Laughing at death: Facebook, the ‘haha’ reaction, and death coverage on local US news pages

ABSTRACT

Is it inappropriate to “like” a Facebook post announcing death? That was the concern of Meta’s leadership in 2016, when they implemented emoji reactions (e.g. sadness and love) to help users more appropriately respond to tragic events. The expectation was that these new reactions would promote compassion in the face of grief, but in the years since many observers now label the “haha” laughing emoji as a particularly menacing reaction used to troll the misfortune of others. To date though, little academic work has sought to explore such claims. Across hundreds of millions of reactions to death on the platform spanning nearly a decade, this paper demonstrates the size and scope of the haha trolling phenomenon on Facebook—including its exponential growth during the pandemic. Using the automated tools of natural language analysis, this study also sheds some light on the potential motivations behind this socially deviant behavior.

KEYWORDS:

• Facebook
• haha reaction
• laughing emoji
• death trolling
• RIP trolling

Introduction

In 2014, during his second town-hall-styled Q&A event, Mark Zuckerberg was asked about adding a dislike button to Facebook. The logic of the question was quite simple. The like button was an overwhelming success, so why not give people the opportunity to express the opposite response? In fact, the question itself was so overwhelmingly popular that Facebook flew the questioner to the Q&A to ask the CEO in person.

In hindsight, Zuckerberg’s (2014) answer to the dislike question was quite insightful. He claimed two immediate concerns. On the one hand, the company was uncomfortable with creating an upvote/downvote system, similar to other websites: as he saw it, “that’s not something we think is good for the world.” But on the other hand, Facebook recognized users desired a more diverse set of emotional responses. The example he gave as justification features as the focus of this paper:

“people will tell us they don’t feel comfortable pressing like, because pressing like isn’t the appropriate sentiment when someone lost a loved one” (Zuckerberg, 2014)

In early 2016, Facebook introduced a new set of emotional reactions to complement the well-established like button. The five reactions—love, wow, haha, sad, and angry—theoretically allowed the platform to capture a more diverse set of appropriate reactions, something Zuckerberg identified months earlier. Nearly half a decade later, Facebook continued to fine-tune this emotional reaction set by installing an additional engagement button, care. Added largely in response to the pandemic, the company suggested users (again) demanded the new reaction to fill a perceived gap in compassion on the platform (Guynn, 2020).

These two examples point to an interesting relationship between socially appropriate (emotional) engagements online and unfortunate events like death. While cultures respond quite differently to the death of loved ones, we know emotional reactions to grief such as sadness are largely universal (Rosenblatt, 2007). Meaning, Facebook’s repeated attempts to guide its users to emotionally react to death in socially acceptable ways introduces new options for individuals to violate those same normative expectations.

Put another way, these reactions may offer new tools for expressing mild deviant behavior online—a practice of deliberately violating social norms online, something academics have been charting since the earliest days of internet-based social communities (Suler & Phillips, 1998). While violating norms on social media in some cases may be relatively harmless, online bullying research demonstrates the very real-world consequences of such deviant behavior (Hinduja & Patchin, 2010) for both victims and perpetrators (Hay, Meldrum, & Mann, 2010). In turn, whether actively trolling the death (and grief) of others online, or merely using emoji reactions on Facebook to mock the pain of others, these new reactions on Facebook offer the space to grow—and potentially reinforce—problematic behavior.

Beyond these theoretical concerns, journalists and pundits claim to have observed such mild deviant behavior expressed via these new reactions on Facebook, particularly the use of the haha reaction in socially inappropriate contexts. In fact, the practice of using the laughing emoji to troll others has led some to insist the platform remove the haha reaction altogether (Driver, 2021). To date though, no academic study has sought to document nor explore this alleged behavior.

This paper addresses just such a need. Across roughly 4.4 million Facebook posts, and 800 million reactions spanning nearly a decade, this study investigates the use of the haha reaction in response to death news on US local news pages. As the data indicates, the laughing emoji does indeed appear—increasingly—in these inappropriate settings. Using natural language processing automated tools, this article further examines the types of death news that prompt the haha reaction, and the potential theoretical motivations for this behavior.

