1. What is ‘liking’?

The internet has transformed the scope and scale of communication; it is now easy, fast, and cheap to communicate with someone anywhere in the world, even a total stranger, and we can amass and address enormously large audiences from the comfort of our homes. More interestingly, the internet, and especially social media, has also changed what it means to communicate. As Neri Marsili observes, we can use social media to perform not only the same communicative acts we perform offline, but also communicative acts like retweeting and ‘liking’, which are only made possible by the complex interfaces of social media platforms (Marsili, Reference Marsiliforthcoming). Internet communicators have also developed and/or popularised other innovative forms of communication, including reaction GIFs, emojis, bodily reaction initialisms like ‘LOL’, and memes.

These forms of communication are now commonplace for the 3.8 billion active users of social media, each of whom uses it for an average of almost two and a half hours a day (Kemp, Reference Kemp2020). Yet they are rarely discussed in contemporary philosophy of language, which still tends to focus on face-to-face, one-on-one spoken interaction.Footnote ² It is therefore high time that we expand our scope. In this paper I shine a light on one particular form of online communication: the act of ‘liking’ a social media post.

This tiny act could seem inconsequential or frivolous. After all, to ‘like’ a post is simply to press a button. Yet it is of huge social significance. With ‘likes’ come considerable power; one can now make a career out of being a social media ‘influencer’, i.e. a person who deliberately cultivates ‘likes’ and other forms of engagement on social media to attract sponsorships and advertising deals. In addition, the data constituted by ‘likes’ have been marshalled by social media platforms, advertising agencies, and political campaigns for financial and political gain.

‘Liking’ posts can also have legal ramifications. In 2020, judges in Zurich ruled that ‘liking’ a social media post could constitute legally actionable defamation or slander, if the post you ‘like’ has defamatory or slanderous content. This is because, the court ruled, ‘liking’ such a post amounts to both endorsement and dissemination of its content.Footnote ³ Many people around the world have found themselves in hot water for ‘liking’ posts. For example in the UK, India, and Thailand, people have been questioned or arrested for ‘likes’ which were deemed equivalent to harassment, heresy, or defamation.Footnote ⁴

There is a need, then, for careful examination of what it means to ‘like’ a post, and what it means to amass ‘likes’. Such is the project of this paper. For simplicity and space, I will focus on what it means to ‘like’ a text post, rather than a photo or video. Though one can ‘like’ posts on most social media platforms, I will focus on Facebook ‘likes’, because Facebook, with over 2.4 billion users, is the largest of these platforms. Much of what I say about ‘liking’ text posts will apply to ‘liking’ other content, and much of what I say about Facebook ‘likes’ will apply to ‘likes’ on other social media platforms, even if some platforms have slightly different communicative norms.

I will also set aside the ‘reaction’ functions Facebook recently introduced, which include ‘Care’, ‘Angry’, ‘Wow’, ‘Haha’, ‘Sad’, and ‘Love’. Marsili observes that these reactions have a more straightforward communicative value than ‘likes’ (Reference Marsiliforthcoming). For example, he suggests that to react with ‘haha’ is to express amusement, and to react with ‘angry’ is to express anger. In contrast, ‘liking’ a post is ‘vaguer and less defined’ (Marsili, Reference Marsiliforthcoming). Indeed, the Facebook ‘like’ button does not have a face emoji like the others; it is a thumbs up symbol, which might suggest that ‘liking’ has a gestural function rather than an expressive function.Footnote ⁵

I will also focus primarily on ‘likes’ enacted and accrued by individual social media users. These, I suggest, are paradigmatic ‘likes’, whilst ‘likes’ enacted and accrued by corporations, like Adidas or McDonald's, are somewhat aberrant. ‘Liking’ is fundamentally a way for individuals to interact with each other, and its adaptation by corporations is parasitic on its basic interpersonal function. That corporate ‘likes’ may sometimes not quite fit the theory of ‘liking’ I will advance is therefore not concerning.

It is tempting to think of ‘liking’ as equivalent to asserting the proposition ‘I like this’. After all, when Leah Pearlman, a product manager at Facebook, first introduced the function in a blog post, she wrote: ‘When your friends share something great, let them know you like it’.Footnote ⁶ Presumably, those who regard ‘liking’ as a form of endorsement, like the judges in the previously mentioned Zurich case, have something like this interpretation in mind. To endorse something is to declare one's approval of or support for it. ‘Liking’, on this interpretation, does just that.

