Digital Moral Outrage, Collective Guilt, And Collective Action: An Examination of How Twitter Users Expressed Their Anguish During India's Covid-19 Related Migrant Crisis

Abstract

This paper examines the intersection of group-based expressions including digital moral outrage, collective guilt, and collective action on Twitter, following the tragic incident of 8 May 2020, in which 16 migrant workers were run over by a train after the Indian government imposed a sudden COVID-19-related lockdown. Twitter data were gathered immediately at three different times - May 8–15, May 16- 23, May 24-May 31, and 4598 tweets were manually coded. The analysis revealed that digital moral outrage was the most frequently expressed emotion. It, however, gradually decreased, signaling digital outrage fatigue. Collective guilt and sympathy constituted the second-largest portion of the total tweets, and tweets reflecting collective action by the community progressively increased. The network of relationships among different group-based emotions, the promotion of one-sided narratives and virtue signaling on social media platforms are discussed.

Poverty, humanitarian crises and gory narratives have the potential to evoke strong emotions among people (Krauth-Gruber & Bonnot, 2020). Every time an image of a distraught citizen is splashed across mass media, the angry public urges a strong corrective action from the responsible authorities. The strong emotional reaction that engulfs the masses when a tragic humanitarian crisis erupts in a community is described as moral outrage- an anger that emanates from the violation of a moral standard (Batson et al., 2007; Crockett, 2017). Moral outrage is often linked with collective guilt- a perception that ingroup members, or people in the unaffected group, are responsible for the plight of the outgroup or the suffering group (Krauth-Gruber & Bonnot, 2020; Rothschild & Keefer, 2017). Digital platforms such as micro-blogging site Twitter have made the expression of moral outrage easy, pervasive and with fewer repercussions than the offline world (Crockett, 2017). Crockett (2017), however, views the digital moral outrage as less effective in fulfilling long-term goals as outrage fatigue soon grips the users and people move from one problem to another very easily on these platforms, without making much effort to take any actions on the ground. Spring, Cameron and Cikara (2018), meanwhile, assert that digital moral outrage is a critical force for collective action as social media sites can act as knowledge-sharing platforms that, in turn, can increase public participation.

The humanitarian crisis during the COVID-19-related lockdown imposed by the Indian government in March 2020 led to a mass exodus of poor domestic migrant workers, estimated to be 100–130 million (Infante, 2020), from India's industrialized metropolitans to the rural hinterland (Dutt, 2020). After the government suspended public transportation, many workers decided to walk several hundred miles to reach their homes. While migrants’ hardships were ignored initially by the government and middle class, and workers were also viewed as virus-spreaders, their problems came to the forefront on May 8, 2020, when a freight train ran over 16 migrants in Aurangabad, Maharashtra as they slept on rail-tracks after completing their day's journey on foot. The images of mutilated bodies of men, women and children triggered a tsunami of emotional reactions across the country. To express their anguish at the plight of the poorer sections of the society, the 17 million Twitter users in India took to the social media platform to urge the government to take corrective action.

Limited research exists on the nature and frequency of group-based emotions expressed during a real-life situation on digital platforms. Furthermore, despite a humanitarian crisis of colossal proportions, limited research exists on India's migrant crisis and its social media use, especially from the perspective of group-based emotions and actions. Integrating the theoretical concepts of digital moral outrage, collective guilt and sympathy, and collective action, this study reveals the dominant group-based emotions prevalent on Twitter after the triggering incident and examines the changes in the frequency of these emotions over a three-week period.

Literature Review

Moral Outrage

Moral outrage, a strong group-based emotion, is an anger that emanates from the violation of a moral standard (Crockett, 2017). Moral outrage occurs when people put the responsibility on the system; it is an absence of self-blame, but the anger is directed at an institution or an authority (Baston et al., 2007; Crockett, 2017). Moral outrage, however, is different from personal anger and empathetic anger. Personal anger stems from the realization that one's own interests are harmed ( Krauth-Gruber and Bonnot, 2020). Moral outrage, on the other hand, an emotional response when one's aim is to restore fairness and justice, generally toward an out-group member. It is different from empathy as empathic anger is for the interests of a person for whom one cares about, and that person's interests have been thwarted (Batson et al., 2007). Perceived injustice can trigger moral outrage- both in the ingroup and the outgroup members, and these emotions are likely to shape social change (Rothmund, Baumert, & Zinkernagel, 2014; Thomas, McGarty, & Mavor, 2009; Saab et al., 2015). Moral outrage is also described as an action-oriented emotion as it often, but not always, precedes collective action against the perceived injustice (Saab et al., 2015; Thomas et al., 2009). Moral outrage, thus, is an anger that is expressed for an unrelated group where the underlying principle is that justice has not been delivered to a particular group.