Before proceeding to that evidence, though, this paper now turns to the theoretical foundation upon which the current study is developed.

Literature review

As the bulk of thanatology research demonstrates, bereavement sparks a wide array of differing behavioral responses across cultures. For example, while sadness and grief permeate the trans-cultural experience of death, expressions of these emotions vary considerably (Rosenblatt, 2007). Funeral processes perhaps best reflect these immediate differences: where some indigenous cultures may offer multi-day community events and rituals to help the dead make the afterlife journey and release the grieving back to life (Nikora et al., 2010), others observe much more secularized and commodified funeral practices emerging in the West (Walter, 2005).

Funerals also represent the clearest example of the emotional stimulus of death, though the power of that stimulus can vary. For example, it is well accepted that the death of others (e.g. strangers) generally reduces death’s emotional potency (e.g. Marti-Garcia et al., 2016). Even still, limited familiarity is not always such a barrier, as the public sharing of collective grief (with strangers) during tragedies is another well-observed phenomenon (Neimeyer, Wittkowski, & Moser, 2004). The many environments for acknowledging the death of others (e.g. friend/work networks, social media, obituaries in newspapers, etc.) offer new forums for such potential behavioral responses to emotions prompted by death. Among this research, media representations of death (Barnes & Edmonds, 2015; Hanusch, 2008; Walter, Littlewood, & Pickreing, 1995), and responses to it, feature as the most relevant collection of literature to the current paper.

This journal’s readership needs little education along these lines. Media unquestionably has the potential to provoke emotions and, in turn, behavioral responses. Exposure to death via media is no different. As a great deal of work shows, entertainment and news media depictions of death can indeed prompt audience emotional responses. From affecting audiences into charitable donations (Maier, Slovic, & Mayorga, 2017), and promoting life-saving behaviors (see discussion in Brown & Basil, 2010), to inflicting moral panic (Burns & Crawford, 1999), death in media can produce an array of responses. In fact, some research claims that our very understanding of the causes of death itself is warped as a result of distorted news media coverage of death (Frost, Frank, & Maibach, 1997). Paradoxically, some literature even suggests that too much (media) exposure to death—and similar—stimuli actually emotionally desensitizes audiences, instilling a compassion fatigue that audiences and journalists supposedly experience (Moeller, 2002). Regardless of these debates, though, the collective body of evidence from this field reinforces the emotional potency of death representations in media, and the power such mediated emotional stimuli offer in provoking behavioral responses.

Social media represents a (relatively) new sphere in which individuals are exposed to and respond to death. In turn, these social media platforms offer new spaces to assess the differing behavior of online grievers (Walter, 2015). To date, most explorations of death engagements on Facebook specifically involve the memorialization of user profiles (Brubaker, Hayes, & Dourish, 2013). Among this research, maintaining an online connection with deceased friends and family via their online profiles features as the main focus (e.g. see Pennington, 2013). While that offers some insight into emotional responses to death on Facebook, such work does little to explain how individuals might respond to the death of strangers on Facebook.

This knowledge gap is particularly important, given the Facebook environment affords academics the option to inspect emotional responses to the death of others at a mass level. As established in the introduction of this manuscript, the social media giant recognizes the utility of catering to differing emotional responses on the platform. To draw from a new example, we know that Meta has used Facebook emoji reactions (e.g. love, care, haha) as algorithmic metrics to promote content among its users (Merrill & Oremus, 2021). While emoji reactions are subject to a number of different motivations (e.g. solidarity, sarcasm, mocking, grief, etc.), we know these engagements can still offer important insights: for example, emoji reactions can not only tell us generally about user emotional responses, but also can—roughly—predict post content in certain contexts (Kuo, Alvarado, & Chen, 2018). Take misinformation spread as another example. Internal Facebook research demonstrates the correlation of misinformation with the angry reaction on their platform (Merrill & Oremus, 2021). Variations in Facebook reactions have also been used to differentiate between post topics in other contextual domains (e.g. Freeman, Alhoori, & Shahzad, 2020). Given the focus of the current study, potential differences in Facebook reactions offer the prospect for new insights into human emotional responses to death online.

