Social Media Bullshit: What We Don’t Know About facebook.com/peace and Why We Should Care

If we live in media (Deuze, 2012), then at least some of our knowledge of our social lives must come from media. For instance, we gauge popularity—ours and others’—through numbers of likes, comments, retweets and shares; we garner a sense of other people’s taste through “if you liked that then you’ll like this” recommendation systems; indeed, academic knowledge often relies on data gleaned from social media platforms. A consequence of this is an “asymmetry in social knowledge” (Couldry & Hepp, 2017, p. 131) which is in turn reinforced by the opacity of the media structures in and through which we live (J. E. Cohen, 2012). This has led researchers to call for greater openness and transparency on the part of social media platforms (Burgess & Bruns, 2015; Driscoll & Walker, 2014; Rieder, Abdulla, Poell, Woltering, & Zack, 2015), calls that have only intensified since the US presidential elections and the Brexit referendum of 2016, and more recent revelations about the role of Cambridge Analytica.

It is in this context that I introduce Facebook’s “A World of Friends” page. This page claims to show “how many new friendships formed just yesterday” between Facebook users from the opposing sides of three different protracted conflicts: Russia and Ukraine, India and Pakistan, and—the focus of this article—Israel and Palestine (see Figure 1). The message for people visiting the page is clear: Facebook is a force for world peace.

A crucial mechanism in conveying this message is the presentation of large and precise numbers, in which, by and large, we trust (Porter, 1995). Numbers such as these are part of a long-term trend toward quantification—“a constitutive feature of modern science and social organization” (Espeland & Stevens, 2008, p. 402)—and convey a sense of objectivity (Kovacic, 2018) that derives from the centrality of statistics both in policymaking and in everyday life (Alonso & Starr, 1987). Moreover, this kind of quantification, or metrification (Beer, 2016; McCosker, 2017), is precisely what we expect from companies such as Facebook; after all, social media users are by now quite used to seeing affective responses to content transformed in real time into numbers, such as Likes and Shares (Gerlitz & Helmond, 2013; Gerlitz & Lury, 2014).

Having seemingly created a proxy for grassroots-level ties between Facebook users from the opposite sides of political divides, Facebook appears to be serving up data that would allow for a slew of questions about possible correlations between political events and friending behaviors to be answered, such as, What is the impact of acts of violence or periods of sustained violent conflict on Facebook friending? Does news coverage influence friending? What about economic conditions? Do political events, such as peace talks or treaties, impact on rates of friending? For researchers of political communication and conflict resolution, for instance, data such as these could be uniquely valuable.

And indeed, a number of scholars have referenced the facebook.com/peace page. For example, computer scientist Newton Lee cites the page in two books (Lee, 2014, 2015), at one point attributing a “27% drop in Facebook daily Muslim-Jewish friendships” to the Israel-Hamas conflict in November 2012 (2015, p. 120). Communication scholar Ronit Kampf (2011) refers to the page to assert that “Israelis and Palestinians talk to each other on the website on a daily basis” (p. 394), while Gonzalez, Kampf, and Martin call (2012) it a “highly impacting” initiative. Conflict resolution scholar Tom Woodhouse also refers to the page (Woodhouse, 2014, p. 25), and Amichai-Haburger, Hasler, and Shani-Sherman (2015) mention the page approvingly in the context of Facebook being “the most common platform for launching intergroup contact initiatives” (p. 520).

However, as I was thinking about the potential of the page for research, the numbers began to seem unfeasibly large: for instance, if around 200,000 Facebook friendships are made between users in Israel and the Palestinian territory every day (see Figure 1), then it would only take 5 days for one million such ties to be forged. Is this feasible when there are only 1.7m Facebook users in the West Bank and Gaza Strip? Considerations such as this led me to try and make sense of the data independently, as well as to ask Facebook if they could explain them to me.

