Tag Archives: Grice

Prejudice and Presupposition in Offensive Language

  1. Updating an old distinction: Frege on sense and tone[1]

In a much-discussed example from his Posthumous Writings (from the piece called “Logik” , written in 1897), Frege makes an analysis of the difference between two similar sentences:

(1) That dog howled all night

(2) That cur howled all night

The two sentences, Frege says, express the same thought:

[T]he first sentence tells us neither more nor less than does the second. But whilst the word ‘dog’ is neutral as between having pleasant or unpleasant associations, the word ‘cur’ certainly has unpleasant rather than pleasant associations and puts us rather in mind of a dog with a somewhat unkempt appearance. Even if it is grossly unfaith to the dog to think of it in this way, we cannot say that this makes the second sentence false. True, anyone who utters this sentence speaks pejoratively, but this is not part of the thought expressed (…) It might be thought that the second sentence does nevertheless tell us more than the first, namely that the speaker has a poor opinion of the dog. In that case, the word ‘cur’ would contain an entire thought.

I have quoted Frege at length because the selection contains many ideas that we may summarise as follows:

– The two sentences express the same assertive content, so that if (1) is true then (2) is true;

– However, (2) expresses also a tone or colouring given the pejorative expression “cur”, which suggests a negative attitude towards dogs;

– The term “cur” may be thought to contain an entire sentence expressing a derogatory attitude towards dogs;

– But the sentence ideally contained in the word “cur” is not expressed, but hinted at with the use of the pejorative word; a person unaware of the derogatory meaning of “cur” would interpret (2) as intending exactly what (1) means.

Therefore, we need to distinguish between:

(a) The thought expressed, which has to do with the truth or falsity of the state of affairs described (we may speak of the truth conditional content of the sentence);

(b) The thoughts “which the speaker leads others to take as true although he does not express them”.

The distinction is reminiscent of a distinction already made by Frege in his 1879 masterpiece, Conceptual Notation (Begriffsschrift), where he insists that we have to distinguish between sense and tone:

(a) The sense of a sentence is what pertains to the truth.

(b) The tone or colouring of a sentence is what pertains to pragmatic agreements.

Although Frege does not use the term “implicature”, widely applied by the philosopher Paul Grice in his analysis of implicit communication, many authors have considered his distinction as a forerunner of Grice’s idea of conventional implicature. Following this lead, David Kaplan (1999) suggested developing the Fregean distinction between sense and tone with the following analysis: in pejorative expressions we have to distinguish a descriptive part and an expressive part; both have the same information content (they refer to the same individuals when used to refer), but the pejoratives express also an attitude that we should take into account.

Consider two sentences concerning a crime:

(3) That nigger is the culprit.

(4) That man is the culprit.

Both have the same truth conditions; they are both true or false depending on the person in question having committed the crime, provided that with “that man” and “that nigger” the speaker intends to refer to the same individual. But while the descriptive part of (3) and (4) have the same function in helping the hearer, maybe together with a gesture, to refer to the individual in question, the expressive part of (3) creates a problem because it expresses a strongly negative attitude towards a class of individuals just because of the colour of their skin.

A possible reaction to this difference could be, “I don’t care about expressive aspects or tone: what counts is the truth of the matter”. The problem is just to answer correctly to the questions:

– Is that man the culprit or not?

– Did that dog howl all night or not?

If we are interested only in the objective truth of the matter, who cares about different shades of linguistic expressions? Actually, this reaction has been more and more powerful since the diffusion of “politically correct language”. Sometimes exasperated by the societal request or even imposition to use politically correct language, many people have begun to think that such a language is only an imposition that hides the real beliefs: political correctness comes across as if people abandon their prejudices, while those prejudices continue to stand as solid rock hidden by a pretentious and insincere use of politically correct jargon. After having been exposed to the excesses of politically correct language during his stay in the United States, Flavio Baroncelli, a political philosopher from Genoa, thought of a way out of the difficulties of politically correct language, by individuating—with a sarcastic humour he often used in his interactions with colleagues—its particular properties and possible virtues.

 

  1. A suggestion by Flavio Baroncelli (1996)

Commenting on the (sometimes correct and sound) reactions to politically correct language, Baroncelli reminds us that:

 There is not only a question of truth but also a question of appropriateness.