Much of the previous discussion largely assumes, be it on or offline, that emotional responses to death are genuine and empathetic. There are, of course, other possible responses to death worth considering here. For example, we know that laughing at the (imagined) pain of others is a staple of slapstick comedy: research here suggests the facial expressions of victims provide the necessary context to prompt laughter among audiences (Manfredi, Adorni, & Proverbio, 2014). Such lighthearted reactions to misfortunes clearly run counter to the empathetic responses identified above, but there are more (socially) problematic responses. To borrow a phrase from Bishop (2014) there is also a “darker, sinister and transgressive side of cyberspace in the form of abuse and vitriol (i.e. anonymous trolling)” that we should also consider.

“RIP trolling” presents one such instance. Here online users are known to deliberately provoke and abuse mourners on social media memorial pages, including on Facebook (Phillips, 2011). A number of motivations exist to potentially explain this behavior. Where some speculate that trolls extract enjoyment out of these activities (Bishop, 2014), others describe it as a service to calling out “grief tourists” (i.e. inauthentic grievers; see Phillips, 2011). Motivations aside, the trolling of grievers itself represents a clear divergence from socially expected behavior.

Trolling, whether “RIP” or otherwise, historically relied on direct interaction to achieve the desired result. That is, such efforts manifest in replies on social media, or in the comments section on news stories. As a result, some news organizations routinely turn off comments on stories likely to attract similarly unpleasant engagements (Green, 2018).

Facebook reactions, however, feature as a less confrontational and less time-intensive method of trolling the pain and misery of others online. Perhaps for these reasons, journalists and public commentators claim to observe the deliberate use of Facebook’s laughing emoji, haha, as a new means of mocking others (Ghaffar, 2018; Walters, 2018). To date though, little academic attention has been directed at exploring the haha reaction as a death trolling mechanism. Research in related areas nevertheless hints the emoji has been adopted in other trolling endeavors. Al-Zaman and Ahona’s (2022) investigation of Facebook reactions to rape in Bangladeshi news offers one such recent example, where they observed “haha” made up over a quarter of all such reactions.

The COVID-19 pandemic introduces new avenues for RIP trolling to take place online. The Herman Cain “award” represents the most popular example: a facetious prize dedicated to individuals who publicly rejected the dangers of COVID only to succumb to the virus itself (Barth, 2021). Here entire subreddits were dedicated to publicly mocking such individuals, reflecting a clear evolution of RIP trolling during the pandemic.

In contrast, those who reject the lethal nature of the coronavirus may also see opportunities to troll COVID-related death content. Take conspiracy theorists who believe that COVID-19 is a hoax or a “plandemic” (Kearney, Chiang, & Massey, 2020), such individuals might theoretically respond to the news of COVID deaths in the community with disbelief, rejection, and consider the news worthy of derision. Partisanship, particularly in the US, may prominently feature in this mix as well. Polling has consistently shown Republican supporters are more likely to downplay the risks of the virus and believe the US government and media institutions exaggerate infection and death rates (e.g. see Kaiser Family Foundation, 2021; Pew Research Center, 2020). As a consequence, the trolling of COVID death news online theoretically seems plausible both from individuals who believe in the severity of COVID-19 and those who dismiss it. These similar potential reactions to COVID-related deaths—despite dichotomous beliefs—therefore hint at the natural experiment promise of research in this space during the pandemic.