This article presents my failed efforts to verify the numbers published by Facebook, and my subsequent conclusion that they are unreliable. It is in reaching this conclusion that this paper contributes to theoretical discussions around numbers, social media, and knowledge. Specifically, I argue that the “A World of Friends” page is an exemplary instance of what I term social media bullshit. “Bullshit” is used here as a technical term to refer to a particular category of talk. In Frankfurt’s (1986/2005) seminal text, he distinguishes bullshitting from lying by referring to the speaker’s stance toward what he is saying: the liar knows the truth of the matter but wants us to believe a falsehood; the bullshitter, according to Frankfurt, does not care about the truth, but only about creating a certain impression. Nielsen (2015) has suggested that social media are a fertile site for the spouting of bullshit because “we know so little and because there is a lot at stake” (p. 2). In this article, though, we are not concerned, as Nielsen is, with bullshit about social media, but rather bullshit produced by a social media company. As we shall see, this particular type of bullshit relies on the semiotics of large, precise numbers, as well as on the vastly asymmetrical power relations between Facebook and the individual researcher. Therefore, while this article focuses on a single webpage, the implication is that there may be more social media bullshit out there. It is my hope that this article will help others detect it.

I could only reach the conclusion that the data on facebook.com/peace are bullshit after trying hard to make sense of them. This, therefore, informs the structure of the current paper. To start, I detail the ways in which I tried to verify the data published by Facebook, cross-referencing them with data available from other sources, including a survey I commissioned, and laying out my efforts to get answers from Facebook. Then, having established that the data are not what Facebook say they are, I suggest that they are bullshit, and that they are bullshit because Facebook does not care about their truth status; rather, I argue, the page serves a different purpose, namely, to reinforce Facebook’s image as a force for peace in the world.

This is a somewhat precarious undertaking. At the time of writing (November 2018) I have been waiting for over a year for Facebook to get back to me and explain how the numbers that they publish are compiled. In the absence of any input from them, all I can do is lay out my interpretation, replete with the knowledge that Facebook can refute it. (Last minute update: on Feb 9 2019, Facebook took the page down).

In October 2009, the Stanford Peace Innovation Lab and Facebook launched peace.facebook.com (see Figure 2). The objective of the page was “to measure peace on the Internet,” to which end it displayed the number of “friend connections created each day between people of different countries, religions, and political affiliations.” These included Albanians and Serbs, Indians and Pakistanis, Greeks and Turks, Democrats and Republicans, atheists and believers, and—the focus of this paper—Israelis and Palestinians.² The page also carried the results of an ongoing, one-question survey of Facebook users: “Do you think we will achieve world peace within 50 years?” Facebook’s own framing of the page was unambiguous, with the following text atop the page, beneath the heading, “Come Together”:

Facebook is proud to play a part in promoting peace by building technology that helps people better understand each other. By enabling people from diverse backgrounds to easily connect and share their ideas, we can decrease world conflict in the short and long term.

The page retained this format for around 4 years with only minor changes. As can be seen in Figure 3, the number of “Israel-Palestine Connections,” as defined by Facebook, rose steadily for about 3 years, before declining and then climbing once more. From October 2009 through May 2011, a daily average of around 11,640 ties were formed (or 4.2m friendships a year); from June 2011 through June 2013, that daily average stood at around 24,240 (or 8.8m friendships a year).³

A former Facebook employee involved in setting up peace.facebook.com explained to me in a lengthy email correspondence how the data were compiled. He said that the numbers were based on “self-reporting, whatever users had entered into their Facebook profile,” and that they were “computed nightly based upon our data analytics. Each night, a process would count up the new friend connections in these categories and those numbers would go live the next day.” It is of course possible that users would incorrectly self-report their location, nationality, ethnicity, religious affiliation or politics. However, even given this, the former Facebook employee wrote that he was “quite confident it was being done accurately at that time.”

In June 2013, the site was seemingly mothballed: the Wayback Machine has no records of it from June 22, 2013 until November 22, 2015, and a 2014 book that cites data from August 2012 notes parenthetically that “the interesting friendships statistics did not seem to be working any more in August 2014” (Lee, 2014, p. 222). However, a screenshot of the page dated February 11, 2015 is available from a later book (see Figure 4; Lee, 2015).

