both news and social media, which (ideally) discover and share political information. Whatever the amount of publicity required for democratic legitimacy, it should be clear that this threshold is not met in Silent Democracy or Balkanised Voters.

Dogwhistles, Microtargeting, and Information Balkanisation
With all those conceptual pieces of the puzzle now in place, let us now turn to the main substantive thesis of this paper: Content-Neutral (Substantial). My claim is that dogwhistles and microtargeting will tend to result in information balkanisation. They thereby 18 Thank you to a reviewer for suggesting many of these ideas. 19 I have in mind the social forms of deliberative extension suggested by extended theories of mind and distributed cognition, according to which, 'The emphasis on finding and describing "knowledge structures" that are somewhere "inside" the individual encourages us to overlook the fact that human cognition is always situated in a complex sociocultural world and cannot be unaffected by it' (Hutchin 1995, xii). The idea is that much of an individual's cognition is partially distributed socially, in the minds of others with whom we are informationally connected via communication. See also Clark and Chalmers (1998). 20 C.f. Goodin and Saward (2005). They discuss a similar case to Balkanised Voters but approach the question from the aggregative democratic tradition. I understand their paper as claiming that Peter would have a democratic 'mandate to rule' but no 'policy mandate'. The deliberative democrat would, I think, prefer to say that Peter has won neither kind of democratic mandate. I discuss their arguments in more detail later. undermine publicity, reduce public accountability, and prevent adequate democratic deliberation from taking place. These practices thereby threaten the democratic legitimacy of election results and the existence of democratic mandates (assuming the deliberative conception of democracy). It is appropriate to condemn such practices as anti-democratic and judge it morally impermissible (other things being equal) on that basis. 21 Let us briefly run through these argumentative steps.
A dogwhistle will predictably result in information balkanisation as follows. A dogwhistle creates two groups, each of which has access to different information. The 'insiders' receive the 'hidden' message, and the general audience receives the 'surface-level' message. The function of the dogwhistle is to keep the content of the hidden message away from a segment of the voting body-to deny them access to certain salient information. This is a clear example of what I have called information balkanisation. Indeed, a dogwhistle's very function, as a speech-act type, can be understood as generating desirable (from the perspective of the speaker) information balkanisation. This is why dogwhistles are intentionally performed. Given that dogwhistles cause information balkanisation and thereby undermine publicity, an election campaign that engages too heavily in dogwhistles could end up failing to produce a legitimate democratic mandate even after (procedurally) winning an election.
Microtargeting also predicably results in severe information balkanisation. By shifting political messaging from large-circulation media such as television, newspapers, billboards, radio etc. to 'micro segments' of the population, as small as one hundred, on privately viewed social media, microtargeted political advertisement campaigns tend to deny the public as a whole access to the information which is influencing voters' political choices. Microtargeted advertisements are, by their nature, invisible to the public as a whole. And, as the deliberative democrat points out, this lack of publicity results in a lack of opportunity for public scrutiny and accountability-key features of an adequate, legitimacy-conferring process of democratic deliberation.
A scene from the fictionalised drama 'Brexit: The Uncivil War' (2019: 1:12:54) captures the dynamic I have in mind. After launching a large microtargeted political advertising campaign on Facebook, the fictionalised Dominic Cummings character asks his chief data analyst 'Why aren't we getting any pushback from this? Why aren't journalists hounding us about it?' The data analyst replies, 'The metropolitan commentariat are not our targets, so they're not seeing our posts on their timeline, so they have no idea what the rest of the country is seeing. So, no-one's reporting it.' Although fictional, this scene showcases the predicable effects of microtargeting on the political information ecosystem: information balkanisation and a resulting lack of publicity. It also conveys the resultant lack of public scrutiny.
In sum, both dogwhistles and microtargeting violate the democratic norm of publicity. I have also proposed that the mechanism through which they violate publicity is information balkanisation. If these practices do violate publicity, it follows from the deliberative democratic view that these practices are anti-democratic.
Let me clarify the nature and scope of the argument made so far. 22 I am not claiming that it is a conceptual or necessary truth that dogwhistles or microtargeting result in information balkanisation and are therefore anti-democratic. The pernicious outcomes of such practices are, I think, highly predictable for the reasons I have just outlined, but they are ultimately a contingent, empirical matter.