This analysis of ‘liking’ could explain why accruing ‘likes’ is enjoyable – we all want to know that other people like our content. It could also explain why ‘likes’ are often ambiguous. Sometimes, we ‘like’ a post because we like its content. For example, we ‘like’ a friend's post about getting a job because we are pleased to learn that they got a job. Other times, we ‘like’ a post because we like the fact that it was posted, even if we dislike its content. We might like a friend's post quoting an offensive remark by a politician, not because we like what the politician said, but because we like the fact that the friend has drawn attention to the politician's bigotry. Just as ‘likes’ are ambiguous, the referent of ‘this’ is often ambiguous, too. When we say, ‘I like this’, we either gesture to the ‘this’ in question, or assume that others will infer its referent based on the context. ‘Liking’ seems to work similarly – we are left to figure out to what exactly it refers.

Yet there are good reasons to doubt that ‘likes’ have this propositional content. If they did, they would likely be regarded as standalone contributions to online conversations, yet this is not the case. Unlike online text, photos, videos, GIFs, and audio, ‘likes’ are not generally regarded as a kind of digital content contributing to online discourse, i.e. as something consumed and engaged with on its own terms. Moreover, if ‘likes’ had propositional content, we would expect there to be disputes about the truthfulness of ‘likes’, in the same way there are disputes about the truthfulness of text posts. Yet this isn't the case, presumably because ‘likes’ are not interpreted as expressing propositions.

Instead, ‘likes’ seem more likely to express some non-cognitive content, like the (real or purported) affective states of users, in response to other peoples’ conversational contributions. Maybe, then, we should model them as equivalent to saying ‘Woo!’, ‘Yeah!’, or ‘Yay!’. This analysis of ‘liking’ reflects the phenomenology of ‘liking’ in many cases. When we ‘like’ a post about a friend's new job, we often want to express positive emotions in response to their news.

Yet any account of ‘likes’ as expressing a positive attitude or emotion, cognitively or non-cognitively, ultimately falls short, because ‘likes’ do not always involve expressing positive attitudes and emotions. For example, sometimes we ‘like’ sad posts, such as those sharing information about bereavements and traumatic events, even though we do not wish to convey that we are happy about these events, nor the fact that the posters had to write about them. We might not feel any positive emotions towards any aspect of the post at all.Footnote ⁷ Sometimes we ‘like’ a post to signal that we share the sentiments expressed; we might ‘like’ a post criticising a political policy because we share the bitterness expressed by poster.Footnote ⁸

In fact, we have good reason to abandon any attempt to identify some stable content, cognitive or non-cognitive, which ‘likes’ express. ‘Likes’ seem to transmit many different kinds of information; their ‘content’ is not stable, and they have no recognisable, conventional ‘meaning’. As such, any content-based definition of ‘liking’ seems doomed. What does seem to be stable, however, is the social function of ‘liking’. This indicates that we may gain more traction with a functional definition.

I propose that ‘likes’ are best theorised as serving a social, or, more specifically, phatic function (a function curiously neglected in philosophy of language). The notion of phatic discourse was first developed by anthropologist Bronisław Malinowski, in his analysis of ‘phatic communion’, a ‘type of speech in which ties of union are created by a mere exchange of words’:

We have various aims when communicating with one another. Sometimes, we communicate because we want to increase the common stock of knowledge in the conversation or achieve an alignment of beliefs. Sometimes, we want to use communication to build social bonds and bring people together ‘by the mere need of companionship’ (Malinowski Reference Malinowski, Ogden and Richards1923, p. 316). Phatic discourse, as characterised by Malinowski, is concerned only with the latter goals. This kind of discourse is often characterised as ‘small talk’, and is exemplified by ritual exchanges of ‘How are you?’ and ‘Fine, thanks, how are you?’, as well as ‘purposeless expressions of preference or aversions, accounts of irrelevant happenings, comments on what is perfectly obvious’ (1923, p. 314). During small talk we might share information with one another, but that information is often arbitrary and/or irrelevant, and its transfer serves an instrumental function. When making small talk with someone about the weather, for example, we likely do not really care about the weather, nor our interlocutor's preferences about the weather, and we may already know everything they tell us about the weather. Yet we exchange these pleasantries because doing so facilitates camaraderie.