This version looks quite different from the original. First, the title is now “A World of Friends,” and the text beneath the heading reads,

Facebook connects people from all over the world even in unexpected places. Here’s a look at how many new friendships formed over the last week.

Second, the multitude of oppositions has been whittled down to three (India/Pakistan, Russia/Ukraine, and Israel/Palestine), though with “Palestine” replaced by the term, “Palestinian Territory.” Third, there is no longer a link to the Stanford Peace Innovation Lab, or indeed any mention of it at all. Fourth, and possibly relatedly, the page has a new URL: instead of peace.facebook.com, the page now sits at facebook.com/peace, with the former site redirecting to the latter, both of which changes suggest a separation from the Peace Innovation Lab.⁴ Fifth, rather than reporting on friendships formed “each day,” the resurrected page shows “how many new friendships formed over the last week.” The figure given for Israel/Palestinian Territory for February 11, 2015 is 63,152, or approximately 9,000 friendships per day, considerably less than the 25,000 per day reported in June 2013.

The next available screenshot of the page is dated November 6, 2015 by the Wayback Machine. Now the figures for Israel/Palestinian Territory are 88,663 “over the last week,” which works out at roughly 12,700 a day, or 4.6m a year, still much less than reported two years earlier. Shortly thereafter, however, the numbers skyrocketed. The next available instance of the page is from an Indian news and entertainment website, which, on February 25, 2016, ran a short piece about the page with a focus on India/Pakistan.⁵ Though unfortunately lacking a screenshot, the short article claims that 154,260 ties had been made the previous day between Facebook users in Israel and the Palestinian Territory, representing a tenfold increase of daily friendships in a period of just 3 months. The numbers for India/Pakistan (from 241,000 per day to 2m per day) and Russia/Ukraine (from 14,000 to 137,000) underwent a similar leap. The next available screenshot, dated April 5, 2016 shows similar numbers: 162,297 for Israel/Palestinian Territory, 1,867,011 for Indian/Pakistan, and 127,409 for Russia/Ukraine. Significantly, it also confirms the return to daily friendship counts, presenting the numbers as showing “how many new friendships formed just yesterday.”

Up to this point, data were attained through the Wayback Machine or Internet searches. On August 4, 2016, however, I started collecting the numbers from the facebook.com/peace page systematically: I would visit the page daily and manually log the numbers in an Excel file. Since December 8, 2016, I have been running a script that collects the numbers into a file and creates a screenshot of the page.⁶ As can be seen in Figure 5, between August 2016 and May 2017 the numbers for Israel/Palestinian Territory held steady at 200,000 to 250,000 new friendships per day. Both India/Pakistan (Figure 6) and Russia/Ukraine (Figure 7) showed a rise in the number of Facebook friendships from August 2016 to February 2017, followed by a decrease thereafter. A dip in the number of new friendships made between Russians and Ukrainians is visible in early 2017.

Since May 7, 2017, the numbers have not been updating, though they are still presented as the number of friendships formed “just yesterday.” Also, at some point between March 11 and May 6, 2017, the original URL, peace.facebook.com, stopped redirecting to www.facebook.com/peace (which showed the current version of the peace page) and started redirecting to peace.facebook.com/peace, which leads to an error page explaining that there is a problem with the way the page is redirecting.

These, then, are the data published by Facebook. In the following, I present a number of reasons for not believing them. In doing so I focus on the Israel/Palestinian Territory case, but some of my arguments apply to the other cases as well. Some readers may view this as overkill. It may seem so obvious that the numbers are not reasonable as to render the following redundant. However, I lay out my thinking and my actions for a number of reasons. First, I felt that I had to give Facebook the benefit of the doubt; if you are going to say that someone has said something wrong, you had better be able to explain why. Second, lacking any concrete proof for doubting Facebook, my strategy is to offer as many counter-indications as I can. And third, I present my slightly exasperating attempts to make sense of the data as a rough guide for how one might go about assessing other data published by social media companies.