How could these practices fail to result in information balkanisation? A political microtargeted advertising campaign may provide highly tailored messaging but avoid providing inconsistent messaging. It could even utilise the power of microtargeting and tailored messaging to provide voters with information to which they were likely lacking prior accessthereby reducing existing information balkanisation. The critique of microtargeting provided here merely concerns a proper subset of instances of microtargeting-those that result, or would predictably result, in information balkanisation. For the reasons already provided and based on the aforementioned anecdotal evidence mentioned (e.g., Wong 2018), I suspect that this proper subset represents the vast majority of actual cases and therefore licences the generic moral claim that microtargeting is anti-democratic in the way I describe.
Likewise, a small subset of dogwhistles could, as an empirical matter, fail to result in information balkanisation. I am thinking of infelicitous dogwhistles, e.g., dogwhistles whose hidden messages are detected and punished or those in which the hidden message remains too obscure for insiders to pick it up. Once again, the kind of dogwhistles this argument condemns as anti-democratic is merely those that actually result, or would predictably result, in information balkanisation.
A slight disanalogy between dogwhistles and microtargeting emerges here. Dogwhistles, unlike instances of microtargeting, as a conceptual matter cause information balkanisation insofar as they are successful instances of their speech-act type. So, we can say (1) that successful dogwhistles necessarily result in information balkanisation, to some degree, and (2) that insofar as a political actor intentionally performs a dogwhistle, the intention is necessarily to perform a speech-act which results in what I have called information balkanisation. Indeed, this is the very reason politicians use dogwhistles. Hence, insofar as we morally condemn political actors for intending to perform actions that have pernicious effects (rather than for the actual outcomes of their actions) it seems prima facie appropriate to condemn political actors for intentionally performing dogwhistles based on the wrongness of intentionally causing information balkanisation. 23 These clarifications made, it is time for the final step in my argument, which moves from the anti-democratic nature of these practices to the moral impermissibility of political actors performing them. 24 I have already mentioned the wrongness of performing these speech-act types. But according to which norms are these practices wrongful? What kinds of norms do political actors (typically) violate when they engage in microtargeting or dogwhistling? Conversational norms? Democratic-deliberative norms? Moral norms?
The conversational, political, and moral normative domains are tightly interwoven and often overlap significantly. 25 A politician who falsely promises or intentionally lies to the public has, plausibly, violated conversational norms which demand that one ought to keep one's promises, or, more broadly, that one ought to do what one says, or only assert what is true, sincere, or justified. The politician has violated democratic-deliberative norms, in abusing public trust. The politician has also violated moral norms against making false promises and lying. A similar mixture of closely-related norm-violations follow, I suspect, from dogwhistles and microtargeting.
My claim is that political actors in a democratic context generally violate moral norms when they violate democratic norms. More precisely, these are norms of political morality-the norms that govern how political actors ought to act-when they engage in dogwhistles and microtargeting. I am making a claim about the morality of political actors qua political actors, and their methods for practicing politics.
I have already provided reasons to think that dogwhistles and microtargeting are anti-democratic practices. But why does an act-type being anti-democratic make it wrong for a political actor (in the context of a democracy) to perform? This might seem like a strange question, but it is plausible that the moral principles which govern political behaviour differ from those that govern ordinary life (Hampshire 1978;Thompson 1987). A Machiavellian about political morality, someone who thinks that political actors ought only to pursue and hold power, would not condemn a political actor for acting anti-democratically (so long as power is achieved and retained that way). But I suspect few would be content to adopt such a radical view of the norms of political morality. A more orthodox view is that, as Sabl puts it, 'the [politicalmoral] test for political officers is not only whether they cause the right outcomes but whether their actions stand to make democratic politics as a whole better or worse [… whether they] help or harm the democratic polity through his or her actions' (Sabl 2002, p. 5, my emphasis). At the very least, it seems highly plausible that one important norm of political morality is that political actors practicing politics in democracies ought not to undermine or damage the democratic polity.
