Unpacking uncivil society: Incivility and intolerance in the 2018 Irish abortion referendum discussions on Twitter

For the purpose of this study, we explore the 2018 Irish abortion referendum discussions on Twitter to look at incivility and intolerance around abortion policy discussions. Abortion discourse is a suitable topic to look at diverse incivil and intolerant remarks, as it is a sensitive and highly polarised sociopolitical issue. Also, this referendum is an appropriate choice to examine online incivility and intolerance, since social media platforms such as Twitter have served as sites of negative language, used by both pro- and anti-choice sides (Sweeney, 2018). There were accusations during the referendum that numerous “trolls” were undermining the integrity of the referendum, which caused social media platforms such as Facebook to ban foreign accounts outside Ireland that were spreading dark ads (Hern, 2018). Twitter introduced new measures to fight spam, abuse, trolls, and hateful conduct, suspending an average of 17,000 accounts per day in May 2018 (Roth & Harvey, 2018).

To examine incivility and intolerance in the referendum discussions, we use a Twitter dataset shared on Harvard Dataverse (Littman, 2018). This dataset contains 2,279,396 tweets collected between 13 April and 4 June 2018. The rehydrated dataset contains 1,842,370 tweets; the lost tweets are due to how Twitter APIs and platform works.

For official information, see Twitter developer at https://developer.twitter.com/en/docs/tweets/post-and-engage/api-reference/get-statuses-lookup

For ethical social media research, this article does not directly quote any tweet in the sample. Due to the searchability of Twitter, we paraphrase tweets in order to prevent the tweets from being reverse searched and the author being identified. However, such paraphrasing is minimal so as not to distort the unique utterance of incivility and intolerance.

This study employs systematic content analysis (Neuendorf, 2002) with the codebook (available upon request) developed by two researchers operating on two units of analysis: users and tweets. Users were coded in terms of gender (female, male, unidentifiable) and their political position vis-à-vis abortion (pro-choice, anti-choice, or unidentifiable). The gender of the users was coded based on their first name in their profile. This has a limitation in that some names are gender-neutral and some users use gender-unidentifiable pseudonyms (e.g., Old Glory). In such cases, their gender was coded as gender unidentifiable.

Political position was coded in terms of the hashtags used in the tweet as well as the overall argument of the tweet. The list of hashtags includes the following: 32 pro-choice (e.g., #trustwomen, #repealthe8th, #Together4Yes, #voteyes); 14 anti-choice (e.g., #JoinTheRebellion, #savethe8th, #loveboth, #lifecanvass); and 7 neutral and ambiguous hashtags (e.g., #8thref, #HomeToVote, #8thamendment, #hometocanvass, #abortionreferendum). Some tweets have both pro- and anti-choice hashtags, and some do not have any hashtags; in such cases the coders read the overall argument of the tweet and coded whether it was written for or against abortion rights.

Tweets were coded to identify whether they contain incivility and intolerance. The unit of coding was at individual tweet level, and considered hyperlinks, images, and audio-video clips included in a tweet – and other tweets if a collected tweet was a part of a thread made by the same user.

The coding of the 3,000 tweets was undertaken by one coder, but two coders have partaken in the process of codebook development and coder training to test inter-coder reliability. They tested 5 per cent of the sample (n = 150) during several rounds of training, revision, and retesting. Reliability for all individual variables in the final version of the codebook achieves substantial agreement in terms of both Cohen's kappa (.75 average) and per cent agreement (93% average).

We have also calculated Krippendorff's alpha, which was equivalent to the Cohen's kappa reported here, achieving substantial level of agreement.

This study needed multiple “consensus binding discussions” (Neuendorf, 2002: 134) among coders to produce very specific rules to code incivility and intolerance in direct relation to abortion discourse. Discussions were necessary to achieve high inter-coder agreement, because although incivility and intolerance are largely commonsensical and socially shared concepts, they also have individual variances. Many would have a shared notion at an abstract level of what incivility and intolerance are (e.g., swearing in political discussions is incivil, denying others’ liberty and rights are intolerant), but there could be many disagreements at an empirical level (e.g., Is “bloody” an incivil word? Is it intolerance to call pro-choice women “baby murderers”?). Having consensus-binding discussions creates more practical rules that can apply to code more common, mundane forms of incivility and intolerance that is particular to the Irish abortion discourse.

Qualitative textual analysis was conducted to examine the ways incivility and intolerance are used by different pro- and anti-choice users and users whose political positions are neutral or ambiguous. Such qualitative in-depth analysis is crucial to developing a thorough understanding of the phenomenon, which cannot be grasped solely by statistical analysis. Qualitative textual analysis enables situated analysis of how incivility and intolerance are expressed and substantiates how the use of incivility and intolerance are different.