Digital Moral Outrage

Digital media has transformed and amplified the expression of moral outrage (Crockett, 2017). When outrage moves online, “it becomes more readily available, requires less effort, and is reinforced on a schedule that maximizes the likelihood of future outrage expression in ways that might divorce the feeling of outrage from its behavioral expression” (Crockett, 2017, p. 769). In an offline world, Crockett (2017) argues, the expression of moral outrage comes at a cost- people usually express moral outrage via gossip, shaming and punishment and there is a risk of retaliation; but it is a lot easier to shame others online to sound virtuous and morally correct. In the digital world, because the frequency of triggering stimuli is extremely high, repeated expression of moral outrage can either lead to ‘outrage fatigue’: constant exposure to outrageous news could diminish the overall intensity of outrage experiences, or cause people to experience outrage more selectively to reduce emotional and attentional demands (Crockett, 2017). Spring, Cameron and Cikara (2019) are of the view that digital moral outrage can lead to sharing of knowledge on social media platforms that, in turn, can lead to organized collective action and increase participation in the public sphere. Spring et al. (2019) contend that moral outrage has mobilizing effects in terms of intergroup relations. Brady and Crockett (2019), while acknowledging the positive outcomes of digital moral outrage, argue that online moral outrage can have more downsides than upsides as online outrage could divert collective action toward issues that are immediately compelling but ultimately ineffective and counterproductive (Brady & Crockett, 2019). Furthermore, experts assert that online moral outrage can silence the marginalized voices as so much noise could be created by a dominant or a majority group (Brady & Crockett, 2019).

Recent studies highlight that the online moral outrage can give rise to negative speech and extremism. Workneh (2021), for example, found that while social media platforms helped youth participation in political discourse during the Ethiopian political unrest, the platforms also led to a rise of outrage communication, hate speech and political extremism. Similarly, Strathern et al., (2020) found that during crisis (firestorms), users often indulge in negative talk. Political scientists have found that politicians across party-lines leverage the impact of moral outrage (Brady et al., 2019), as evidenced in the former US President Donald Trump's expression of moral outrage at the violation of non-negotiable boundaries (Marietta et al., 2017).

Moral outrage, though stemming from anger, is different from digital vigilantism – a process where citizens are collectively offended by other citizens’ activity, and coordinate retaliation on mobile devices and social media platforms (Trottier, 2017). Digital vigilantism entails naming and shaming and online harassment, but digital moral outrage isn’t malicious and does not have criminological motivations (Trottier, 2017). It, however, remains to be investigated in a real-life situation, such as the 2020 domestic migrant crisis in India, whether digital moral outrage is indeed a frequently occurring group-based emotion on the digital platforms, and, if yes, against whom this anger is directed at.

Collective Guilt, Virtue Signaling and Sympathy on Digital Platforms

Guilt often predicts outrage (Rothschild & Keefer, 2017). Moral outrage is not merely a concern with justice alone; it is also an attempt by individuals to appear virtuous to others, and to reduce their guilt (Rothschild & Keefer, 2017; Puryear, 2020). For collective guilt to be felt, individuals must categorize as members of the group that has harmed or is harming another group and must perceive the ingroup as responsible for the harm done to another group. Collective guilt can be for actions and inactions (Krauth-Gruber & Bonnot, 2020).

Online participation behavior is driven by moral compass and a desire for social recognition (Johnen, Jungblut, & Ziegele, 2018). Puryear (2020) argues that digital networks provide near unlimited access to morally relevant discussions that reward outrage expression. Social media provides concrete indicators of social approval in the form of “likes” and “upvotes” that give concrete signals for moral reputation unlike anything in face-to-face reactions (Puryear, 2020). While the risk of condemnation is high in a face-to-face setting, digital platforms provide safe networks of like-minded people to condemn relatively risk-free and enhance their social reputation (Crockett, 2017). This does not mean that online outrage is not genuinely felt (Jordan & Rand, 2020). The digital outrage culture, however, is motivated by virtue signaling. Both moral outrage and collective guilt, especially guilt aroused by action of the ingroup, predict people's readiness to act in favor of the poor and disadvantaged group (Krauth-Gruber & Bonnot, 2020).