I was impressed at that time (the mid-1990s) by Baroncelli’s precise wording. Actually, “appropriateness” is a property of utterances, and it is traditionally connected in the studies of pragmatics to the concept of presupposition, which, in turn, is strictly connected with the concept of prejudice. Although this is not the place to define prejudice, given the abundant literature and different concepts behind different words in different languages (and we may refer to the paper by Oprah Załęska in this issue), I want to provide at least a generic distinction about the term “prejudice”, given that literally “pre-judice” means a “judgment before…”. The question remains “before what”?

Is a prejudice a judgment given before having correct information or is it something that comes before a judgment? There are two ways of taking the term “before” that lead us to see two different aspects of prejudice: we may think of a prejudice (a) as a judgement given in advance, before having proper information; or (b) as something that comes before the actual act of judging and supports the judgement. On the one hand, we have missing information that is normally required to give a proper judgment; on the other hand, we have assumptions, beliefs, and attitudes that lie hidden and are taken for granted, as a common ground on which a judgment is possible. These kinds of opinions or beliefs on which we ground our judgments can be labelled—in contemporary terminology—“presuppositions”.

Frege distinguished the mental act of judgment from the linguistic act of assertion: an assertion is the expression of a judgment. Using the term “cur” instead of “dog”, in asserting (2), I express a prejudice against dogs; while giving a judgment on a situation I rely on a background of tacit assumptions that lie hidden in my judgment. Is this necessarily bad? Not necessarily. Actually, every assertion is based on some presuppositions. If I say that Elena stopped smoking, my assertion presupposes that Elena smoked. However, this doesn’t mean that I have a prejudice against Elena; I just tacitly state that she was a smoker in a previous time. We speak of “prejudices” only when we think that presuppositions are fundamentally wrong, and often these presuppositions are wrong because they select some superficial feature of a class to define the class itself as being negatively characterized by those features (race, gender, and so on).

From this point of view, prejudices belong to presuppositions, to what is taken for granted without or before any speech act (assertion, question, command…). A presupposition is what is taken for granted without the need for being expressed explicitly. Prejudices are a subset of the set of presuppositions.  Studying presuppositions, we study the basic features of prejudice itself, features that it shares with “normal” harmless presuppositions, but that may drastically impinge on our well-being and social life.

A basically accepted definition of presupposition is the one introduced by Robert Stalnaker (2002: 712):

[PRES] A sentence S pragmatically presupposes a belief B when an utterance of S is appropriate only if B is shared by participants to a conversation (or B is taken for granted by participants)

Taking the example above, the sentence “Elena stopped smoking” presupposes the belief “Elena used to smoke”, and this presupposition is triggered or activated by a simple piece of the lexicon, in this case, the verb “to stop” that indicates a change of state that requires having done an action before. If I say, “Carlo gave a talk on prejudices again”, I presuppose that Carlo has already given a talk on prejudice because of the use of the iterative adverb “again”. My interlocutors take for granted those presuppositions either because they already know them or because they “accommodate” the common ground of shared beliefs with those presuppositions. Analogously, if I say, “that nigger is the culprit”, I presuppose that blacks are inferior as such, because I use a pejorative word that requires assuming an attitude of contempt towards blacks. And one who uses this pejorative expression assumes that her interlocutors share the same kind of belief and attitude.

There are at least two apparent problems in applying Stalnaker’s theory and his definition to the case of derogatory words, and they are the following:

(i) In using a pejorative in a case of reappropriation, people do not share the prejudice attached to the term; therefore we should say that their use is not appropriate, but intuitively it does not seem so.

(ii) In contrast, the use of derogatory terms by people with racist prejudices seems perfectly appropriate in their own context of dialogue where the prejudice is shared. Should we accept that?

I give here two short answers to these two problems:

(i) Reappropriation as detachment

The term “nigger” is normally and typically used in contexts where black friends enjoy using the term as a signifier of social bonding; but certainly, they do not share a prejudice against black people. However, they share the knowledge of the prejudice attached to the derogatory term and want to explicitly reject the prejudice by using the term in order to change the presuppositions. Not only is the knowledge of the presupposition shared, but also the understanding that they want to detach the use of the term from the prejudice. It is similar to irony, where a term is not used with its literal meaning, but the literal meaning is intended to produce in the audience the contrary of what is normally intended. In the philosophical and linguistic environment, irony is typically interpreted as an implicature or as an “echoing” of others’ point of view in order to mock the speaker. It is as if the group of people wanting a reappropriation were mocking the usage by racists: in using irony concerning their presuppositions, they detach the term from the prejudice and can use it freely—but they cannot leave other people to use it.