Our first research question asks what the mapping of incivility and intolerance is. It turns out that incivility and intolerance do not account for the majority of the Twitter discussions. Overall, only 14 per cent (n = 413) of tweets contain incivility, and 5 per cent (n = 138) include intolerance. In short, there is neither a substantial amount of incivility nor intolerance in Irish abortion policy discussions on Twitter.

These percentages are interestingly lower than the substantial amount of online incivility and intolerance other international research has found: for example, Coe and colleagues's (2014) study finds that in the comment section in an American local newspaper website, more than one in five comments (22%) contain incivility. Rossini's (2019) study on a Brazilian Facebook page and newspaper websites finds that the presence of incivility is substantial both on Facebook and the websites (41% in the news websites and 37% on Facebook) although the presence of intolerance is much less frequent (10% on Facebook and 4% on news websites). Their findings are not directly comparable to this article's context of the Irish abortion discourse on Twitter, but such disparities require further exploration in this matter: Could the low presence of incivility and intolerance in the Irish Twitter discussions be attributed to Ireland's online civic culture, the nature of abortion discussions, or our method? Another aspect to be taken into account is that there were tweets lost during the rehydration process. When we were rehydrating the dataset in March 2019, we were not able to retrieve 437,026 tweets. This might have been because some users themselves deleted their tweets or changed their accounts to private, or because Twitter deleted and suspended accounts that violated Twitter guidelines. Despite the loss of some tweets, it does not pose enough of a problem as to invalidate our findings, because the size of the rehydrated dataset is still massive and accounts for 80 per cent of the original dataset.

In terms of the relationship between gender and online uncivil discourse, our analysis shows that gender has statistically meaningful relationships with both incivility (χ² = 9.707, p < .01) and intolerance (χ² = 7.960, p < .05) (see Table 2). In terms of incivility, it is inferable that users with female names use incivility less frequently than users with male and gender-unidentifiable names. Users with female names have the lowest percentage of incivility within each gender pool: 12 per cent of female tweets are incivil, whereas 16 per cent of male tweets and 15 per cent of gender-unidentifiable tweets are incivil. It is interesting that even though the overall discussion is largely led in terms of numbers by users with female names (1,122 tweets by female names and 742 by male names), it is users with male names who use incivil language more on average.

Similar inferences can be made in the analysis of intolerance as well. Users with female names use intolerance less frequently than users with male names and especially less than users with gender-unidentifiable names. 3 per cent of tweets by users with female names contain intolerance, whereas 5 per cent of tweets by male-name users and 6 per cent of tweets by gender-unidentifiable users include some form of intolerance. However, the effect size of such association might not be significant at a practical level (incivility φ = .057; intolerance φ = .052) (Glen, 2016). Taken together, it might be suggested that gender is related to the tendency to be more readily incivil and intolerant, but not to such a large extent.

We also examine incivility and intolerance in relation to users’ political positions (see Table 3). These results indicate that the political positions of Twitter users have significant associations with both incivility (χ² = 37.269, p < .001, φ = .112) and intolerance (χ² = 313.463, p < .001, φ = .326). Anti-choice users use incivility more often, and use intolerance considerably more often, than pro-choice users and users whose positions are neutral or ambiguous. This is another interesting finding. Although the landscape of the Irish abortion discussion on Twitter is largely pro-choice dominant (2,237 pro-choice tweets and 496 anti-choice tweets), it is the anti-choice tweeters who are more frequently incivil and intolerant. Within the positions, 23 per cent of anti-choice tweets (n = 112) contain incivility and 20 per cent involve intolerance (n = 99), which means that one in every five anti-choice tweets contains some form of incivility and intolerance. This amount is substantially larger than the amount of incivility and intolerance in pro-choice tweets (13% [n = 279] and 2% [n = 39] respectively).

So, here, we can see a pattern. We have analysed above that users with male names hold anti-choice views more often than users with female names. This means that male names and anti-choice views overlap to some extent, and they – who are minority in number in the Irish abortion discussions – tend to express more incivil and intolerant tweets. This suggests that there is a relationship between the form and content of populist communication. It is plausible to suggest that these incivil and intolerant tweets express “the cultural backlash” from conservatives, who were presented with a discursive opportunity to express their feelings of resentment, anger, and a sense of loss and be susceptible to misogynistic, sexist, and other intolerant appeals (Norris & Inglehart, 2019).