Sometimes, entire conversations have a phatic function, and other times, individual communicative acts within conversations have a phatic function. They might have this incidentally (for example, one might make an assertion about the weather for phatic reasons), or they might have it essentially; some acts are definable in terms of their phatic function. Consider, for example, the speech act of greeting someone, or gestures like waving, shaking hands, smiling, and nodding. To understand what these acts are, we must understand the phatic function they fulfil. ‘Liking’, I propose, is best theorised as a similar kind of essentially phatic act. It seems to fulfil at least two phatic functions.

Firstly, ‘liking’ helps institute and maintain connections by enabling social media users to make posters aware that their content is being seen, i.e. it provides evidence of uptake. Uptake is to be understood as the hearer's receipt of another person's message; it involves hearing/reading that message and coming to a particular interpretation of it.Footnote ⁹ Posting online can feel like shouting into the void, as we do not get any immediate signs of uptake from our interlocutors; we cannot see them meeting our gaze or nodding, nor can we hear them performing interjections like ‘Oh’, ‘Yeh’ and ‘Mhmm’. The latter behaviours, known in linguistics as ‘backchannel responses’ (Yngve, Reference Yngve1970), tell us that someone is actively listening to us speak.Footnote ¹⁰ ‘Likes’ serve a similar function. Just like the backchannel response, ‘Mhmm’, ‘likes’ need not have any content, but rather they simply signal to the poster that their post has been seen, i.e. that they aren't shouting into the void.

‘Liking’ each other's posts is often considered a requirement of online friendship. Many people ‘like’ friends’ posts routinely and out of a sense of obligation, without really reading or engaging with them. This is good evidence that ‘liking’ functions to signal uptake; we expect our friends to listen to us, not to ignore us, and so ‘liking’ posts helps reassure people that they have an audience, which is still listening and engaged.Footnote ¹¹

Secondly, ‘likes’ function to anoint the ‘positive face’ of the poster.Footnote ¹² This claim requires some unpacking. Politeness theorists Penelope Brown and Stephen Levinson characterise a person's positive face as their positive self-image (Reference Brown and Levinson1987, p. 61). When we speak of ‘saving face’, this is what we have in mind; we like to think that we are held in high regard by other people, and we try to avoid situations in which others might come to think less of us; we try to preserve our positive face.Footnote ¹³ Politeness dictates that we should preserve and anoint the positive face of other people, too; this is known as positive politeness. We can do this in a number of ways, including attending to a person's interests/wants/needs, showing approval and sympathy towards them, agreeing with them, presupposing or asserting common ground with them, joking with them, extending offers and promises, and giving them gifts (1987, p. 322).

‘Liking’, I propose, is a similar kind of positive face preservation. Facebook was probably not so-called because its founders took an interest in sociolinguistic work on politeness. Yet it is striking that one of the most well-known functions of the platform seems to have this face-preserving function. I argued earlier that ‘likes’ do not necessarily express positive emotions, nor convey that the ‘liker’ actually likes the post. Yet their function, overall, is still positive. When you ‘like’ a post, your ‘like’ will be added to a tally next to the post, and if you have ‘friends’ on the platform, they may see on their newsfeeds that you have ‘liked’ the post. To ‘like’ a post is therefore to bestow a small gift upon the poster, specifically a gift of recognition, (sometimes) increased visibility (due to social media algorithms), and elevated social standing in the eyes of other users. Unless they have tightly controlled, private accounts, social media users communicate in the knowledge that their communicative acts could be seen by many others. Positive face is closely tied up with how we believe we are seen in the eyes of others, and so the publicity of ‘likes’ contributes to their face-anointing effects.

The backchannel, uptake-signalling function and the politeness function of ‘liking’ may sound similar, and indeed most backchannel responses are also acts of positive politeness, but the two functions can come apart. One can engage in backchannel responses that are not forms of positive politeness – for example one could signal uptake by staring intensely at someone as they speak. And one can engage in positive politeness strategies which are not backchannel responses; for example, one might initiate a conversation with an act of positive politeness (a handshake, a compliment, et cetera). ‘Liking’ seems to have both functions.