Apparently this problem would deserve a deeper analysis, but it is at least useful to have an insight from actual discussion on the subject, like the wording of one famous rapper, Ice Cube: “A slur is like a knife. You can use it as a weapon or you can use it as a tool. It’s been used as a weapon against us by white people, and we’re not gonna let that happen again by nobody, because it’s not cool. It’s in the lexicon, everybody talks it, but it’s our word now. You can’t have it back.”[2] Not everybody agrees on the idea or practice of reappropriation, and some take a more radical stance similar to the one held by Jennifer Hornsby (2001: 129) concerning pejoratives in general: “Derogatory words are ‘useless’ for us. Some people have a use for them. But there is nothing that we want to say with them. Since there are other words that suit us better, we lose nothing by imposing for ourselves a blanket selection restriction on them, as it were.” In particular, with the term “nigger”, Oprah Winfrey claims that the term “should not be a part of the language, of the lexicon”[3].

(ii) Appropriateness of hate speech in small groups

It may sound awkward to say that the use of derogatory terms is “appropriate” in small groups, but it is just a consequence of the definition. And it helps in understanding the working of prejudices. In fact, if an expression is appropriate if its presuppositions are shared by the participants in a conversation, then a pejorative term is perfectly at home in a conversation among racists, because they certainly share the prejudices attached to the pejorative term. And knowing that using a term presupposes a common ground of racist beliefs may help us to acknowledge other people’s perspective—also in order to find ways to contrast them. However appropriate in small groups, racist or hate language should be legally forbidden
 in public—as it happens or should happen, in Italy, where promoting Fascism is a felony punished by the law. A public offence always invites the possibility of legal action, and we have many cases of public debate on that, as well as on situations where the speaker did not intend to offend. (The quotations from the previous section come from a discussion of the use of the term “nigger” by a notorious white television personality.) At the same time, we cannot actually “forbid” using slurs, including derogatory and offensive language, in private conversation. Besides—and this is not so different from reappropriation—it is well known that derogatory language is often used in groups or pairs as a joke or as a sign of confidence. (I may use derogatory language and you are not offended because you know that I don’t mean it.)

But we have invented “politically correct language” where even in private conversation people tend to adhere to a kind of language that avoids pejoratives and offensive terminology. And in this particular fashion, developed to some extremes in the United States, Baroncelli makes his provocative challenge: with politically correct language, racism becomes a “gaffe”.

 

  1. A provocation by Flavio Baroncelli: “Racism is a gaffe”

In what follows, I try to present Baroncelli’s idea without his humour (and therefore missing something relevant, but I cannot be him). Let us take again our examples (3) and (4). Following the definition [PRES] above, the sentence (S) “that nigger is the culprit” is appropriate if it presupposes the sharing of the tacit belief (B) “coloured people are inferior as such”. Now imagine a situation of a classroom in a scholarly educated town for which we may assume that (B) is not shared among the participants in the conversation. Let us imagine that the classroom is brought to a court to assist a case in which—let us say—the former president of the US is accused of having wiretapped Donald Trump. What will happen if a less educated girl—seeing the once president of the US accused of the crime, and maybe unaware of the role of the person in front of her—utters “that nigger is the culprit”? Other students will look at her in a very curious way and will judge her with mixed feelings of astonishment or embarrassment and maybe take distance from her. At this point, facing the reactions of her companions, she will realize that she has made a gaffe.

But what is a gaffe? By common definitions (e.g. Wikipedia), a gaffe is:

To say something true but inappropriate in social context.

By this definition, a sentence is inappropriate in a social context when the presuppositions are not shared. Using the case of politically correct language, Baroncelli, on the one hand, puts racists in a humiliating situation, whereby they are unable to understand the social place they are in, and on the other hand puts politically correct language users in a ridiculous situation, making them reduce racism to a mere gaffe.

Yet there is something deep in this analysis, and it is the attempt of analysing the interaction of different presuppositions in different contexts. The point is that there are always many social contexts and they have complex relations; in small local contexts, you are allowed more liberty. As we have hinted at before, slurs and offensive language are easily used in small groups of friends, xenophobes or not, and offensive language among friends may also be a sign of friendship: you are not offended, but take the slur as a joke, as a colourful way to say something that could be also expressed in “educated” language. Youngsters are used to this (although sometimes there are periods when bad examples by adults get over the fence; Italian television during the Berlusconi era became a means to foster far too much vulgar language[4]).