Finally, we explore how the use of incivility and intolerance changes across the time of the abortion referendum debates from April to June 2018. Our chronological analysis shows that total participation and incivil and intolerant tweets drastically increased from 24 May, a day before the vote. The figures below (Figure 1 and 2) indicate that substantial amounts of tweets were made over a period of three days: 24–26 May, a day before the vote until a day after the vote. The day of the vote had the highest number in terms of participation, incivility, and intolerance. Of all referendum tweets, 44 per cent (n = 1,302) were made in these three days; so were 34 per cent of incivil tweets (n = 141) and 30 per cent of intolerant tweets (n = 41).

Figure 1

Figure 2

We can infer that incivility and intolerance are political and contextual products rather than intrinsic personality traits. Incivility and intolerance are heated expressions of political and ideological beliefs and populist sentiments, probably as a last spurt to promote a side to win the referendum on 24 and 25 May – and additionally, expressions of emotions – excitement or frustration and resentment – when users have discovered their side has won or lost in the referendum on 26 May. Even when debating such an emotive topic as abortion, it was only during the final day and in the immediate aftermath that discourse took an incivil and intolerant turn. However, it should be noted that the similarities between the 3,000-tweet sample and the 1.8+ million-tweet dataset have not been tested, and this analysis may not be representative of the entire dataset, nor of the complete abortion discourse at a national scale.

Our second research question asks whether incivility and intolerance are distinct concepts. The results suggest that they are, although they co-occur on occasion. Table 4 highlights that 36 per cent of intolerance entails incivility, whereas it is 12 per cent of the incivil tweets in the sample (n = 49). This indicates that intolerant tweets are more likely to be incivil and intolerant at the same time, but not vice versa (χ² = 55.580, p < .001, φ = .137).

Furthermore, interestingly, two-thirds of the intolerant tweets are not incivil (n = 89, 65% of intolerance) – which means that they have exclusionary and intolerant messages, but disguise themselves in a rational and civil tone. This substantiates the main argument of the article – what is problematic in uncivil society is the intolerant content – and less the incivil tone – of political communications.

This qualitative analysis is part of a still ongoing effort for Dayei Oh's doctoral project, and in this article, we only include a limited range of preliminary findings. Qualitative textual analysis shows that incivility as a tone has some ambiguous potential that can serve both for and against democracy. Incivility can be rhetorical and expressive (reasonable), or harassing and distressing (unreasonable). On the other hand, intolerance is antidemocratic and anti-pluralist.

One of the most popular deployments of reasonable incivility is to call out perceived social injustice in a legal system, society, and culture. Pro-choice users, especially, employ incivility for this purpose, criticising the “archaic”, “illiberal”, “barbaric”, and “cruel” anti-abortion laws in Ireland. Anti-choice users – although not as extensively as those who are pro-choice – also passionately and aggressively criticise contemporary society, which is heading downhill to a “shameful”, “vampire and clown” “cult of death” that does not recognise the sanctity of human life.

Reasonable incivility is also used from both pro- and anti-choice sides to condemn opposing institutions and opposing groups of people. Many pro- and anti-choice incivilities target major institutions, such as the government, political parties, the Catholic Church, mass media, and so forth. Even when users criticise groups of citizens, they tend to attack a group of people as a collective entity (e.g., “those campaigners are disgusting”), instead of harassing and insulting individual users in a direct way (e.g., “you are disgusting”). Furthermore, users often attack the ideas, arguments, and actions of their opponents that are relevant to the abortion discussion, instead of attacking their opponents’ personal attributes. This corroborates the finding of previous incivility research that the public distinguish personal- and public-level incivility (Muddiman, 2017), as well as personal attack and issue-based attack (Stryker et al., 2016).

Moreover, reasonable incivility is a way to assert one's determination and provide empowerment for marginalised actors, such as women. Even though abortion and crisis pregnancies are women's issues, Irish women's voices and lived experiences have been largely peripheralised and silenced in the andro-centric public spheres of religion and politics (Earner-Byrne & Urquhart, 2019). Breaking the long silence, Irish pro-choice women tweeted empowering statements like “Shove it up your ass, we will not back down!”, “Get your hands off my fucking uterus”, “Don’t fucking dare to tone-police us”, and so forth. Such incivil rhetoric is chosen by women to make their voices heard and recognised and to express their determination and resilience. Such expressive incivility is different from insulting incivility, which is used to distress and harass a specific individual with vulgar names and swear words.