Described in this way, ‘liking’ seems to have an equivalent in offline discourse: smiling or nodding. Such gestures typically convey to a speaker that you have registered their speech – they signal your uptake – and they are also polite, in so far as they preserve the speaker's positive face, especially when audiences are looking on. This analysis can accommodate the fact that we ‘like’ posts we do not actually like and do not wish to convey that we like. If a person tells us a sad fact, we still signal our uptake and preserve that person's face by meeting their gaze, nodding, and performing other phatic gestures. We ‘like’ sad posts for similar reasons.

This analysis is very minimal, in that instead of attributing a fixed meaning to ‘liking’, it attributes a general function. This enables it to accommodate the fact that ‘likes’ can have different social significances in different contexts. The identities of the ‘liker’ and the poster, the relationship between them, the content of the post, and the general context, will all affect how exactly ‘likes’ are interpreted. A ‘like’ out of the blue by a person who rarely interacts with the poster might convey something quite different from a ‘like’ by a friend who routinely ‘likes’ all posts. In different contexts, ‘likes’ can be olive branches, flirtatious gestures, attempts to get back in touch, routine demonstrations of friendship, or bestowals of prestige (if performed by celebrities). Yet what all such ‘likes’ have in common is a basic social function.

We are now in a better position to consider how ‘likes’ should be treated in law. Is ‘liking’, as the Zurich case held, a form of both endorsement and dissemination? Let us tackle the endorsement question first. Endorsement is ‘the act of saying that you approve of or support something or someone’.Footnote ¹⁴ I have argued against interpretations of a ‘like’ as either an assertion of the claim ‘I like this’ or a positive expressive act akin to saying ‘Yeah!’, as well as against the idea that ‘likes’ have any stable content at all. Therefore ‘likes’ are not necessarily endorsements (though they could be endorsements in some contexts). The same is true of smiling or nodding at someone – these are not necessarily endorsements of the person or what they are saying. ‘Liking’ a post does not necessarily mean that one endorses the post's content, and so ‘liking’ defamatory content should not necessarily count as endorsement of that content. To treat ‘likes’ as endorsements by default would be to attribute too substantive a conventional meaning to the act of ‘liking’ something.

Whether a ‘like’ is a form of dissemination, meanwhile, seems to depend on the structure of the social media platform. In the Zurich case, judges held that by ‘liking’ defamatory posts on Facebook, the defendant ‘made them accessible to a large number of people’. Given the defendant had other Facebook ‘friends’, this is true, regardless of his intentions. ‘Liking’ posts on Facebook does share them with one's broader network, in so far as the ‘like’ and the ‘liked’ post may appear on the newsfeeds of the ‘friends’ and ‘followers’ of the ‘liker’ (depending on the obscure workings of the Facebook algorithm), as well as, sometimes, on the profile of the ‘liker’. Thus a ‘like’ is likely to constitute an act of dissemination, if dissemination is understood merely as sharing content with other people, and provided the ‘liker’ has at least one ‘friend’ or ‘follower’.

That a ‘like’ can be an act of dissemination does not entail that ‘liking’ should constitute litigable dissemination in cases where the content ‘liked’ is defamatory or hateful. One should also factor in the size and identities of the audience, as well as the possibility of accidental ‘likes’. A ‘like’ can be performed with a single click, so it is incredibly easy to accidentally ‘like’ something and thereby disseminate it. Establishing mens rea for ‘liking’ could therefore be difficult. That is, if it is required at all; if a disseminating ‘like’ ends up causing harm, then perhaps the ‘liker’ could be culpably negligent or reckless, even if they did not explicitly intend to ‘like’ the post.

It is very easy to perform unintentional communicative acts online; one can accidentally press buttons on social media platforms, send emails to the wrong people, or paste the wrong text into instant messenger. Internet users can also have their online identities hijacked and have messages sent to others in their name. I lack the space to examine the difficulties of attributing intention and authorship to online communicators, but these issues should be taken seriously when considering online crimes.

2. Being ‘liked’

I turn my focus now away from the act of ‘liking’ and towards the ‘like’ tally, which records and displays the number of ‘likes’ a post has received. If one regularly accrues many ‘likes’ on many posts, one develops a kind of power or status, with significant economic, political, and social ramifications. ‘Like’ tallies both institute this power, i.e. bring it into being, and serve as a measure of it, telling us how much of it social media users possess. In this section I consider how to theorise this power. I will argue that accrued ‘likes’ are a digital form of what Pierre Bourdieu called ‘social capital’.