Let us start with the very category of “Palestinian Territory.” This is an odd category as the West Bank and Gaza Strip, occupied by Israel since 1967, are usually referred to as the “Palestinian territories” in the plural, and with a lower-case t. “Palestinian Territory” is not a term in common usage, and it was not used in the first version of the page. Nor is it a category that appears anywhere else in Facebook. For instance, when creating ads on Facebook, “Palestine” is a location one can choose for targeting purposes, but not “Palestinian Territory.” Moreover, when looking at the peace page in different languages it becomes clear that the category, “Palestinian Territory,” was added manually (as opposed to being extracted from a pre-constituted list of countries). Facebook renders the names of all the other countries in the language the user is choosing to view the site, except for “Palestinian Territory,” as illustrated in Figure 8. This raises a number of questions: What exactly is the “Palestinian Territory”? What is its relationship with the areas defined as A, B and C in the Oslo II Accord? Does it include Israeli settlements in the West Bank? Does it include East Jerusalem, and the Old City of Jerusalem? Is it based on self-reporting, as the first version of the page was? Is it based on users’ IP address, or the language predominantly used by the user? Is this a case of Facebook trying to avoid controversy by avoiding the word “Palestine” (which some might take as including present-day State of Israel as well)? These questions destabilize the category of “Palestinian Territory” even before we have started to look at the numbers themselves.

To put it bluntly, the numbers reported by Facebook since February 2016 are suspiciously high, both in relation to the earlier numbers reported, and in relation to the size of the Facebook populations in Israel/Palestine. By way of illustration, an average of 20,100 new friendships per day were reported during 2012. From August-December 2016, however, an average of 232,600 new friendships per day were reported. What could account for this huge increase? One possibility is that it is a function of Facebook’s growth. However, I find this unlikely: from Q1 2012 to Q1 2017, Facebook’s global population grew from 901m to 1,900m,⁷ representing just over a twofold increase, compared to an 11-fold increase in the number of reported Israel/Palestine friendships. Zooming in, there were 3.7m Facebook users in Israel in September 2012, and 4.9m by June 2016, representing an increase of only 33% (compared to the 1,100% increase in friendships reported). Perhaps the increase in friendships can be explained by a rapidly growing Palestinian population of Facebook? Again, it would seem unlikely. There were 711,000 users in the West Bank and the Gaza Strip in December 2011 and 1.7m by June 2016. This is a much larger increase than among Israelis, but still far smaller than the growth in the number of friendships across these points in time.⁸Figure 9 illustrates this, along with the parallel data for the India/Pakistan case (there are no data for Russia/Ukraine from the earlier period).

Perhaps, though, the seemingly disproportional increase in the number of friendships is the result of a change in how Facebook is counting them. We know that the earlier version of the page was based on self-reporting; perhaps this version is based on Facebook being able to know where people really are. In other words, perhaps the numbers reported now are correct, and the earlier numbers were far lower than they should have been.

However, some back-of-the-envelope calculations challenge the credibility of the data published since 2016, regardless of their relationship with the earlier numbers. Staying with the Israel/Palestine case, if there are around 1.7m Facebook users in what Facebook calls “Palestinian Territory,” and if around 200,000 new friendships are made between those users and Facebook users in Israel every day, then this implies that, on average, every single Palestinian Facebook user makes an Israeli Facebook friend every 8.5 days. This, quite frankly, is hard to believe. Even the parallel implication that every single Israeli Facebook user would have to have added a Palestinian Facebook friend at least once a month does not seem plausible. Not only are the Jewish and Arab populations of Israel deeply segregated (Lissitsa, 2015), the vast majority of Israelis have no contact with Palestinians from the Occupied Territories.

However, to better understand whether Facebook’s data are feasible, I surveyed 1,005 Jewish Israeli Facebook users, asking them the following question: “As far as you can recall, how many Facebook friendships have you made with Palestinians living in the West Bank or Gaza Strip (in other words, Palestinians who are not Israeli citizens) in the last six months?” Nearly 80% said none; only 1.7% gave an answer above 10 (for the last six months).

