There are two approaches that we can take when theorising with natural language.
One approach is to be a gardener, who respects the complexity and organic unity of natural language. The gardener takes pains to understands subtle distinctions, and takes a conservative attitude that is wary of change.
Another approach is to be a forest manager, who tries to maintain a traversable landscape by trying to control and shape language. Whereas the gardener prioritises natural order, the forest manager focuses on the utility of language as a tool for describing the word. The forest manager is happy to reshape usage or to propose abandoning portions of language. Sometimes the only way to maintain a forest is by starting a controlled forest fire. For example, Sarah-Jane Leslie advocates giving up on generic constructions — such as ‘sharks attack bathers’ or ‘Muslims are Terrorists’ — because of their tendency to support prejudiced generalisations, and David Coady thinks we should stop using the phrase ‘Conspiracy theory’ because of its use to close down legitimate discussion.
My sympathies are with the forest manager: I worry that language can embody collective vices as well as collective virtues, that taking a particular language too seriously runs the risk of mistaking linguistic quirks for features of reality. In this piece, I want to apply the forest manager’s attitude to the phrase ‘fake news’ (I will use quotation marks to refer to terms, and will avoid scare quotes). I think that this term should be a particular concern: its contested meaning raises concerns about whether it refers to anything, and its use as an attack word raises concerns about its use as a propaganda term. I want to make the case that we should completely abandon the term ‘fake news’. This proposal has been mooted in a number of places, notably by Claire Wardle of First Draft news. However, I think that running the argument with tools from the philosophy of language makes particularly clear why we need to stop talking about ‘fake news’. (I should say that similar argument could be made with ‘post-truth’, ‘alternative facts’, as well as non-English terms such as ‘Lügenpresse’.
Let’s start off by thinking about what ‘fake news’ means; what kind of things it refers to. Consider what kinds of stories we can truly apply the phrase ‘fake news’ to (setting to one side uses of the phrase applied to institutions). Although it is easy to identify some paradigm examples — here the Pizzagate story inevitably gets wheeled out — when we try to generalise, we quickly run into a set of difficult questions:
Does ‘fake news’ apply only to stories spread online, or can it also apply to print media? Critics of news institutions are keen to use the term to accuse traditional news institutions of fakery (sometimes with good reason), but often when people talk about ‘the problem of fake news’ they have in mind some kind of online phenomenon, and the solutions that are proposed typically involve social media companies changing how stories are shared online.
Does ‘fake news’ apply only to stories spread by (purported) news institutions, or does it also apply to people making up claims, or reposting items from news institutions? When QAnon posts a cryptic message, or when you or I repost a story without reading it (which appears to be a worryingly common behaviour) does that count? Are we all purveyors of misinformation?
Does ‘fake news’ apply only to malicious stories, or can satire and news parody count? As Axel Gelfert points out, in the mid-2000s ‘fake news’ primarily applied to satire that wore the clothes of traditional media — think of the Daily Show, or the Colbert report. Has the term undergone a shift in meaning? In 2016 it seemed to apply primarily to false clickbait stories that were spread in order to generate income (Macedonian clickbait farms being the paradigm example), but since then it has appeared to shift again, often functioning as a catch-all term for information dysfunction.
Does ‘fake news’ only refer to false stories, or can it apply to partially or even completely true stories? It is natural to think that the interesting phenomenon is something like lying. We might think of lying as saying something you believe to be false with the intention that someone else believes what you said. If ‘fake news’ is an analogous to ‘lie’, then we might expect that it only applies to false stories (or at least to stories which the speaker believes to be false). However, we might also consider cases of misinformation that involve making true claims. Consider what Michael Lynch calls the internet shell game, where malicious actors flood a news environment with a range of stories on both sides of a question, making it difficult to tell the true from the false. In this case the purveyor of misinformation asserts both true and false information to bamboozle her audience. If this practice counts as ‘fake news’, then one might wonder whether ‘fake news’ is closer to Harry Frankfurt’s notion of bullshit than lying. On Frankfurt’s account bullshit is speech which has no consideration for the truth, unlike truthfulness which aims for the truth, and lying which aims away from the truth. This sense of bullshit just concerns the intentions of the speaker, meaning that true claims which are made without consideration from the truth count as bullshit.