This analysis of reasonable incivility is not an exhortation to being incivil and hostile to each other, but rather to acknowledge the rhetorical function of incivility in the sites of political struggles. Zerilli provides a feminist analysis of incivility in that the trope of civility is often deployed against feminist and anti-racist arguments. She adds:

Uncivil public behaviour is symptomatic of a more general democratic deficit of public space in which grievances can legitimately be raised and meaningfully addressed by fellow citizens and their elected representatives. If some citizens are more prone to shout, that may well be because those in power are not listening.

In the sample, there is problematic and unreasonable incivility that operates as menace, abuse, and distressing attacks at named individuals. For instance, both pro- and anti-choice tweets sometimes call opponents harsh names, such as “Fucking gobshite”, “Communist scumbag feminist goblins”, “Spineless cowards”, and so forth. Some tweets express threatening remarks at politicians, public figures, and users with opposing views, such as “Get in the sea”, “Crawl under a rock”, and so forth. Such hostile antagonism does not contribute to the deliberative and persuasive side of the public sphere and civil society and should be discouraged.

However, some incivility lies between menace and humour. A pro-choice tweet, for example, says “Get in the sea if you are so fond of forcing people to have to cross the sea to get abortion”. Is “get in the sea” in this tweet an online joke to signal disdain, or a threat and harassment to tell someone to drown themselves (equivalent to “Go kill yourself”)? There were some precedent cases where online menace and jokes are indiscernible. In 2016 in the UK, Labour Party member of parliament Thangam Debbonaire accused a University of Bristol student of sending her a death threat – “Get in the sea” – on Twitter. While the university agreed with Debbonaire and disciplined the student to write a formal letter of apology, others criticised Debbonaire for overreacting to a popular online joke (Tait, 2016). Overall, qualitative analysis of incivil tweets show that the distinction between reasonable and unreasonable incivility cannot be sharply made and needs further deliberation.

In contrast, intolerance is anti-pluralist and antidemocratic. Many intolerant tweets enter the abortion discussion in bad faith. The worldview of the intolerant users involves a simplistic moral dualism where “we” are righteous, good, and represent deliverance, and “they” are immoral, evil, and represent demise. In doing so, inter-group conversation or compromise seems undesirable or nearly impossible.

In their demonising rhetoric, intolerant anti-choice tweets compare pro-choice users to “the devil”, “Satan”, “Nazi”, and “Hitler”, and compare abortion to “killing”, “murder”, “slaughter”, and “holocaust”. Some intolerant pro-choice tweets also use the Nazi trope to demonise the anti-choice side (e.g., “his language is reminiscent of Nazi propaganda”). Nazi analogies can be inappropriate hyperbolic comparisons and argumentative fallacies if the association is based on superficial similarities (e.g., Nazi accepted X. Group Y accepts X. Therefore, Y is Nazi) (Curtis, 2004). Only a thorough comparison of history and political ideologies between Nazism and one's opponents can make the Nazi trope more than an unreasonable and fallacious intolerance.

Furthermore, a couple of pro-choice tweets use the trope of Islam and its fundamentalist terrorist organisations to demonise opposing political parties and politicians (e.g., “misogynistic like the Taliban” and “While shouting for abortion and LBGTQ+ rights, he promotes the Muslim Brotherhood and Hamas who slaughter LGBTQ+ and their babies to death”). Similar to the Nazi trope, such Islamic fundamentalist analogies can be antidemocratic in that they exacerbate antagonistic polarisation between “us” and “them” based on fallacious logic. For instance, an Irish politician supporting an Irish-Palestine solidarity campaign was denigrated as promoting Islamic fundamentalist groups and their intolerant anti-LGBTQ+ crimes. The association between supporting Palestine and promoting Islamic fundamentalism and their anti-LGBTQ+ crime lacks rigorous analysis of politics and ideologies. The superficial intolerant analogy simply derails deliberation by shunning any reasonable contribution from the person denounced as an “Islamic terrorist sympathiser”. It also reproduces the discriminatory representation of Islam as extremism and violence, which can further marginalise the Muslim communities living in Ireland and Europe.

Lastly, some of the intolerant anti-choice tweets are not just intolerant of abortion rights, but they express their rejection of other liberal changes in sociocultural values and non-conventional lifestyles, such as LGBTQ+ rights; for example, criticising a sex education bill for “promoting abortion”, “peddling the myth that there are more than two genders” and “explaining how two men have sex” to “our” children. These tweets demonstrate the deep entanglement between anti-abortion intolerance, anti-liberal backlash from social conservatives, and the rise of populism, which reaffirms that anti-abortion and anti-feminism are often part and parcel of the bigger populist phenomenon.