The prior discussion has, so far, offered insights into the motivations for mocking the pain of others in relation to death. Beyond these motivations, there are also real-world consequences of such deviant behavior worth detailing here. For example, a host of cyberbullying literature demonstrates detrimental outcomes spanning both the mental (Tian, Yan, & Huebner, 2018) and physical (John et al., 2018) well-being of victims (e.g. increased anxiety and depression to self harm, suicidal thoughts, and substance abuse). In fact, perpetrators of cyberbullying also share some of these concerning outcomes with their victims (Kowalski et al., 2014). Though some may differentiate online trolling from cyberbullying in terms of severity, such debates nevertheless generally accept both forms involve some degree of malevolence and bullying (March & Marrington, 2019).

Viewed through the lens of these negative outcomes, a study of new forms of death trolling via Facebook reactions becomes all the more valuable. In other words, the Facebook reaction set theoretically presents new methods by which users can bully the grieving of others, or be bullied themselves, offering the potential for real-world negative outcomes similar to those observed in other environments. While those potential negative outcomes do not feature as this study’s current focus, the prospect of these effects demand the field first documents the presence of the alleged mild deviant behavior online.

As this literature review has shown, there exists a real need to investigate how users respond to death news online. Such a contribution would strengthen our current knowledge of death responses on social media (e.g. beyond memorialization), and broaden our narrow understanding of how Facebook reactions may be used both as a means of trolling and (politically motivated) resistance to death news. Beyond these immediate gaps, the prospect of a natural experiment presented by the pandemic, and the potential for real-world harms to emerge as a result of this behavior provide even further incentives for study.

The current paper addresses these needs by exploring Facebook haha reactions to death news. Given the prior discussion, the study presents the following two research questions:

RQ1:

Does death news prompt haha reactions on Facebook?

This literature review has outlined a number of reasons why we might expect to see haha reactions to death news on Facebook. Public commentators have offered anecdotal evidence along these lines, and the aforementioned academic literature indicates that these death trolling activities should be observable. To date though, no study has documented such a phenomenon. That is the objective of this first research question.

RQ2:

If RQ1 reveals the expected behavior, what type of death news prompts more haha reactions on Facebook?

From the pandemic, to slapstick comedy literature, and RIP trolling, it is clear that responses to death occur in a variety of different contexts. However, no study to date has sought to identify the types of death news that tend to spark greater levels of trolling—should it exist. Answering this question would provide greater clarity about the context in which this trolling takes place online, and serve as a foundation for further investigations into this potentially problematic behavior.

Methods

In order to examine Facebook reactions to death news, this paper collects data from CrowdTangle, the Meta-owned and approved data extraction API. More specifically, this study draws from the CrowdTangle-identified, US-based, local news Facebook pages.

US local news pages were chosen for three reasons. First, journalists have observed comments sections on their stories as spaces for trolling activity, as previously discussed. Meaning, news presents the most likely venue to observe such behavior. Second, coverage of deaths in the community may prompt slightly stronger emotional responses given proximity, in contrast to national-level reporting where a death may occur quite literally on the other side of the country. This is not to say that local news does not report about deaths at a national or international scale (e.g. war and conflict elsewhere), but arguably local news content nevertheless provides a marginal methodological advantage over national level coverage. Finally, the US’ well-documented struggles with COVID (both politically and otherwise) make it an ideal focus for study given the natural experiment property discussed previously.

Using the query “death OR dead OR died” this study extracted over 4.4 million posts on Facebook from February 25, 2016 (when reactions were implemented) to March 23, 2022. Those posts provoked over 800 million engagements, arising from nearly 6 thousand local news pages in the United States.

To differentiate between the various types of death news that spark these reactions, the content for each post was first assembled and then classified via automated natural language processing tools. In terms of content assembly, the Facebook post’s text and other textual descriptions of (any) images and links were combined and processed using standard language parsing procedures (e.g. removing non-alphabetical characters and stopwords, converting to lowercase, etc.)