Can these questions be resolved by thinking about the ordinary meaning of ‘fake news’? I don’t think so. I think it is simply unclear where the ordinary concept sits on these questions. All natural language terms have hard cases — is a forty foot high feline that emits gamma radiation and can walk through walls properly called a cat? — but here we don’t even have a clear sense of core cases. Researchers can’t agree on whether ‘fake news’ online refers to online content, or whether satire should count.
Maybe we can defer to experts in communications, technology, or philosophy to determine the meaning of the ‘fake news’. As Hilary Putnam points out, although ordinary speakers are not reliable at applying ‘elm’ and ‘beech’ to the appropriate trees, these terms remain meaningful, because we defer to tree-experts to determine their meaning. No such luck with ‘fake news’. Not only are ordinary speakers confused; there is no academic consensus. Researchers writing about ‘fake news’ disagree on all of the questions raised above. The most common operationalisation is to simply label some news sites as ‘purveyors of fake news’, and to treat everything that they produce as instantiating the phenomenon. But the choice of sites will depend on one’s political points of view, and there is inevitably massive disagreement about which sites to target. Figures on the right leverage a perception of media bias to slur establishment news institutions, establishment newspapers focus on sites that spread false and politically motivated claims, and critics of digital capitalism focus on the role that social media and search engines play in manipulating the information environment. It should come as no surprise that academics — like everyone else — use ‘fake news’ as a weapon to attack their political opponents.
To diagnose what’s gone wrong in this discourse, let’s take a sharp left turn from 21st century misinformation to consider an early twentieth century critique of metaphysics. In The Elimination of Metaphysics through the Logical Analysis of Language, Rudolf Carnap launched a radical critique of metaphysics. Whereas past critics had alleged that metaphysical speculations about the fundamental nature of reality were either false, inadequately based in evidence, or merely sterile, Carnap contends that metaphysical claims are simply nonsense. By his lights terms like ‘the absolute’ and ‘the non-Ego’ are utterly devoid of meaning. When Hegel writes “The Ego as the self-positing, and the non-Ego as the self-oppositing are such ideal factors,” according to Carnap’s he has simply said nothing. If metaphysics does not say anything, what is it for? Carnap suggests that it might play an expressive function, similar to classical music. If metaphysical discourse is for anything, it is not for talking about the way the world is.
I suggest that Carnap’s diatribe against early twentieth century metaphysicians gives us a helpful diagnosis of the problems the contemporary discourse around ‘fake news’. Carnap couches his diagnosis of meaningless in a verificationist theory of meaning, according to which terms must be empirically verifiable to be meaningful. This approach to meaning of unpopular, but we can update the critique. Contemporary theories of meaning either appeal to speakers’ beliefs about the properties associated with a term, the ways speakers are disposed to use a term, or historical patterns of usage. All of these criteria fail in the case of ‘fake news’: speakers (including supposed experts) have different beliefs about what ‘fake news’ means, different speakers use the term in incompatible ways, and its use is changing over time, cutting it off from its historical usage.
I think that this suggests that ‘fake news’ doesn’t mean anything. ‘Fake news’ is nonsense, and speakers who utter sentences containing this phrase literally say nothing. Someone who says ‘CNN is fake news’, might as well have said ‘fake news is bryllg’. I don’t think this is a cast-iron argument — what is? — but the mere possibility that ‘fake news’ is nonsense ought to concern us. It would be an unhappy situation if a large swathe of academic and public discourse has nonsense running through it, or if public anxiety about misinformation is directed toward a nonsense term.