Sympathy (feeling of compensation for the plight of others) can motivate action (Harth, Kesseler & Leach, 2008). Sympathy is a response to a disadvantaged outgroup's suffering which involves feeling compassion for them (Thomas et al., 2009). Sympathy focuses on the disadvantaged group's plight rather than the advantaged group's actions (Harth et al., 2008). Studies have found that sympathy is a positive predictor for collective actions among advantaged groups (Saab et al., 2015). When people in the advantaged group feel sympathetic toward the suffering group, they are more likely to help, especially in the form of donations or other monetary help (Harth et al., 2008). Thus, digital platforms, by increasing the frequency of moral outrage and collective guilt and sympathy, should contribute to mobilizing collective action by the advantaged group to address injustice. Limited research exists on the frequency and the nature of collective guilt and sympathy on digital platforms. Hence, there is a need to investigate how the emotions of collective guilt and sympathy are expressed on the digital platforms, and how these intersect with digital moral outrage.

Collective Action or Collective Behavior

Research linking group-based emotions and people's intention to take collective action to mitigate injustice have often found that moral outrage and collective guilt influence people's collective behavior (Krauth-Gruber & Bonnot, 2020). In the previous studies, scholars argued that people act when they feel a relevant group-based emotion (emotion-based route) and/or believe that they can achieve change (efficacy-based route) (Van Zomeren et al., 2004). Theoretical frameworks such as relative deprivation theory (Runciman, 1966), and social identity theory (Tajfel & Turner, 2004) have clarified how a disadvantaged group mobilizes resources to take collective action against a perceived injustice.

Studies have documented the use of social media platforms for collective action and social change (Foster, 2015; Zhang & Pan, 2019). However, as the emotions of moral outrage and guilt move online on digital platforms, it is unclear if these group-based emotions result in a similarly meaningful collective action as in the offline world (Brady & Crockett, 2019; Crockett, 2017). Also, previous studies demonstrate that people may get deeply angry about an injustice, yet not engage in any kind of action. According to this perspective, people will engage in collective action only if they have the resources to effectively mobilize their resources for a meaningful action (Linewebber, Barr-Bryan, & Zelditch, 2015; Martin, Brickman & Murray, 1984). Therefore, collective action can be a desirable outcome of moral outrage, but it is certainly not a necessary outcome. Digital users may just express outrage to mitigate their guilt but do nothing to help the needy if they don’t think they have the resources to act meaningful. Thus, the role of social media as a resource mobilization tool for collective action, and collective action's relationship with moral outrage and collective guilt and sympathy must be explored when social media become the main tools for emotional expression.

Pro-Establishment/ Status-Quo Expressions

Prior research suggests that people can advocate pro-establishment views if they belong to an advantaged group and are not morally outraged by the event. According to the system justification model of collective action, the group-based anger and perceived group efficacy should motivate distinct forms of protest for those who are advantaged versus disadvantaged by the status quo (Jost et al., 2017). That is, people should have a different response to changing the status quo if they are high in system-based anger, group-based anger, and perceived group efficacy should be more likely to engage in system-challenging protest, whereas those who are advantaged and high in group-based anger and perceived group efficacy (but low in system-based anger) should be more likely to engage in system-supporting protest. Thus, group identification and intensity of group-based emotion both play a role in deciding whether an individual will participate in changing the established system. Therefore, not everyone will be outraged or express their anger on social media platforms; it is plausible that there will be voices in the advantaged group that will support the establishment. It remains to be seen, however, how the counter-narrative pans out on social media platforms especially in relationship with digital moral outrage.

Migrant Crisis due to the COVID-19 Lockdown in India

On March 23, 2020, in a televised speech at 8 p.m., India's Prime Minister Narendra Modi announced an immediate nationwide lockdown (beginning on March 24, 2020) and issued stay-in-place orders to contain the spread of Covid-19 (Balakrishna, 2020). The state governments sealed the borders the next day and suspended interstate travel (Balakrishna, 2020). The order was initially for only three-weeks but was later extended by several weeks as the infection rates increased. PM's sudden announcement triggered an exodus of seasonal, migrant workers from metropolitans to rural areas in eastern states of Uttar Pradesh and Bihar, many of whom were unsure of when the lockdown will be lifted and began their journeys on foot to reunite with their families (Balakrishna, 2020, Chakravarty, 2020; Dutt, 2020). Terming the domestic migrant crisis the worst that India faced since 1947 (Partition of India), Infante (2020) estimated that more than 100 million workers were displaced during the lockdown. Barkha Dutt (2020), the award-winning journalist who traveled with migrant workers during the lockdown, squarely blamed the government for causing a humanitarian crisis during a pandemic. She wrote: “The mass exodus of workers from cities was because they had been forgotten by policymakers, politicians and the media. Orphaned by the system and left without wages or work, laborer after laborer, in state after state, told me that if confronted with death, they would rather die at home, with the people they loved.”