For classification of post content, this study utilized the automated topic machine learning algorithm LDA (i.e Latent Dirichlet Allocation) to categorize the content at a mass scale. Stanford’s MALLET (Machine Learning for LanguagE Toolkit) was used to produce the coded dataset. To clarify, MALLET’s LDA implementation classifies documents (i.e. Facebook posts) into a predetermined number of categories (i.e. topics) based on the words that exist within each document. To arrive at the appropriate number of topics, optimal coherence (UMASS) and log likelihood harmonic means were first identified across nearly forty preliminary models. These diagnostics indicate models that offer the best statistical fit for the data—in this case, five, six, and eight category models. Next, the ideal model was chosen subjectively by comparing these diagnostic-identified potential models (e.g. by determining how the algorithm sorts the same data with a larger number of available topics). This qualitative comparison revealed the larger topic models separated Facebook posts on seemingly less relevant categories (e.g. parental relationships to the deceased and community participation) making them less appealing for the current study. The final topic model, consisting of five categories selected via these diagnostic procedures, was compiled over 50 thousand iterations.

To answer both research questions and to further explore the data, the analysis is conducted in the following sequence. To begin, the results section outlines the broad structure of Facebook reactions in response to local death posts over time, which will address the first research question. Next, the topics—identified by the automated analysis—are visually expressed via a comparison cloud, outlining the general types of death-related content on Facebook within this dataset. Third, the analysis presents a time series of haha reactions split by topic, exposing potential relationships between specific topics, time, and death trolling reactions on the platform—addressing the second research question. To finish, the results section then offers a brief qualitative interpretation of the data, helping to provide a bit more insight into the nature of the findings presented within.

Results

Figure 1 displays the daily accumulation of over 800 million Facebook reactions to death posts from 5.5 thousand US local news pages, across nearly seven years of data. Fluctuations in these daily reactions are interpreted via locally estimated scatterplot smoothing curves. The figure therefore visually represents the daily expressions of approximately 386 million sads; 231 million likes; 83 million angries; 58 millions wows; 32 million loves; 14 million cares; and 11 million haha reactions to death news (i.e. addressing RQ1).

Figure 1. Reactions to Facebook death posts over time.

Beyond sheer quantity, the figure clearly depicts inverse relationships between reactions: that is, where some become more fashionable over time, others decline as a result. Take the introduction of emoji reactions in February, 2016. Here the data indicates “like”, the most prominently used reaction on Facebook at the time, is quickly replaced by “sad” as the preferred reaction to death news within a twelve month period. Similarly, likes in response to death posts continue to decline until mid-2018, as other reactions grow (e.g. angry, wow). The same type of inverse relations occurs with the introduction of care reactions in early-2020, which coincides with the decline of sad, like, angry, and wow.

The solid magenta line in Figure 1 depicts the daily number of haha reactions to death posts during this time period. While comparatively far fewer in quantity (in contrast to other reactions), the laughing emoji still managed to accumulate roughly 11 million reactions over those seven years. The figure also reveals two clear differences between haha and other reactions. First, haha appears not to possess an inverse relationship with care reactions. In fact, a general additive model regression from early 2020 to 2022 reveals every reaction type—except haha—is statistically related to care (i.e. p > .05; see Table 1 in Online Supplement). As a result, the laughing emoji exhibits an unusual apparent independence, unlike other Facebook reactions.

Second, the haha reaction sees heightened popularity at a far later date than most other emojis. The red dots in Figure 1 indicate the top of the LOESS estimated curve for each reaction (i.e. a day where the rolling average reaches its maximum value). For reactions such as like, the top of the curve starts when reactions were first introduced (i.e. February, 2016). Other reactions see a maximum at some point prior to mid-2020 (e.g. sad: February 2020; angry and wow: November 2019; loves: March 2020). In contrast to those theoretically socially appropriate reactions, the rolling average for haha reactions peaks in November 2020. In other words, the laughing emoji appears most popular much later than other reactions, with the exception of the newly created care engagement.