I conclude that a practice being anti-democratic makes its performance generally wrongful-morally impermissible, all things being equal. Hence, since both dogwhistles and microtargeting, insofar as they predicably result in information balkanisation, are anti-democratic practices. So, Content-Neutral (Substantive) is true.

Objections and Contextualisation
I will now respond to two potential objections and contextualise this paper in relation to existing work on these topics. The objections are that; microtargeting is nothing special compared with traditional advertising and my argument 23 I specify that a wrongful dogwhistle is most plausibly intentional because, as Saul notes, it also seems possible for harmful dogwhistles to be performed unintentionally via the 'Unwitting use of words and/ or images that, used intentionally, constitute an intentional dogwhistle, where this use has the same effect as an unintentional dogwhistle.' (Saul 2018, p. 368). 24 Thanks to an anonymous reviewer for pushing me to develop this important step in my argument. 25 A fuller utilisation of the resources of argumentation theory than I can provide here would, I suspect, substantially supplement this cursory account of the conversational and argumentative norms and rules that might be violated by microtargeting (and dogwhistles). According to Frans van Eemeren's work on strategic manoeuvring, while some 'adapting to audience demand and particular ways of making topical choices' (Eemeren 2010, p. 272) can be a legitimate rhetorical framing of one's argument in a way that is likely to be compelling to one's audience, such adaptations to audience can also go too far and derail the dialectical side of argumentation, the side of argumentation which aims to resolve disagreements on the merits of the case. In such cases, adaptation to audience can become increasingly argumentatively fallacious, suggesting that microtargeting may violate certain conversational/argumentative norms due to over-audience-adaptation. I sincerely thank an anonymous reviewer for suggesting this promising avenue for future development of the arguments in this paper. proves too much, namely that it also condemns ordinary private political conversations.

Two Reductio ab Absurdum Objections
Perhaps there is nothing special about microtargeting. In a sense, the differences between microtargeting and oldfashioned targeting are merely in degree rather than in kind. Advertising placed in traditional media is also targeted based on demographic segments to raise the efficacy of advertising dollars. And we do not condemn ordinary targeting of political advertisements. So, the objection goes, to be logically consistent I either need to bite the bullet and condemn what intuitively is permissible, traditional political advertisements, or abandon my claim that microtargeting is morally objectionable. The objection has the form of a reductio ab absurdum.
But there is an important political-moral asymmetry between engaging in traditional and microtargeted political advertising. A traditional targeted political advertisement in a left-wing newspaper will mostly be seen by left-wing voters, but it will still be accessible to right-wing voters and media. So the placement of such an advertisement is plausibly net increasing the publicity of the information conveyed in the advertisement. It is offering that information for consideration and scrutiny by the public as a whole. Journalists, for example, reporting on the coverage of newspapers and on the campaigns of the politicians, will likely discover the messages in traditional targeted advertisements like this. Microtargeting is very different. Because it seems to generate such strong information balkanisation, it systematically undermines publicity with no compensatory mechanism of discovery and scrutiny, unlike traditional advertising. It is this failure, and the resulting damage inflicted on the quality of democratic deliberation, that makes the practice antidemocratic, and it is also this failure which distinguishes it, morally, from traditional political advertising. In sum, traditional targeted advertising may be morally dubious to the small extent it reduces publicity. But overall, it seems more likely increase the publicity of the information it conveys as it is sufficiently public for public scrutiny etc. to occur. Even if the trade-off does not quite work out in the traditional advertisement's favour, it will still be far less morally objectionable than a typical microtargeted political advertisement.
A second potential objection, like the last, takes the form of a reductio ab absurdum. It holds that my argument proves too much. If it is wrong to balkanise salient vote-influencing political information via dogwhistles or microtargeting, then it is equally wrong to have an ordinary dinner table political conversations amongst friends-this activity also problematically undermines publicity! We avoid this absurd conclusion by appealing to a distinction between public and private political speech and the differing normative standards of ordinary morality and political morality. Both dogwhistles and microtargeting, as they have been considered here, are forms of public political speech performed by politicians or their campaigns; they are attempts by those politicians and campaigns to influence public opinion, broadly understood, and the population's voting patterns. It seems both true and justified that we have different normative standards for these special forms of political speech performed by political actors and aimed at mass, public communication (i.e., the standards of political morality) than those we have for private political speech aimed primarily at interpersonal communication (i.e., the standards of ordinary morality). In this paper, I am only interested in the political-moral normative standards of these more public forms of speech performed by political actors in their capacities as political actors. I am not, then, forced to condemn political talk at the dinner table (though it may be objectionable for other reasons).