If ‘likes’ were acts of endorsement, then the number of ‘likes’ a person accrues on their posts could be a measure of the esteem in which they are held. Geoffrey Brennan and Philip Pettit state that to hold someone in ‘esteem’ is to take up an evaluative attitude towards them, based on a positive rating of the person against some characteristic, compared with other people (Brennan and Pettit 2004a, p. 17). If many people take up this attitude towards you, you can be said to possess the resource of esteem. Brennan and Pettit suggest that esteem could play ‘an especially important role on the Internet’, because the internet enables us to access large, enthusiastic audiences (Brennan and Pettit, 2004b, p. 140).

Yet ‘likes’ are not necessarily acts of endorsement, and so ‘liking’ a post doesn't always signal that you hold the poster (or their post) in esteem. Indeed, some people regularly accrue many ‘likes’ but are not held in high esteem by the people ‘liking’ their posts; how many ‘likes’ someone has is not always a good guide to how people think of them. In addition, ‘likes’ (and the power that comes with them) seem to be a tradeable commodity, but, as Brennan and Pettit stress, esteem cannot be traded: ‘However much people want esteem, it is an untradeable commodity: there is no way that I can buy the good opinion of another or sell to others my good opinion of them’ (2004a, abstract). If esteem could be traded, it would cease to be esteem. Yet ‘likes’ can be traded. Some users buy the services of ‘click farms’, where workers are paid to ‘like’ thousands of social media posts a day.Footnote ¹⁵ Other users do not buy but rather solicit ‘likes’; they tell users that their ‘likes’ will constitute entries into competitions, or they ask their followers to ‘like’ them in exchange for ‘shout-outs’ or as a favour.

For these reasons, we should not think of accrued likes as a reliable measure of the esteem in which a person is held. A better way to conceive of them can be found in Pierre Bourdieu's work on capital. Capital, for Bourdieu, is accumulated labour (1986, p. 241). Ordinarily when we think of capital, we think of economic capital, i.e. assets like financial resources, land, and property, which are ‘immediately and directly convertible into money’ (1986, p. 242). It is tempting to think that our social order is determined wholly by the distribution of such capital and the systems which facilitate its exchange. Yet Bourdieu argues that it is ‘impossible to account for the structure and functioning of the social world’ unless one accounts for the various forms of non-economic, immaterial capital (1986, p. 242). If the social world were a casino, Bourdieu suggests, we gamble not only with the black chips of economic capital, but also with the blue chips of cultural capital and the red chips of social capital (see Field, Reference Field2003, p. 14).

Cultural capital is a form of cultural knowledge and socialisation. A person who is au fait with contemporary literature and can talk about it articulately, for example, possesses cultural capital.Footnote ¹⁶ Cultural capital facilitates upwards social mobility and the acquisition of economic capital. The aforementioned person could leverage their cultural capital to get into Oxford or Cambridge, and as a result of the academic qualification then conferred (itself a form of cultural capital), they may find it easier to get a high-paying job.

Social capital, meanwhile, is ‘the sum of resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalised relationships of mutual acquaintance and recognition’ (Bourdieu and Wacquant, Reference Bourdieu and Wacquant1992, p. 119).Footnote ¹⁷ The extent of one's social capital depends both on how many connections one has, and the capital (economic, cultural, social) possessed by the people with whom one has connections. Just like cultural capital, social capital facilitates the acquisition and maintenance of economic capital. The phrase ‘It's not what you know but who you know’ alludes to the power of social capital; even if you lack economic and cultural capital, you can mobilise and exploit connections with people who do have these forms of capital in order to acquire them.

When thinking about social capital, Bourdieu had in mind the connections facilitated by old boys’ networks, unions, secret societies, and sports clubs. Yet social media has radically transformed the nature and extent of our social connections. Now we can develop connections with people of all demographics, even if we have never met them in real life, and we can have these interactions with thousands of people around the world at any time and at low cost. These kinds of connections are facilitated and constituted by functions such as ‘liking’. And so I suggest that the ‘like’ tally both institutes and measures a digital form of social capital.