However, Israel has a large Arab minority, upward of 20% of the population. Perhaps these Facebook users could account for the friending between Facebook users in Israel and the “Palestinian Territory”? This once more takes us into the realm of back-of-an-envelope calculations, but even if we make the most generous assumptions, it still does not seem possible to reach the numbers of friends that the peace page claims are being made. Let me explain: If 200,000 Facebook friendships are made every day across Israel and the “Palestinian Territory,” then around 73m such ties are made annually.⁹ Israel has 4.9m Facebook users, but how many of those are Jewish? Direct data are unavailable, but we do know that 83% of Jewish adults use the Internet, compared to 53% of Arab adults (Lissitsa, 2015). This means that Israel’s online population is more or less 85% Jewish and 15% Arab. If Facebook uptake is similar across these two online groups (i.e., if 85% of Israeli Facebook users are Jewish), then there are about 4.165m Jewish Israeli Facebook users, and 735,000 Palestinian Facebook users in Israel. If 4.165m Jewish Israeli Facebook users each make 5.66 Palestinian friends from the “Palestinian Territory” per year (which is what my survey found), then they account for 23.57m of the 73m–86m friendships made annually across this particular line of conflict. This leaves a 49m–62m shortfall to be met by Israel’s Arab Facebook population of 735,000, requiring each of them to befriend between 67 and 84 Palestinians from the West Bank and Gaza Strip annually, or somewhere in the region of six per month. Likewise, each Facebook user in the “Palestinian Territory” would have to have friended around 45 Israelis in the last year for the numbers to add up. When we recall that the average number of Facebook friends per user is around 350, with the median closer to 200, this makes the number of friendships claimed by Facebook even less believable.

Looking at the other two cases,¹⁰ the numbers presented by Facebook would require each of Pakistan’s 27m Facebook users to have befriended 33 Indian Facebook users over the last year, and each of Ukraine’s 5.6m Facebook users to have befriended 9.7 Russian Facebook users over the last year. These numbers seem counterintuitive, as they imply that Palestinians in the West Bank and Gaza Strip are friending Israelis four to five times as much as Ukrainians are friending Russians, despite the much closer ethnic and linguistic ties between Ukraine and Russia, and the mingling of populations during the Soviet era.

So far, then, I have shown that the numbers published by Facebook throughout 2016 seem excessive, both in relation to the earlier version of the site, and in relation to the size of the Facebook populations of the areas under consideration. Also, comparing the different cases with one another only increases the grounds for skepticism.

However, not only are the numbers themselves illogical, they do not correspond to events in the world either. By way of example I present the data for Israel/Palestine for October 2016, a month with two major Jewish religious festivals, which are also national holidays. As I shall discuss presently, friending between Israel/Palestine appears to follow a weekly cycle (according to Facebook). As Figure 10, shows, however, this cycle is not at all interrupted by these important holidays in Israel, when religious people, who constitute a significant percentage of the Israeli population,¹¹ turn off their computers and mobile devices. It is slightly mystifying, therefore, to see that there is no less friending on the Jewish New Year or Yom Kippur than there is in the weeks before and after those festivals.

Nor do the data appear to relate to incidents around the Israel-Palestine conflict. For instance, three incidents of Palestinians being killed by Israeli forces made no difference to the amount of friending reported the following day or two, which it should, given the number of Israeli friends Palestinians need to make in order for the numbers to make sense. More pertinently, the quantity of friendings shows no statistically significant correlation with other monthly fluctuating data, such as Palestinian casualties, or number of house demolitions by Israel in the Occupied Territories (see Figure 11).¹²

At this point, I feel that I can state with confidence that whatever the numbers published by Facebook are, they are not the number of friendships made between Palestinians in the Occupied Territories and Israelis. What, then, might they be? One possibility is that they are algorithmically generated, with each conflict’s numbers randomly produced within certain pre-defined criteria. However, if the numbers are randomly generated, then surely one would not expect to see such a strong weekly pattern as we see in the Israel/Palestine case. Oneway analysis of variance (ANOVA) and Bonferroni post hoc tests (which compare the average for each day of the week against the average for each of the other days of the week) show that there are statistically significant differences between the amount of friending on different days of the week. Specifically, Sundays see higher levels of friending than every other day, apart from Mondays, which in turn see higher levels of friending than Wednesdays and Fridays. Why might this be? Social media experts in Israel with whom I consulted could not make sense of this, as they do not accord with what they know about engagement on Facebook, specifically that the number of user log-ins does not vary by day, and is certainly not higher on Sundays than other days. This does not prove that Israelis and Palestinians are not friending each other at higher rates on Sundays, but cross-referencing that with other data about Facebook use at the very least raises questions (not to mention eyebrows).