If ‘fake news’ does not have any meaning, what are we doing when we use it? As Austin points out, language has plenty of uses besides describing the world. That’s why we have grammatical moods other than the declarative. We can use language to name things (Austin gives the example of naming a ship), to express emotions, and to direct others. Building on the Austinian idea that saying is doing, a number of feminist philosophers have explored the ways that language can be used to do evil things. Slur terms can express hatred, and can rank people as inferior. Declarative language can covertly serve a directive function: repeated use of ‘animals’ connected to the MS-13 gang (which has associations with El Salvador) directs hearers to treat central American migrants as sub-human. And, as Jason Stanley argues in How Propaganda Works, propaganda terms like ‘welfare’ can cue up bad ideologies, including both false claims (such as ‘blacks are lazy’), and problematic habits (such as vivid thoughts of prominent welfare cheats).
’Fake news’ has a rich expressive meaning, and often functions as an epistemic slur. Applying ‘fake news’ it to a news story seems not to describe the story, but to express disdain toward the story, the institution that produced it, and (in some cases) toward people that believe the story. As a number of writers have pointed out, a sentence like ‘CNN is FAKE NEWS!’ is often functionally equivalent to ‘I don’t like CNN!’ Applying ‘fake news’ to a story also has a directive function, exhorting the hearer to distrust the institution that produced that news story, and others like it. We can think of this kind of speech as epistemic policing. Applying ‘fake news’ to a story directs the hearer to disbelieve the claims made in the story, as well as manipulating their wider patterns of trust, leading them to deem certain sources unreliable and biased.
It has been less remarked upon that ‘fake news’ often functions as propaganda. Some of this propaganda is fairly blatant. Recent laws in Kenya, Malaysia, the Philipines, Tanzania, and Uganda put ‘fake news’ front and centre, and look very much like attempts to legitimate government censorship. In western democracies, the manipulation can be a little subtler. Consider the following tweet from @realDonaldTrump (chosen at random):
This tweet is manipulative at a number of levels. For starters, it presses Trump’s own interpretation of a news story, tarring any alternative narrative with allegations of fakery, viciousness, and obsession with the first lady’s appearance (hence the allusion to facelifts). The wider message here is about trust: if the mainstream media — itself a term that is approaching a slur — cannot be relied upon to give unbiased reporting, then they cannot be trusted. If you can’t put your trust in the mainstream media, then you will need to look for alternative media sources that signal their non-partisan, objective status (this is a theme in Fox news, and the Sinclair broadcasting company). The intention is that the repetitive drumming of these messages causes its audience to shift their patterns of trust away from traditional news media, and towards alternative sources that are controlled by parties sympathetic to the president’s line.
Stanley points out that one of the most worrying effects of propagandistic speech is that way that it can implicitly introduce a whole host of ideological claims into a conversation. ‘Fake news’ cues up an ideology of media manipulation, according to which the mainstream media are systematically trying to manipulate the public into believing things that align with the interests of the liberal elite. Even without asserting these ideological claims, they can easily seep into the background of a conversation.
This function of ‘fake news’ talk can be a little hard to diagnose, because it is masked by the association of the term with classic enlightenment intellectual values of truth, objectivity, and free speech. Trump’s tweet is an example of what Stanley calls undermining propaganda: speech that appeals to legitimate democratic ideals, whilst at the same time working to undermine those ideals. The term signals a commitment to truth, and the avoidance of fakery, but the effect of labelling stories from legitimate news institutions as ‘fake news’ to get us further away from enlightenment ideals. Epistemic policing of stories has the effect of sapping legitimate trust in news institutions. Accusations of fakery make it extremely difficult to have a reasonable discussions about whether news stories are true. Warnings about media manipulation that are wrapped up with ‘fake news’ are themselves a form of manipulation. For example, the script used by local anchors in the Sinclair broadcasting company was littered with warnings about news bias and manipulation, but was itself deeply manipulative, wrapping a company line in the trappings of a local broadcast.