The number of ‘friends’ one has may also be a measure of social capital.Footnote ¹⁸ Indeed, it may initially seem a better guide to social capital than ‘like’ counts, since friendships are by definition forms of social connection. I consider both ‘friends’ and ‘likes’ guides to social capital, but ‘likes’ are a better guide to current social capital. This is because ‘liking’ a post is a form of active engagement, whereas ‘friending’ someone, once completed, requires no further action. You can amass thousands of inactive ‘friends’ who never interact with you or your content. ‘Friending’ someone merely gives a relationship formal recognition, whereas ‘liking’ posts is a form of relationship upkeep. The more ‘likes’ you regularly accrue, the more people are actively engaging with you.

This account of ‘likes’ is concordant with my account of the ‘liking’ function. Social capital, for Bourdieu, reflects relationships of mere ‘acquaintance and recognition’, and I have argued that ‘liking’ is a vehicle for signalling your recognition of someone's presence online. This account can also explain why one can have many ‘likes’ but lack popularity or esteem. One can have considerable social capital, but still be widely disliked.

This account is also consonant with talk of the ‘like economy’ (Gerlitz and Helmond, Reference Gerlitz and Helmond2013). This term reflects increasing recognition that ‘likes’ are a form of capital which is not obviously economic nor cultural. The power that comes with ‘likes’ is often referred to in contemporary internet discourse as ‘clout’, a word which historically seemed to refer to a form of social capital. In 1973, journalist Mike Royko wrote about clout and its role in Chicago society, defining it as ‘political influence, as exercised through patronage, fixing, money, favors, and other traditional City Hall methods’. He reported people saying things like ‘I don't need a building permit – I got clout in City Hall’.Footnote ¹⁹ ‘Clout’ has lost its connection to political influence in particular, but the idea that it is a kind of social currency persists. This is presumably why it is now used to refer to high engagement levels on social media.Footnote ²⁰

In addition, the social capital model captures the idea that one has to work hard to acquire ‘likes’. The same is true of social capital, which is the product of accumulated social labour. Bourdieu stresses that acquiring and maintaining social capital requires an ‘unceasing effort of sociability’ (1986, p. 250). One must work constantly to transform contingent relationships into ‘social relationships that are directly usable in the short or long term’ (1986, p. 249). For social media ‘influencers’, this ‘unceasing effort of sociability’ is a full-time job. To build an internet platform through ‘likes’, one must constantly produce content, solicit engagement, and interact with other users (especially those with many ‘likes’ themselves).Footnote ²¹

Modelling ‘likes’ as a form of social capital can explain why with more ‘likes’ often comes more economic capital. ‘Likes’ cannot be immediately and directly converted into money, nor can they be easily transferred to other people. Yet high numbers of them can generate lucrative advertising deals and career opportunities. All of this is congruent with Bourdieu's argument that the more social capital we acquire, the greater our access to economic capital, even if the two forms of capital cannot be directly exchanged. This explains why successful social media influencers often enjoy rapid upwards social and economic mobility, even if they start out with very limited capital.

Prior to social media, social capital was hard to measure. Indeed, Bourdieu himself never specified how social capital could or should be measured. ‘Like’ tallies have made social capital (or at least one form of it) both more visible and more measurable online. I explore some of the effects of this in the following section.

3. Should we like ‘likes’?

Social media is a complex ecosystem, of which the ‘like’ function is one small part. Teasing out the individual effects of this system's discrete but interrelated functions and capabilities can be difficult. Indeed, in many ethical debates about social media, for example concerning surveillance capitalism, privacy, manipulative algorithms, addiction, mental health, identity construction, online abuse, political activism, and fake news, social media is treated as a single structure; social media as a whole is said to cause this or that problem. Nonetheless, in this section, I try to home in on the impact the ‘like’ function in particular has had on online communicative practices. I do not claim that ‘liking’ is the sole cause of the effects I identify; similar points could well be made about other social media functions. The following considerations are also somewhat speculative; to draw more concrete conclusions would require robust empirical evidence.

The ‘like’ function seems prima facie to have improved online communication in at least one way, by making it easier. Before the introduction of the ‘like’ function, social media users had two options after reading someone's post; they could leave a comment, or they could do nothing. Commenting requires creativity, concentration, and time, and so many chose instead to ‘lurk’, consuming content silently and passively. The ‘like’ function changed this, by offering users a quick and low-cost way of interacting with posters. If we think interacting with others is a good thing, then anything that makes such interactions easier (and as a result, probably more frequent) is surely good, too. ‘Likes’ certainly make communication more satisfying for online ‘speakers’, who can now know that they are being heard, and that their communicative efforts have not gone ignored.