Adjacent Work
I want to briefly highlight what is novel about the argument provided in this paper and how it is distinct from some existing arguments.

Goodin and Saward's Account
The first existing argument worth briefly distinguishing from my own is Goodin and Saward's in 'Dog whistles and democratic mandates ' (2005). They argue that political actors may fail to obtain policy mandates if they perform dogwhistles. This can be understood as a content-neutral critique of political dogwhistles according to which using dogwhistles undermines their user's policy mandate by delivering an electoral win based on mixed messages.
Besides the obvious, that my account addresses microtargeting in addition to dogwhistles, there are two key differences between our accounts.
The first is that my argument is moral rather than political. Specifically, as mentioned before, my conclusions fall under the umbrella of 'political morality'-how moral agents ought to behave in political contexts. Goodin and Saward's paper is about a political question-under what circumstances do political actors enjoy the benefits of a policy mandate? This is important because, write Goodin and Saward, 'A policy mandate-where it exists-gives political actors a particularly strong hand.' (p. 473) 'Our concern here,' they explain, 'is with the ways in which dog whistle politics weaken that hand.' They conclude that 'Parties practising dog whistle politics might win a mandate to rule but doing so in a way that deprives them of the possibility of winning a policy mandate. ' (p. 476) In other words, their conclusion is practical, political, and disjunctive: either a politician can engage in dogwhistle politics, or he can enjoy the benefits of a policy mandate if he wins, including the 'particularly strong hand' it provides its possessor. Dogwhistle politics can, they ultimately conclude, be 'fundamentally counterproductive' for the politician who seeks a policy mandate.
Notice that nothing in this argument entails the distinctively moral thesis that my argument pursues, Content-Neutral (Substantial). Content-Neutral (Substantial) claims that political dogwhistles (and microtargeting) are, all things being equal, morally wrong. Goodin and Saward do not make this claim. Nor does such a claim obviously follow from what they say. Goodin and Saward seem to be offering politicians a strategic/political choice, rather than a moral one. 26 There is also a second crucial way in which the account offered here differs from Goodin and Saward's. They argue that dogwhistles can undermine democratic policy mandates, but that they cannot undermine the democratic mandate to rule. It might even be objected that my argument oversteps its bounds by claiming to show that dogwhistles and microtargeting are anti-democratic tout court, i.e., that these practices threaten the democratic mandate to rule.
Let me briefly explain. Goodin and Saward distinguish the ordinary democratic mandate from a policy mandate. The former is simply the right to rule-to legitimately hold a democratically elected office. The latter is an entitlement to have one's political opponents to defer, or at least not obstruct, the implementation of one's policies. To acquire a mandate to rule, they claim, one needs only fulfil the procedural requirements for being elected to power. The use of dogwhistles does not threaten the acquisition of a democratic mandate to rule. However, to acquire a democratic policy mandate requires more than this. Goodin and Saward do not explicitly state what they think a policy mandate requires, but they do suggest that there must be something like a 'collective intention' behind the 'joint action' of electing a representative with a policy (p. 474). Their claim is that those who engage in 'dog whistle politics undermine the possibility of a […] policy mandate.' (p. 472) Why? Because: Politicians engaging in dogwhistle politics are […] not telling everyone what specific policies they propose to implement if elected. Instead, they tell one group of voters one thing, while allowing (and indeed, encouraging) another group to do another. If they win the election on the basis of such mixed messages, what does their victory add up to in substantive policy terms? Nothing, we suggest. (p. 473) Goodin and Saward claim that the standard democratic mandate, on the other hand, -the mandate to rule-is not threatened using dogwhistles. I disagree. For all the same reasons the use of dogwhistles undermines a policy mandate, it would also seem to undermine standard democratic mandates. In other words, I agree with Saul's critique that 'Goodin and Saward's arguments do not go quite far enough. If they are right about the policy mandate, then the mandate to rule may also often be undermined.' (Saul 2018, p. 380).