A similar weekly pattern is also visible in the Ukraine/Russia case. Here, Mondays see statistically significantly less friending than Wednesdays to Saturdays. The India/Pakistan case, by contrast, reveals no differences in friending by day of week. In brief, then, there is a statistically significant but inexplicable pattern in Israel/Palestine friending, a statistically significant but inexplicable and different pattern in Russia/Ukraine friending, and no weekly pattern in the case of India/Pakistan.

The last day that the data on facebook.com/peace were updated was Sunday, May 7, 2017, so obviously since then the claim that the numbers refer to friendships made “just yesterday” is not true. However, it is not true in an uninteresting way. Accordingly, in this section, I discuss the page prior to its freezing, when the numbers were being updated daily. If the numbers reported then did not correspond with the number of Facebook friendships actually made between Facebook users in Israel and the “Palestinian Territory,” or for that matter between Facebook users in Russia and Ukraine, or in India and Pakistan, how are we to understand them?

One possibility is that the numbers really did report something, just not what it is claimed that they were reporting. Maybe in the past, in the first version of the page, they used to report new Facebook friendships, but now they measure something else. Another possibility—one that makes me feel uncomfortable—is that Facebook was simply somehow making the numbers up; perhaps a random number generator did that.

Perhaps, though, this is the wrong way to look for an explanation, and it is here that I take recourse to the concept of bullshit, a notion flagged as deserving of attention among social media scholars (Nielsen, 2015). However, where Nielsen refers to the bullshit talked about social media, I am talking about bullshit produced by social media. As mentioned earlier, in his seminal text on bullshit, Frankfurt argues that bullshit is a function of the bullshitter’s stance (Frankfurt, 1986/2005). A la Frankfurt, to say that Facebook is bullshitting is not the same as saying that Facebook is lying; it is to say that Facebook does not really care whether the data are true or not (and also, that they are not true). Thus, if they are bullshit, the point of the data published on facebook.com/peace is not to tell us exactly how many Facebook friendships were made yesterday between Israelis and Palestinians. The point of the data is to show us that Facebook is a force for world peace. As Frankfurt puts it, “the fact about himself that the liar hides is that he is attempting to lead us away from a correct apprehension of reality.” What the bullshitter wishes to conceal, however, “is that the truth-values of his statements are of no central interest to him; what we are not to understand is that his intention is neither to report the truth nor to conceal it.”¹³

Given Facebook’s refusal to engage with my questions, it is impossible to know what their attitude to the truth of the matter here is. However, having devoted so many words to dissecting their claims, it would be remiss of me not to suggest an alternative explanation. My best guess, then, is that the numbers published from early 2016 (after the tenfold increase in reported Facebook friendships) to May 2017 were massaged somehow because the numbers prior to that did not feel big enough to support the ideological aim of the page. Recall my informant’s comment that “Numbers are always so big around here.” Because we are used to such large numbers when it comes to social media, posting that 88,663 friendships between Israelis and Palestinians had been made over a period of a week might not have seemed enough. In other words, the numbers are (maybe) bullshit because what drives them is not their relationship to reality, but rather their ability to create a certain impression. Their “referential accuracy” (Andrejevic, 2013) is not important; they need to feel right. This kind of approach is also reported by Helen Kennedy (2016), who quotes a social media mining worker as saying that for some clients, “whether that data is accurate or not is irrelevant” (p. 109). As Kennedy notes, what would seem to matter is that data appear to be “real-time and high-volume.” The data published on the peace page meet those criteria, particularly following the tenfold increase of early 2016.