Figure 2 presents a comparison cloud for the LDA generated topic model. Words in these clouds reflect differences between death post topics; meaning, a word is used more frequently in one topic over another, with color separating the topics. The size of words, and their placement, in the comparison cloud further represents this difference between topics. That is, larger words closer to the center indicate an increased difference between the frequency of word use in topics.

Figure 2. Death topic model comparison cloud.

Diagnostics suggest that five general topics best fit death content on Facebook local news pages in the United States. Coverage of famous death(s) represents the first topic identified by the algorithm. Here post content involves death announcements for celebrities, artists, sports stars and their upcoming funerals—consisting roughly 23% of total death related posts. The next topic, disasters, involves news of death associated with heatwaves, hurricanes, tornados, floods, bridge collapses, etc., representing 14% of the sample. The third topic, legal proceedings, consists of reports on prosecutors, charges, pleas, judges, lawsuits, etc.—making up 28% of the coverage. Crime scene news, such as car crashes, shootings, police barricades, etc. represents 21% of the topics. Finally, disease coverage made up 14% of death news reporting. This final topic is dominated by COVID-19 pandemic reporting, but also contains discussion of other outbreaks (e.g. West Nile, E. coli, opioid overdoses, etc.). Collectively these five topics therefore (generally) represent all the types of death news that exist on local news Facebook pages in the US.

The (approximately) 11 million haha reactions to death news were distributed among these five topics as follows: 15% famous deaths; 17% disaster; 9% legal proceedings; 23% crime; 37% disease news. The allocation of laughing emoji reactions to these topics clearly differs from the proportion of coverage each topic received. For example, despite being one of the least covered topics (i.e. 14% of all death-related content), disease posts attracted nearly 4 million total haha-reactions. Crime news also attracted a notable collection of roughly 2.5 million haha-reactions. As a result, the data quite clearly indicates that disease coverage, and to some extent crime news represent topics that collectively attract the bulk of haha-reactions (i.e. addressing RQ2).

In order to visually address RQ2, Figure 3 depicts the quantity of laughing emoji reactions each of these topics received (per day) on Facebook. Much like what was observed in Figures 1, 2 shows the growth of haha reactions to these topics over time. LOESS estimates suggest that most peaks for haha reactions occur prior to mid-2021 (e.g. disaster: April, 2021; crime scene: September, 2020; legal proceedings: November, 2020; disease: December 2020; famous death: March 2021).

Figure 3. Haha Facebook reactions to death by topic.

While all the peaks for the topics appear to reflect moderate increases in daily haha reactions around a specific time period, none mirror the exponential growth that disease death news received between 2020 and 2021. This growth most obviously coincides with the death toll rising during the COVID-19 pandemic. Prior to the pandemic, it appears that crime scene reporting and natural disasters represent the next two largest (earlier) peaks associated with death news.

Beyond purely automated analysis, a brief qualitative reading of the top haha-attracting posts within these classified topics can also help infer some of the motives behind deviant haha reactions to death news. For example, as expected, competing (political) perspectives of COVID-19 and vaccinations operated as one potential motivating factor. Notably, three of the top ten haha-attracting posts in the disease category involved news of a Stanford University study that claimed tens of thousands of President Trump’s followers caught coronavirus and hundreds died as a result of attending the candidate’s rallies during the 2020 presidential election. In contrast, one top ten post in the famous death category covered a vaccinated woman’s family who blamed her COVID-related death after exposure to the unvaccinated (likely classified based on obituary style language). In both cases politics, competing views of the virus, and vaccines all likely feature as potential motivations for this behavior.

Laughing at the misfortune of others also appeared prominently in the results. One top ten post in the legal proceedings category featured a “drive-by gone mad,” where a passenger accidentally killed their driver during an attempted drive-by shooting. Laughing at criminal attempts gone wrong presents a (perhaps) more lighthearted amusing reaction to laughing at the pain of others—much like the prior slapstick comedy discussion. In fact a number of top haha-attracting crime posts echo similar instances where perceived criminal wrongdoing results in death: e.g. attempted carjacker crushed by vehicle; ATM thief dies by explosion; man’s attempt to jump subway turnstile results in fatally breaking neck, etc.