Terms that are used for undermining propaganda are honeytraps for defenders of the ideals being undermined. There are lots of examples of defenders of enlightenment ideals to taking up ‘fake news’, which is understandable given its association with their values. This attempt to appropriate right-wing propaganda is deeply problematic. For starters, the use of ‘fake news’ by mainstream figures lends the term legitimacy, strengthening its propagandistic uses, and making it harder to tell apart propaganda from legitimate speech. Although these figures may be well-motivated, the cueing of ideology by propaganda terms is independent of intentions, meaning that ideological content can be smuggled into the background of a conversation without anyone noticing (much as asking ‘when did Japan last win the world cup?’ smuggles in the claim that Japan has won the world cup). When the defenders of enlightenment values use ‘fake news’, they are liable to introduce the ideology of media manipulation. The establishment narrative about public opinion being duped by malicious actors can look like the beginnings of a parallel ideology of manipulation. Perhaps the conspirators are different — we should be worried about Russian interference, and Macedonian troll farms, rather than liberal bias — , and the allegations may have some basis in fact, but in using ‘fake news’ to cue up ideology we are still introducing a set of conspiratorial presuppositions into the conversation without giving them proper consideration.
The honeytrap worry deepens when we realise that establishment figures might use ‘fake news’ to engage in the same kind of pernicious epistemic policing that we saw in the Trump tweet. For example tools like the MediaBiasChart — which sorts providers of news based on their reliability and political bias — already looks pretty close to a visual directive about which sources to believe. Defenders of enlightenment values who engage in epistemic policing are effectively undermining their own values. Even if we have a grip on who is spreading false and unjustified claims, if you buy into the idea that we have a right to freedom of thought, the way to oppose these claims is by persuasion, not by issuing commands about what to believe.
‘Fake news’ is not only meaningless; it has been weaponised, becoming an epistemic slur term and a tool for authoritarian propaganda. What should we do about it? There are two options. We could try to reclaim the term, giving it a determinate meaning, and stripping it of its slurring function, or we could abandon the term.
The reclamation project can look appealing. An optimistic take is that giving ‘fake news’ a new meaning would strip the ideologues of one of their most effective tools, help us to think and talk about the new epistemic challenge facing democracy, and redirect public concern toward that problem. I am not an optimist. Reclamation projects are hard, costly, and slow — consider how long it took to shift the meaning of ‘queer’ — and there is no guarantee of success. Focusing on the definition of ‘fake news’ makes it easy to think that contemporary western Democracies face exactly one epistemic problem. In actuality, there are lots of epistemic challenges, and we need lots of different conceptual tools — both new and old — to think about them.
Although repetition has driven home the idea the 2016 US presidential election was the start of a new and worrisome threat to democracy, when we get down to it, the problems are perennials. People have always lied and bullshitted, newspapers have always sensationalised and got things wrong, politicians have always obfuscated and engaged in propaganda. Recent US politics is strewn with examples of misinformation campaigns aimed at Black people: think of the war on drugs, myths about welfare, and ignorance of the extent of racial segregation and inequality in the prison system. If anything is new it is that the latest round of misinformation targets white members of the middle classes, rather than Black people, immigrations, LGBT+ people, or working class people. If the problems are old ones, then perhaps the place to start is by using terms that have been used to understand the epistemic situation of oppressed groups. For example, Charles Mills’ notion of white ignorance seems like exactly the right tool for thinking about the racist aspects of authoritarian ideology, Misogyny seems like the right tool for understanding why Clinton did not win the 2016 election, and Marxist notions of ideology seem like the right tools for understanding the epistemic effects of propaganda.
Banning ‘fake news’ ought not to be seen as restriction on speech, but an invitation to centre conceptual tools from marginalised groups and develop new concepts to think about the epistemic challenges of democracy. We might be burning down the tangled thicket of ‘fake news’, but the hope is that new shoots will poke through the ashes.
(This is a shortened version of a paper, forthcoming in Inquiry https://philpapers.org/rec/HABSTA. I benefitted massively from discussions at the Fake Knowledge conference in Cologne. Videos of some of the talks are here http://concept.phil-fak.uni-koeln.de/36447.html)