Furthermore, it is not as if the average visitor to the page will know any better; the huge majority of Facebook users do not live in any of the three conflict areas. They have no idea of the size of the Palestinian population of Facebook and whether that population could really be making that many friends with Israelis. They do not come back to the page again (as I did), in order to realize that new numbers are (were) presented every day, which is what raised my suspicions in the first place. In fact, it does not even really matter that the page stopped updating. The impact for the one-time visitor is the same, and that is to make them think that people from opposing sides of violent, intractable conflicts are making friends with one another on Facebook, and maybe even to encourage them to do the same—these may be “reactive” numbers, in Espeland and Sauder’s (2007) sense. Moreover, the visitor is encouraged to view these three vastly different conflicts as fundamentally similar in that all of them can be overcome by everyday people reaching out to one another on social media. Another effect of this is to flatten the power relations between the different parties: no longer are we dealing with strong versus weak, but rather with everyday people with no structural demands of one another.

This is crucial for Facebook, whose mission statements have long stressed global interconnectedness. In 2009 its statement was, “Facebook gives people the power to share and make the world more open and connected.” This was updated in June 2017 to, “Give people the power to build community and bring the world closer together.” In between, in 2012, the letter attached to Facebook’s IPO paperwork said that “There is a huge need and a huge opportunity to get everyone in the world connected,” but that this is based on “the relationship between two people” because “Personal relationships are the fundamental unit of our society.” The letter talks specifically about the bottom-up nature of the social change Facebook will bring about: “By giving people the power to share, we are starting to see people make their voices heard on a different scale from what has historically been possible. These voices will increase in number and volume. They cannot be ignored. Over time, we expect governments will become more responsive to issues and concerns raised directly by all their people.” We could also note the banner image on Mark Zuckerberg’s own Facebook page: a map of the world lit up by interactions between Facebook users.¹⁴

The facebook.com/peace page thus appears to offer empirical support for long-held beliefs about the potential for peace inherent in communication technologies. For instance, Charles Frederick Briggs wrote about the telegraph in 1858 that “It is impossible that old prejudices and hostilities should longer exist, while such an instrument has been created for an exchange of thought between all the nations of the earth” (Briggs & Maverick, 1858). Echoing this, Sheryl Sandberg called the peace page “profound,” and said, “Is it harder to shoot at someone who you’ve connected to personally? Yeah. Is it harder to hate when you’ve seen pictures of that person’s kids? We think the answer is yes.”¹⁵ The IPO letter even cites “inventions like the printing press and the television,” stating that “They brought us closer together.” The peace page seemingly provides evidence that this is working, that grassroots connections are being forged between Facebook users, even if political leaders are not talking to one another.

So Facebook publishes a page with dubitable data. Who cares? Are there not bigger fish to fry? Probably, but this case nonetheless contributes to broader discussions about trust, social media, their opacity, and how we can try to establish the veracity of numbers that Facebook et al. publish in the context of the quantification and datafication of social and cultural life, and given the centrality of social media in the lives of so many (Gerlitz, 2016; Kennedy, 2016).

Some of the wider issues at stake can be seen in the following: In September 2016, Facebook was involved in a brouhaha regarding “its metric for the average time users spent watching videos,”¹⁶ which were erroneously reported as being higher than they actually were. Facebook apologized, corrected the error, and rolled out new metrics.¹⁷ Also, it was reported that Facebook claimed that its ads could reach 10m more 18- to 24-year-olds in the United States (41m) than there are 18- to 24-year-olds in the United States, according to census data (31m).¹⁸ In these instances, the numbers provided by Facebook are problematized, and questions are asked about their production. These, though, are not instances of bullshit. Facebook responds to and fixes problems in its metrics, because the trust of advertisers is of critical importance to its business model. The referential accuracy of the data is what gives them value; without it, Facebook loses money.