What politically correct language teaches us is therefore the need to take care of different presuppositions contained in our lexicon and in different contexts where these presuppositions are or are not shared. Only with this awareness can people avoid making a gaffe, when they involuntarily use a pejorative expression in an environment that rejects the prejudices attached to the term. Often young and old people are not aware of prejudices of this kind. An aunt of mine, Maria Bianca Penco, in a report of her travel through Italy after the second World War, wrote something like “….and we met groups of niggers…”. She did not have another lexical item, like “black”, and we had to explain to her that “nigger” is now a pejorative term with such and such presuppositions. She was happy to learn, and she felt enriched and changed her lexicon. But young people are not excusable; they need to learn as soon as possible (and this is the duty of teachers) the presuppositions attached to the lexicon they use.

If in a local small context you are allowed to use slurs, in a larger context you receive social censorship (or even denunciation). The main thing to teach in this regard is that what seems normal in your small environment may be inappropriate if uttered in a larger context. Understanding this implies understanding the stereotypical presuppositions triggered by derogatory words (whose force people are often not aware of), and getting to the roots of prejudice.

What then is the role of politically correct language? Through realizing having made a gaffe, a person may learn the power of the prejudices hidden in language and emotionally react to them; a person may learn more about others and about social history, and, taking a careful attitude towards the use of lexicon in a public environment, the racist himself may find a way to change. As Baroncelli says:

It is not important just having different words; what is relevant is the effort of changing. It is the way we train the animals we are.

Last, but not least, there is also a particular form of prejudice: assuming that others share racist stereotypes while they do not. This attitude, this presumption, may be considered a kind of prejudice and may be felt very offensive. If you attribute a presupposition to a social group where the presupposition is not shared, your utterance in not appropriate, and therefore you make a gaffe. More than 10 years after Baroncelli’s book, I have been struck by an apology made by Microsoft. In the US, Microsoft deployed advertising that depicted three experts in discussion around a table: a white woman, a white man, and a black man. When the company began to use this advertising in Poland, it cancelled the image of the black expert and put in his place a white person, probably thinking that the Polish cultural environment might not have been ready to positively accept a black figure. Many people in Poland reacted strongly, feeling themselves to be judged as culturally inferior by Americans; eventually, on August 26, 2009, Microsoft re-introduced the original picture (with the black expert, as you can see from a journal article commenting on the fact[5]) with a comment, which sounds mysterious unless you know the entire history, saying:

Microsoft apologizes for the gaffe.

 

 

  1. Baroncelli 20 years later

Baroncelli’s main lesson is the search for awareness of the clash of contexts, from contexts of face-to-face conversation to different kinds of contexts of public interaction. What is new after 20 years? The World Wide Web  was invented in 1994; the first University homepage in Genoa (the Faculty of Literature and Philosophy homepage) was launched in 1996, the same year of the publication of Il razzismo è una gaffe. Twenty years later, we realise that two aspects could not have been foreseen:

(1) When derogatory expressions pass by ignorance from the context of private or small-group conversation into the context of social networks.

(2) When derogatory expressions are used on purpose in structured ways in social networks to convey the prejudice presupposed by those words.

If considered with care, (1) is exactly the kind of problem Baroncelli was trying to denounce: you cannot use offensive language out of a restricted context without paying consequences or making others pay consequences. The enormous consequences of offensive language on the Web have attracted public attention; (some) people are beginning to understand that they cannot write the first thing that passes through their mind without having or provoking dangerous consequences. Public offence can have provocative consequences both for the writer and for the offended. It depends on the strength of the offended person, who can be devastated—if young or inexperienced—or can devastate the writer, who may be denounced by the public. The novelty in the social space since the 1990s is the wide variety of social networks, from Facebook to Instagram or YouTube and Twitter. The varieties of contexts on the Web are a novelty that we still have to learn to fully manage and master, trying to find software that could check tens of thousands of pages coming online every minute[6].

However, the analysis made in the previous section, concerning the sharing of presuppositions in different contexts, still keeps its original flavour and interest. And Baroncelli’s legacy might be a warning for teachers to work with students to better understand different levels of contexts of reception.

The second aspect above, concerning the use of social networks for actual intentional spreading of prejudices, fake news, and offensive or hate language, is really something new, and it was unpredictable in the nineties. We can no more speak of a “gaffe” inside a context, but we are facing a new way of spreading prejudices through new means. Here I abandon philosophical and linguistic analysis, and give a short comment on some common news.