That said, this benefit will be outweighed, or at least reduced, if it turns out that the ease of ‘liking’ has led us to forego more time-consuming but also more meaningful interactions in favour of higher levels of less meaningful interactions. It may be that ‘liking’ is so easy and takes so little effort that it encourages us to be lazy. Rather than come up with a thoughtful response to someone's post, we can just ‘like’ it and continue on our way. In this way, the ‘like’ function might actually be making us worse listeners/readers; it is a form of pseudo-engagement which absolves us of the guilt of not responding to others’ posts but creates the bare minimum of human connection. These things are hard to assess without empirical data, but the worries just outlined make me reluctant to conclude that the ‘like’ function has ultimately improved online communication simply by making it more prevalent. The quality of our interactions seems to matter as much as, if not more than, their quantity.

‘Like’ tallies, meanwhile, might seem to have improved online communication by providing social media users with a new kind of knowledge, in the form of quantified uptake. When we post something on social media, we get clear, objective information from our ‘like’ tally of how many people have acknowledged our post. This knowledge seems useful for a number of reasons.Footnote ²² On a personal level, ‘like’ data can help us gauge who and how big our audience is, and so it enables us to better tailor our communication to them. More generally, ‘likes’ are a useful way of measuring other people's social capital. This is particularly useful in political spheres; ‘like’ tallies give us concrete data about the influence of political figures. This can help us take the political temperature during the run up to elections and referenda, and it can also alert us to the growth of political movements. Offline, social capital can be very hard to measure, but online we can see its distribution clearly.

However, just as some kinds of interaction are more valuable than others, so too are some kinds of knowledge more valuable than others. ‘Like’ tallies do give us information we previously lacked, but this information seems to have had a number of corrosive effects on internet discourse. These effects seem worrying enough to offset any particular benefits ‘like’ data may offer. I.e., it seems there may be some things we are better off not knowing.

In particular, ‘like’ tallies seem to have created a new motivation in our online communicative practices, which can override our commitment to truth. With ‘likes’ comes social capital, which as noted earlier can transmute into other forms of capital and generate various forms of power and influence. For this reason, many of us will be driven to produce content that attracts lots of ‘likes’. Of course, in offline interactions we also desire social capital, but the ‘like’ tally makes the capital-producing potential of online communication perpetually salient, and it also reminds us that online, unlike offline, we have almost infinitely large potential audiences.

Getting ‘likes’ is also addictive; it releases dopamine in the brain that makes us want to come back for more. One study shows that the same circuits in adolescent brains that are activated when they win money or eat chocolate are also activated when they see large numbers of ‘likes’ (Sherman et al., Reference Sherman, Payton, Hernandez, Greenfield and Dapretto2016). ‘Likes’ are a good example of ‘gamification’, which C. Thi Nguyen defines as the introduction of ‘game-like elements’ into our lives (Reference Thi Nguyen2020b, p. 189). Nguyen observes that gamification can be intentional, as in the case of technology like the FitBit, or unintentional, if it occurs as a result of metrics and quantification introduced to collate information (Ibid). Metrics, Nguyen suggests, can ‘look a lot like points’, and they can encourage us to ‘redefine our notions of success’ in the terms specified by those metrics, in order to get more ‘game-like pleasures’ (Ibid).Footnote ²³ Driven by the desire to accrue these points, we may ‘lose ourselves in the pursuit’, leading us to set aside all other values and commitments (Reference Thi Nguyen2020b, p. 192).

The drive for ‘likes’ seems to encourage many of us to produce or share shocking and attention-grabbing content, even if we are unsure of its veracity and even it is divisive or hostile. If our audience has thousands of posts to sift through, we need to say something dramatic to get their (and the algorithm's) attention.Footnote ²⁴ Thus the ‘like’ function might partially explain why fake news proliferates on social networking sites.Footnote ²⁵ Our desire for engagement with others, and the social capital that comes with it, can make us care less about whether the claims we make and share online are true, as well as whether the content we share has been deliberately designed by others to trigger our biases and vulnerabilities, or to serve some nefarious political goal. This makes social media users more vulnerable to manipulation and can lead to the dissemination of harmful ideologies.