When it comes to data about cross-conflict friending, though, it would not seem to care, suggesting that we have here a different kind of data, namely, bullshit data. Here, what counts is the data’s semiotics, or their character as “material-semiotic objects” (Maldonado Castañeda, 2015). As Helen Verran (2013) argues, numbers can be used performatively, but in the current case, it appears that there is nothing but the performance. An important aspect of this performance, and part of what makes the page seem credible, is the precision of the numbers. Facebook does not say that approximately 200,000 friendships were formed yesterday between Israel/Palestinian Territory; Facebook tells us that precisely 195,435 such friendships were formed. We assume that Facebook can know this because we know they know so much about us, and the precision even when dealing with such large numbers contributes to the effect of its credibility. Indeed, in their book on numbers and policy, Funtowicz and Ravetz (1990) suggest that “[d]ata expressed as a lengthy string of digits” can produce a “spurious appearance of accuracy” (pp. 10–11); they call this “hyper-precision or even pseudo-precision” (p. 11). By not rounding the numbers up or down, the large and precise numbers on the peace page enjoy their appearance of accuracy. Borrowing from Andrejevic, Hearn, and Kennedy’s (2015) discussion of Wernick (1991), we might say that the numbers on the peace page “prey parasitically on modes of human representation” (p. 381) with which we are already familiar.

The relevant mode of representation here is that of “metric culture” (Ajana, 2018; Beer, 2016), which itself builds on “[s]trict quantification, through measurement, counting, and calculation”; according to Porter (1995), these are “among the most credible strategies for rendering nature or society objective” (p. 74). A feature of metric culture, according to Kernighan (2018), is that it renders people “number numb,” and indeed this may account for the uncritical usage of the peace page in peer-reviewed articles. From this perspective, the current article is an attempt to restore some sensitivity to our responses to numbers, which may be helpful in reading other data pushed to us by social media platforms. While scholars critical of APIs (see, for example, Lomborg & Bechmann, 2014; Rieder, 2016) have pointed to their opacity, and have raised issues to do with how APIs leave much of the database out of reach (John & Nissenbaum, 2018), and how the companies publishing them can change them at will and without forewarning, the question of how accurate the data are that social media companies deign to release is rarely raised. Here, though, that is precisely the question: what are these data? Where are they coming from? What do they signify? And most importantly, how can we know?

These questions—which have very broad application—have stood at the center of this article. My attempts to answer them shed light on a different kind of relationship between social media companies and researchers (and others) who are interested in them. When user metrics are misreported, Facebook admits it and they get fixed (at least sometimes); when bots impact on social media, the relevant companies (eventually) engage in some kind of discussion, however frustrating that may be. The issue with facebook.com/peace, however, is frustrating in a different way, because Facebook really could provide data about cross-conflict friending and unfriending. When metrics go wrong, we can accept that sometimes large companies can make mistakes; we may even accept that social media companies are reasonably protecting their IP and trying to prevent further abuse by not sharing everything they know about bots. But why make up data about Facebook friendships, or refuse to explain why the numbers are to be believed? And if these data are bullshit, then what about other data? When Facebook says it has 2bn users, should we believe them?

To end, let us try to define social media bullshit, or at least posit some of its characteristics, with the hope that future work will offer refinements. Accordingly, I propose that social media bullshit is (1) bullshit produced and published by a social media company (2) in an opaque manner; (3) it refers to a social media activity (in this case, friending); (4) it appeals to the datafication of social life by social media companies and (5) it makes explicit or implicit claims as to the importance of social media in social processes; (6) it draws on the semiotics of big data by taking the form of large, precise numbers that seem up to date; and finally, (7) it comprises data on which the publisher refuses to comment, reflecting the asymmetrical power relations between company and researcher.

The objective of this article is not to cast doubt on all quantitative statements made by social media companies. It may be technically difficult, but I believe that Facebook could tell us how many Friendships were made “just yesterday” between Israelis and Palestinians. (In fact, if they thought it would increase spending on the site by advertisers, they probably would.) The objective of this article, rather, has been to posit a new category of data-based claims—the category of bullshit—and to offer some ways of testing whether certain data are indeed bullshit. When we live our lives in media, and when we are reliant on those media to help us attribute meaning to those lives, then I believe there is value in calling bullshit on attempts to shape our perceptions of those media.