Who is right here? First, consider why Goodin and Saward think dogwhistles undermine policy mandates. They provide an analogy in which California voters have two ballot proposals to vote yay or nay on. Proposition 1 would outlaw gay marriage, while Proposition 2 would outlaw abortion. The numbering of the propositions is garbled by a hacker. Some voters, on their voting screens, see Proposition 1 as outlawing abortion, while others see it as outlawing gay marriage. Imagine the State Electoral Commissioner, after having found the fault, declares that Proposition 1 has secured 52% of the results, and has carried, while Proposition 2 secured only 41% of the results, and has failed. 'What on earth are we supposed to make of this?' ask Goodin and Saward, 'Is gay marriage outlawed?' They conclude, and I agree, that there can be no policy mandate generated by such a vote. They also claim that a vote for a candidate or policy based on dogwhistles will have the same kind of structure. Because voters will be voting for the same candidates for radically different reasons, it will be unclear what voters really voted for, and it will therefore be impossible to identify any candidate that has a mandate to implement their policies.
But the same logic can be applied to the right to rule. Imagine that instead of propositions, the California voters were voting for candidates. The winning candidate uses dogwhistles, which allow him to present different versions of himself to different voters. The voters submit ballots in for the winner (or enough of them do), but they have voted for radically different candidates. Would the winner of this election really have the full democratic mandate to rule? The same that an honest non-dogwhistling politician would? I think not.
In sum, Goodin and Saward's content-neutral critique of dogwhistles is distinct from my own in being political and practical rather than moral in nature. I also claim that the use 26 At one point, Goodin and Saward claim that dogwhistles are 'morally tainted.' It is possible to read their paper as giving a reason for this; dogwhistle politics undermines a certain kind of democratic mandate-a policy mandate (2005, p. 476). The problem with this moralistic interpretation of their argument is that their claim that dogwhistles are 'morally tainted' is much more plausibly understood as a (dialectically impotent) assumption than as the intended conclusion of their argument. Furthermore, that dogwhistles undermine policy mandates would not make it morally wrong to perform them unless we assume that politicians have a general moral duty to generate policy mandates. Goodin and Saward never venture to make this substantial claim.
1 3 of dogwhistles undermines the general democratic mandate to rule. And, of course, our accounts differ is that my analysis extends the content-neutral critique to microtargeting.

Saul's Account
The second existing argument I want to briefly distinguish from my own is that provided by Saul (2018).
Saul's focus is developing a detailed taxonomy of the various forms that dogwhistles can take and how each can generate problematic perlocutionary effects (and how some of these may be countered). She also discusses ways in which dogwhistles could be considered anti-democratic which I do not. One might, for instance, think that dogwhistles and (some) microtargeted advertisements tend to bypass voters' rational scrutiny, that they might not be fully aware of the message they are receiving, or how their reactions are being primed and their prejudices activated by a dogwhistle. In such cases, dogwhistles may undermine democratic deliberation directly, in the mind of each individual by manipulating them, rather than via publicity suppression. Dogwhistles might also be performed unintentionally and yet have the same noxious perlocutionary effects as an intentional performance, which may be anti-democratic in some sense but would not, I think, be obviously immoral.
Saul's argument is plausibly read as providing primarily content-neutral normative critiques of the performance of political dogwhistles. Most significantly, as discussed briefly in 5.2.1., and most relevant to my own argument, Saul argues that dogwhistles, particularly covert dogwhistles, may undermine democratic mandates in a more radical way than Goodin and Saward suggest because of covert dogwhistles' greater manipulative potential (Saul 2018, pp. 379 − 380). 27 She claims the use of dogwhistles may undermine the dogwhistler's democratic right to rule, and not merely the entitlement to a policy mandate.