The diffusion of offensive language[7] increased sharply during the “Brexit” referendum in the UK (June 23, 2017). In June 2017 in Great Britain we had 5,468 records of hate speech (40% more that one year before), and in July–September 2016 there were 14,300 hate crime reports. We have to consider these to represent only a small part of actual hate crimes, given that most are not denounced. There is a strong hidden support to hate speech grounded on prejudices, which politicians have used to support their party (think of the UKIP, which had a fundamental role in deciding Brexit and disappeared in the June 2017 elections). Similar statistics come from the US after Donald Trump’s election, as a sign that prejudices are not typical of Europe, but are spreading around, supporting different political agendas (we don’t have statistics about hate crimes between Sunni and Shia populations, which go beyond what we know in Europe).

Statistics typically report only actual hate crimes in the streets, expressing prejudices against “other Europeans” or against “non-Europeans” just because of the emergence of nationalism. Is nationalism enough to explain the diffusion of prejudices and hate crimes? Not really, although we already know that propaganda in the Nazi period made great use of prejudices shared or imposed on a great part of the population. What is new today is the way in which hate crimes and offensive language are diffused through the internet, where neo-Nazi and white supremacist channels are always very active, and the way in which countless sites deliberately generate and distribute fake news on any enemy. Some YouTube channels reach very high numbers and have therefore a very high influence in generating prejudices. To provide only a few examples:

– Steve Anderson is a famous US pastor who commented on the massacre at a gay nightclub in Orlando as “good news” and said “there’s 50 less paedophiles in the world”. For him, gay people “were not born that way, but they will burn that way”. His YouTube channel has had 33.5 million views.

– Wagdi Ghoneim is a Muslim preacher and a central figure in the diffusion of hate speech; his channel has more than 200,000 subscribers and has had 31 million views.

– Donald Trump’s twitter account has a similar number of followers: 31 million. A peculiar feature of this president of the United States is that he insists on defining the official press as “fake news”: “The FAKE NEWS media (failing @nytimes, @NBCNews, @ABC@CBS@CNN) is not my enemy, it is the enemy of the American People!”[8] In this way he implicitly suggests that his supporters rely more and more on sites that support hate speech (like the sites supporting the news that Hillary Clinton and her campaign chief John Podesta ran a child sex ring—also provoking an assault on an innocent pizza restaurant in Washington[9]).

The novelty of the Web is that hate speech and offensive language not only create a common ground of shared presuppositions, but they do it while making money. According to marketing experts, extremists and hate preachers have made around 300,000 euros from advertisements for household brands and government departments placed alongside their YouTube videos. The above-mentioned sites make money by spreading prejudices; but in having millions of views they use their sites also for advertising normal products, services, and institutions. And they make a LOT of money (gaining something like $4.18 for every 1,000 clicks may not seem like much, but it becomes relevant if you reach millions of visualisations).

In front of this new diffusion of hate language we need reactions, and perhaps Europe may be able to do something about that. We need both institutional reactions and communitarian reactions. Here are some data and suggestions, selected only from recent news. Two examples of institutional reactions: the Home Affairs Committee (British Parliament) in April 2017 asserted that the largest and richest technology firms are “shamefully far” from taking action to tackle illegal and dangerous content, and specifically that “one of the world’s largest companies has profited from hatred and has allowed itself to be a platform from which extremists have generated revenue.” And the Germany Justice Ministry in April 2017 proposed imposing financial penalties of up to 50m Euros on social media companies that are slow to remove illegal material. But reactions from private firms have also been relevant, and McDonald’s, the BBC, L’Oréal, HSBC, Royal Bank of Scotland, Lloyds, the Guardian, Audi, and Channel 4 are among the companies that have decided to refuse to work with web companies if they permit advertisements on their sites with offensive or hate language[10].

As I reported at the end of Section 3, in 2009 Microsoft made an apology for a gaffe implicating that Poland is a retrograde and racist nation; later, in March 2017, Google’s European chief has publicly apologised after online advertising for major brands appeared next to extremist material[11]. As Aristotle taught us, if you ask for excuses you begin to admit there is something wrong. It’s just a first step.

 

 

References

Baroncelli, F. (1996). ll razzismo è una gaffe; eccessi e virtù del “politically correct”. Roma: Donzelli.

Frege, G. (1879). Begriffsschrift. Halle: L. Nebert. English translation in M. Beaney (1996), The Frege Reader. Oxford: Blackwell.

Frege, G. (1897). “Logik”. In H. Hermes, F. Kambartel, & F. Kaulbach (Eds.), Frege Gottlob 1969: Nachgelassene Schriften. Hamburg: Felix Meiner (pp.137–163). English translation in M. Beaney (1996), The Frege Reader. Oxford: Blackwell.