While the previous examples hint at irony as a driving motivation for the laughing emoji, there were also notably more problematic instances worth further exploration. To best explain these potentially problematic instances, the author examined the top 10% haha and care reacting posts, about 15 thousand posts illiciting both high levels of care and amusement in response to death. Like the COVID-related news above, tragedy inflicted on seemingly innocent members of political opponents was clearly present: prompting haha emojis in response to the news of Chrissy Teigen’s newborn’s passing, discussions of Beau Biden’s death from brain cancer; and President Trump’s brother Robert’s fatal stroke.

Laughing at the tragedy inflicted on marginalized communities also features prominently among these 15 thousand isolated posts. For example, the LGBTQ+ community saw haha reactions in response to posts about the Pulse Nightclub shooting and the death of a trans migrant after surviving “rape, kidnapping and assault.” Similarly, characterizations of the killing of gay, trans, Black, and Asian people as “hate crimes” were also present in this collection of top care and haha emoji-attracting posts.

Posts about fatal police shootings also sparked high levels of sympathy and amusement. Discussions of George Floyd were the most apparent example (similarly visible in the word cloud), but a brief survey revealed at least 15 names of other African Americans victims within that isolated data (e.g. Trayvon Martin, Breonna Taylor, Botham Jean). Of those 15 thousand high-care and high-haha attracting posts, over 20% mentioned one of those 15 names. In fact, some of the comments to those posts explicitly asked why others were laughing at the fatal police shootings.

Unquestionably, motivations for these responses involve context, but having surveyed the data it seems at times racism, retribution, and other concerning factors may all partially explain some of the observed behavior.

Perhaps the most obvious problematic, apolitical, example of death trolling via the haha-reaction involved historic video footage. This particular instance, a top ten haha-attracting post in the disaster category, covered the 35th anniversary of the Challenger Shuttle explosion. That post showed video footage of Christa McAuliffe’s parents—in tears—watching their child’s death in real-time as the space shuttle exploded overhead, eliciting over 7 thousand haha-reactions to date.

Where other uses of the haha reaction conform logically to several probable motivations, the Columbia accident clearly differs. Here potential justifications via politics, racism, or reprisals for misdeeds all fall short. Laughing at footage of parents watching their child die in a horrific public accident quite jarringly mirrors the potential RIP trolling motivations described above. Indeed a handful of the post’s comments hint at this problematic motivation: from retelling old jokes about McAuliffe (i.e. “Did you know that Christie McAuliffe had blue eyes? One blew this way…one blew that way”); to remarking about the explosion as “great fireworks”; and questioning the sincerity of McAuliffe’s parents as “actors” in a conspiracy.

Moving beyond potential motivations for the moment, this paper now turns to discussing these results within the context of the research questions.

Discussion and conclusion

This study examined the possible use of haha reactions in response to death posts on Facebook. Having developed a better theoretical understanding for this potential behavior in the literature review, this article offers two primary research questions to explore the phenomenon online. Across a sample of 4.4 million Facebook posts, 800 million reactions, spanning nearly seven years, the results indicate two clear answers to the proposed research questions.

RQ1:

Does death news prompt haha reactions on Facebook?

As previously established, journalists and public commentators have both claimed the laughing emoji is used on Facebook in troll-like responses to news. But to date little academic work has tested this claim, specifically within the realm of death related content. The evidence from the current paper does just that: as observed within, death posts from US local news since 2016 have prompted nearly 11 million haha reactions on Facebook. In fact, it appears use of this reaction in response to death is increasing over time.

RQ2:

If RQ1 reveals the expected behavior, what type of death news prompts more haha reactions on Facebook?