The arguments provided in this paper are, I think, largely supplementary to Saul's. Regarding the ways in which dogwhistles undermine democratic norms, offer an additional and significant way in which these practices can be considered anti-democratic. But the argument provided here is distinctive from Saul's in (a) providing a new way in which dogwhistles can be considered anti-democratic (via undermining the deliberative-democratic norms of publicity), (b) describing a novel mechanism through which publicity can be violated (information balkanisation), and (c) extending the analysis to the parallel threat of microtargeting. The other significant difference between our accounts is that Saul's content-neutral critiques of dogwhistles are most plausibly read as ultimately appealing to political and democratic norms rather than the norms of political morality that I invoke.

Summary and Conclusion
I have argued that Content-Neutral (Substantive) is true; political dogwhistles and microtargeting are unethical practices, other things being equal, regardless of their content (though they are very commonly additionally wrongful because of their content). My argument for that conclusion proceeds from a deliberative conception of democracy, according to which the legitimacy of democratic processes depends on sufficient democratic deliberation of an adequate standard taking place.
As Silent Democracy and Balkanised Voters jointly demonstrate, insufficient publicity of vote-influencing information predicably impairs and degrades democratic deliberation. Both dogwhistles and microtargeting undermine democratic deliberation via information balkanisationthe prevention of the flow of salient political information between individual voters and demographic groups. These practices thereby prevent public access to that information. They are, therefore, anti-democratic. I additionally conclude on this basis that they are morally wrongful practices according to plausible principles of political morality.
The argument provided here supplements the broader literature on these issues in four important ways. Firstly, it expands the arsenal of the content-neutral critiques of dogwhistles developed by Saul, Goodin, and Saward. It does so by identifying a novel way that dogwhistles can be considered anti-democratic (namely by undermining publicity). Secondly, it also identifies a new mechanism for undermining publicity-information balkanisation. Thirdly, it expands the content-neutral critique of dogwhistles to the parallel threat of microtargeting. Fourthly, it links these issues to the norms of political morality to provide a distinctively moral content-neutral critique of political actors' performance of these practices.
Does the analysis provided here have any relevance to policy or the duties of ordinary citizens? Perhaps, though it is beyond the scope of this paper and the knowledge of this author to address these questions in any depth. The considerations raised in this paper may suggest that ordinary citizens have duties to engage in bottom-up attempts to promote greater publicity of salient information. 28 This paper's analysis may also provide considerations relevant to policy development around the top-down regulation of microtargeting. If microtargeted political advertisements severely undermine publicity and thereby undermine the quality of the democratic polity, as I have argued, then perhaps democracies should institute policies which raise the publicity of these messages at least to a level that permits a greater degree of public scrutiny. In any case, putting particular policies to one side, my analysis of the wrongfulness of political microtargeting suggests a concrete goal for any ameliorative regulation: raising the publicity of political messaging to a more acceptable level. 29 29 The least disruptive mitigation method is raising transparency. One way of achieving this is a mandate, for example, that political advertisements be uploaded to a publicly available database. The contents of this database could be reported on and discussed by journalists and the public at large. To their credit, Facebook and Google have both provided searchable databases of their political advertisements after the 2016 US Presidential Election and Cambridge Analytica scandal, (Facebook 2023;Google 2021) but there is currently no legal mandate for them to do so, and these databases are overwhelming for an average citizen to search, with hundreds of thousands of data points added each month of a major election cycle. A more disruptive method of raising the publicity of microtargeted advertisements would be to mandate lower limits on the size of the demographic 'slice' at which a political advertisement can be aimed. Whether and where this is legally achievable is beyond my expertise, but US Federal Election Commission Chair Ellen Weintraub has already 'called on social media giants to "stop the practice of microtargeting" (Wilson 2019) and Google instituted this policy in 2019. Google now only allows its political advertisements to be targeted based on 'broad segments' of the population such as gender, age, and postcode (Cox 2019). A lower limit on audience-size would also be synergistic with a public database mandate by making the databases less overwhelming to search and thereby enabling greater public scrutiny. Am I contradicting my claim that microtargeting is, generically, an anti-democratic practice? Can it be made democratic? I do not think so. The suggested measures would effectively ban full-blown microtargeting and return us to more traditional, coarse-grained forms of targeting.