Gundle S. (1997) “Television in Italy”. In James Coleman and Brigitte Rollet (eds.), Television in Europe, Exeter: Intellect Books, 61-76.

Hornsby, J. (2001). Meaning and uselessness: how to think about derogatory words. Midwest Studies in Philosophy, XXV, 128–141.

Kaplan D. (1999). The meaning of Ouch and Oops. Exploration in the theory of meaning as use. Unpublished.

Penco, C. (in press). Refusing to endorse: a must explanation for pejoratives.

Rovatti, P.A. (2012). Un velo di sobrietà. Uno sguardo filosofico alla vita pubblica e privata degli italiani, Milano: Il Saggiatore.

Stalnaker, R. (2002). Common Ground. Linguistics and Philosophy 25, 701–721.

 

[note] Carlotta Pavese suggested that it is literally wrong to call “prejudices” a “subset of presuppositions”. If a presupposition is expressed it is no more a presupposition. If a prejudice is expressed it is still a prejudice. Therefore I should recommend a lighter rendering of the intuitive idea. I should say that when prejudices are hidden, they work as if they were shared in the common ground, therefore as presuppositions given for granted.The similarity with presuppositions runs as follows:
Like many presupposed contents triggered by a presupposition trigger, also prejudices may be challenged. If you say “Elena stopped smoking” and I know that Elena never smoked, I may react saying: “hey, wait a moment! Elena did not smoke at any time!” canceling the presupposition. If you say “that cur howled all night” I may react saying: “hey, wait a moment! dogs are nice animals; I don’t accept your way of speaking”, putting the prejudice against dogs in the open, rejecting it and pulling it out of the presupposed common ground.
I cannot cancel the prejudice expressing it, but I may refuse to endorse it.

Endnotes

[1] I have developed these hints in Penco (in press).

[2] See https://www.youtube.com/watch?v=gnwiYdFaRfk

[3] See  https://www.youtube.com/watch?v=5A9aPUpHQ6M

[4] P.A.Rovatti, 2012.

[5] See: http://edition.cnn.com/2009/TECH/08/26/microsoft.ad.gaffe/index.html

[6] See for instance Google’s attempt to “flag” hate speech on line:

https://www.ft.com/content/8786cce8-f91e-11e6-bd4e-68d53499ed71

The task is difficult, and any solution has its shortcomings. Think for example of the ontology used by Facebook to avoid and cancel offensive posts. The first solution is to distinguish main “protected” categories and subsets of those categories. This is a tentative ontology that has, among its consequences, the effect that “white man” (main categories) is more protected than “black children” (where “children” is a subset and not a main category). This has been criticised as intentional. However, the difficulty of the task is overwhelming for any ontologist, and we are assisting in the first attempts to provide regulation on the spread of prejudices through hate language.

[7] From now on, unless differently remarked, data comes from The Guardian—a reliable source of information, although not specialised.

[8] Twitter 17 Feb. 2017. Another Trump Twitter on July 27, 2017, was: “So they caught Fake News CNN cold, but what about NBC, CBS & ABC? What about the failing @nytimes & @washingtonpost? They are all Fake News!”
See for instance: https://www.washingtonpost.com/news/the-fix/wp/2017/06/27/trump-renews-attack-on-fake-news-cnn-after-retraction/?utm_term=.49bd0eda471a

[9] See for instance: http://www.independent.co.uk/news/world/americas/hillary-clinton-fake-news-conspiracy-theory-child-sex-ring-edgar-maddison-welch-open-fire-comet-ping-a7456021.html

[10] With results from pressure by the UK government, McDonald and Mark & Spencer’s on Google:
https://www.theguardian.com/technology/2017/mar/21/google-advertising-boycott-hateful-offensive-content
Online petitions are also useful; Sumofus succeeded in making 2,000 companies dissociate themselves from Breitbart and forcing the commerce giant Shopify to adopt hate speech policies. Some gains may also come from websites that actually fight against prejudices:

https://oie.duke.edu/knowledge-base/toolkit/reducingstereotypethreatorg

https://www.nohatespeechmovement.org/hate-speech-watch

[11] “Recently, we had a number of cases where brands’ ads appeared on content that was not aligned with their values. For this, we deeply apologise.” (from link at endnote 8). See also:
”https://www.theguardian.com/technology/2017/mar/20/google-ads-extremist-content-matt-brittin