Using natural language processing tools, this study classified all death related posts on Facebook, US local news pages into one of five topics. Those topics include famous deaths, disasters, legal proceedings, crime scene coverage, and disease news. Among the approximately 11 million total haha reactions, a majority of these laughing emoji responses to death came from disease coverage (~4 million) and crime (~2.5 million). As time series data reveals, laughing emoji responses to disease-related posts exponentially increased during the pandemic. In other words, while the data shows a variety of death-related topics can prompt the haha reaction, COVID-19 coverage features as the most prominent and recent driver of this behavior. Increased use of the haha reaction in response to other death topics also occured during the pandemic as well, hinting perhaps that the behavior became normalized to some degree during the crisis.

While not specific to RQ2, the author also conducted a brief qualitative assessment of the top 10% posts that illicited both high sympathy (i.e. care emojis) and haha responses. The qualitative results indicate some other complex motivations for the haha response. For example, as speculated in the literature review, laughing at the misfortune of others appears to be one major motivation. This is perhaps best exemplified by the diametrically opposed views on the COVID-19 pandemic, where some laughing emoji reactions are deployed to criticize perceived irresponsibility (e.g. covid-related deaths associated with attending political rallies during a pandemic) others utilize the same response to express apparent anti-vaccination positions (e.g. laughing at the COVID-related death of a vaccinated women whose family blames the unvaccinated).

The brief qualitative reading also shows potentially more problematic motivations behind the haha response. For example, there are clear instances where the death of children, or those from marginalized communities attract an unusual amount of haha reactions. In fact, news of African American deaths from police shootings (or interventions) constituted over 20% of the high sympathy (cares), high amusement (hahas) sample. Results like this suggest, if nothing else, there is considerable reason to investigate the haha-emoji phenomenon in much greater detail.

Future research in this space, for example, might use survey experiments to produce more insight into the motives and environments in which this behavior is viewed acceptably by individuals, though identifying users that exhibit such responses may be a challenge. We also know little about the possible detrimental effects of such emoji reactions. Again, cyberbullying research clearly shows negative outcomes both for perpetrators and for victims, and so this new method of trolling the pain of others represents an important area for similar research.

Given the apparent growth of this phenomenon, particularly during the pandemic, academics should test whether these engagements have knock-on effects in other relevant arenas (e.g. empathy, social cohesion, support for institutions, political polarization, etc.). There is also reason to believe this use of the laughing emoji as a death-trolling tool may continue to grow beyond the confines of the pandemic context. The fact that over four million haha reactions were used on death-related content years prior to COVID19, and the laughing emoji’s apparent statistical independence from other engagements, both suggest that death-trolling via Facebook reactions is at least an ongoing—if not growing—problem. As a result, this paper’s results offer a call for a deeper level of understanding related to insincere death-related reactions on social media and the potential consequences of this behavior.

A final word from Mark Zuckerberg’s, (2014) Q&A is particularly relevant here in closing. Nearly a decade ago Facebook’s co-founder recognized the need for users to have a more diverse set of reactions on the platform, lest they inappropriately respond to posts announcing the death of loved ones. He suggested caution was needed before pressing ahead with any changes to the reaction set, though:

“[But] we need to figure out the right way to do it, so that it ends up being a force for good and not a force for bad and demeaning the posts that people are putting out there.” (Zuckerberg, 2014)

The data here shows that a vast majority of users adopted reactions like care, and sad in response to death—much like Facebook expected. This socially acceptable “force for good,” however, is clearly accompanied by more problematic behavior. Across millions of “haha” reactions, the results from this paper hint that those same demeaning motivations that Zuckerberg feared have also now been realized.

Disclosure statement

The author is a member of a team that has been granted a Facebook research award for a separate project. Those funds were not used for the current paper, nor did the award influence this study in any way.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15405702.2023.2287738

Additional information

Notes on contributors

Justin Bonest Phillips

Justin Phillips is a political scientist and Senior Lecturer in the School of Social Sciences at the University of Waikato, New Zealand. He specializes in political communication research, particularly on social media utilizing big datasets.