This also hampers meaningful and productive political deliberation online. If we are not interested in getting at the truth, but only in getting ‘likes’, and if we know that others take this approach, too, we will not be interested in exchanging information, reasons, and arguments with one another, but rather with fighting it out for the most exciting online content. In its early days, the internet was heralded for its potential to improve democracy. Many thought the internet could bring about what Jurgen Habermas calls the ‘ideal speech situation’ (Reference Habermas and Fahrenbach1973), a context in which agents can participate in free, rational, and open democratic debate (Ess, Reference Ess and Ess1996; Dahlberg, Reference Dahlberg1998).Footnote ²⁶ Yet the ‘like’ function has revitalised the age-old worry that vivid rhetoric and emotional appeals will win out over rational deliberation in democracies. It has done this by quantifying social capital and making it ever-present in online communication, thereby making demagoguery a more salient and tempting prospect than ever before.

In addition, the desire for ‘likes’ might also lead us to (consciously or unconsciously) surround ourselves with people who share our interests and commitments. Several social media analysts have observed that social media creates ‘context collapse’ (boyd, Reference boyd2002, 2010; Marwick and boyd, Reference Marwick and boyd2011; Davis and Jurgenson, Reference Davis and Jurgenson2014). In most of our offline conversations, we know who our audience is, so we can predict the reach of our utterances and we can adjust what we say, as well as the style in which we say it, to suit the context. Online, we can never be sure who or how many will hear us. We are rather shouting into an audience of several billion, any number of whom may potentially hear us. Thus there is no identifiable ‘context’. This ‘context collapse’ was the case before ‘likes’ were introduced, but ‘likes’ make it visible to social media users; they are given evidence, via the ‘like’ metric, that people from multiple different audiences could view their content. By making previously invisible hearers visible, ‘likes’ remind us that online communication is a kind of digital panopticon; you never know who is watching.

This context collapse means there is often no recognisable common ground in online discourse. Philosophical analyses of offline conversations typically assume that all participants in a conversation are aware of each other's presence, and work together to develop what Robert Stalnaker calls a ‘common ground’ between them; throughout the conversation they collaboratively construct a set of attitudes and propositions mutually accepted by all conversational participants (Reference Stalnaker2002). Participants can take what is in this common ground for granted, and this enables conversation to proceed at a fast pace. Participants can presuppose the truth of claims already in the common ground, rather than having to assert them outright every time they speak.

If we are indiscriminate about the social connections we make online, it will be hard to know what is in the common ground. And if we do not know which beliefs we can safely presuppose or assert, we will not know what content will get the most ‘likes’. One way around this is to construct a cocoon of ‘followers’ and ‘friends’ who share our world-view and thus who will want to read and then ‘like’ our content. To use C. Thi Nguyen's terminology, these cocoons could take the form of epistemic bubbles or echo chambers (2020a). Epistemic bubbles are relatively homogeneous social epistemic structures, and echo chambers are epistemic bubbles combined with systematic distrust of outsiders. Both can lead to more political polarisation and partisanship, though echo chambers pose more of a threat to cross-partisan debate, because they make people deliberately unwilling to listen to other viewpoints.

In this way, ‘likes’ may have made cross-partisan engagement less likely, and contributed to the documented increasing political polarisation of Western societies (Pew Research Centre, 2014).Footnote ²⁷ That social media fosters political polarisation is not a new observation (see, for example, Pariser, Reference Pariser2011; Sunstein, Reference Sunstein2018), but what is new is my suggestion that the ‘like’ function in particular has played an instrumental role in this phenomenon. This is because the ‘like’ button reminds us constantly of our online social capital, and it strengthens the cognitive and social incentives for producing content that accrues many ‘likes’ – many will therefore adjust their circles (consciously or subconsciously) in order to guarantee a steady stream of ‘likes’.Footnote ²⁸

Thus though ‘likes’ may generate a new kind of knowledge, and make (superficial) engagement with others online easier, they also ‘gamify’ communication, making us inclined to pursue ‘likes’ at all costs, regardless of whether we are sharing meaningful or true content. They may also lead to increased polarisation, as we cultivate homogenous online communities in which we can guarantee we will receive ‘likes’. ‘Likes’ may ultimately not be that likeable after all.