Government's mismanagement of the crisis came to the forefront when workers began to die due to starvation and exhaustion. One such incident occurred on 8 May 2020, when a freight train ran over 16 sleeping migrant men, women, and children, as they slept on the tracks in Aurangabad, Maharashtra after a day's journey on foot (Balakrishna, 2020). The gory images of mutilated bodies were splashed across both the social and the mainstream media. While this incident highlighted the perils of such arduous journeys, this also brought into focus the government's lack of transportation arrangements for the migrants. This incident was followed by another road accident on 16 May 2020 in which 26 workers were killed as their private vehicle collided with another vehicle in Auriya, Uttar Pradesh. Chakravarty (2020) states that over 80 workers died partaking in a journey during the lockdown. It was only later in the month of May, several weeks after the start of the lockdown, that the government announced free of cost trains from the workers (called Shramik trains- worker trains) (Balakrishna, 2020).

India's apex court's role also came under a scanner after the Supreme Court of India rejected a Public Interest Litigation (PIL) on May 16 that requested the court to direct the district magistrates to arrange for the workers’ travel and stay. Court specifically mentioned that the May 8 train incident could not be prevented (Ananthakrishan, 2020). After sharp criticism, the SC, on May 26, directed the central and state governments to provide free food and accommodation to the workers (Ananthakrishan, 2020).

Twitter Users in India

India, with 17 million Twitter users, is the micro-blogging site's third largest user base in the world (www.statista.com). Twitter is considered an urban, elite medium in India that is extremely popular among influencers and opinion leaders (CSDS-Lokniti, 2019). Important politicians, journalists, industrialists, and famous actors are very active on the platform. PM Modi, with over 50 million Twitter followers, connects with his followers directly on social media platforms, often bypassing the mainstream news media (Pal & Panda, 2019). Twitter users were particularly active during the migrant crisis as social media platforms were used to criticize the government over its inaction (Elsa, 2020). The hashtag #StepdownModi was the most trending hashtag on Twitter in India in June 2020 (Elsa, 2020). Many people from the middle class, including professional journalists, used the platform to post pictures and videos of the suffering workers on Twitter to highlight migrants’ plight (First Post, 2020). The demographic profile of Twitter users in India- millennials, educated, urban, elites- is different from that of the poor, domestic workers who were the most affected during the lockdown. Thus, Twitter users in India should be classified as an ingroup (advantaged, majority, non-suffering group), while the domestic migrant workers should be classified as an outgroup (disadvantaged group that faces injustice) in this study.

Aims of the Study

Using the 8 May 2020 train accident as a stimulus for triggering emotional responses, we analyze how Twitter users reacted to the COVID-19 related migrant crisis in India. We specifically aim to find out how the expressions of moral anger, collective guilt and sympathy, collective action and pro-establishment expressions played out immediately after the incident, and which of these reactions were the most frequently expressed emotions and actions on Twitter. Crockett (2017) asserts that digital platforms can increase the volume of moral outrage because it is easy and may sound morally righteous to condemn online as compared to offline. Thus, we pose our first question:

• RQ1: Which group-based expression- moral outrage, collective guilt and sympathy, collective behavior, or pro-establishment tweets - was most frequently expressed by tweeters after the tragic train incident?

Additionally, limited research exists on how the nature and frequency of these emotions and collective action change within a few weeks of the triggering incident. Crockett (2017) suggests that since digital media offers several triggers to the users, repeated exposure could lead to outrage fatigue. We want to investigate how, within the selected three-week period, did the expression of emotions change, if at all, on Twitter. Did systematic anger persist, or did the platform emerge as a tool for collective behavior? Thus, we pose our second question:

• RQ2: How did the group-based emotions and collective action on Twitter behave in the weeks after the triggering incident?

Prior studies have demonstrated a significant relationship between moral outrage, collective guilt, and collective action (Saab et al., 2015). On the digital platforms, though such relationships are not studied extensively. We, thus, propose our third question:

• RQ3: What is the relationship between group-based emotions including digital moral outrage, collective guilt and sympathy, and collective action?

Methodology

To examine the frequency and flow of group-based emotions and collection action, we extracted tweets on three different time periods from May 8-May 31, 2020. Using the MAXQDA Analytics Pro software, we searched tweets using the keywords such as Migrant Labourers India, Shramik, mazdoor, migrants on the road, migrant crisis, trains, and migrants. MAXQDA Analytics Pro software allows extraction of tweets for up to one week prior to a given date, and up to 10,000 tweets can be extracted per word search or hashtag search. We started extracting tweets on May 15, one week after the Aurangabad freight train incident. Using the above-mentioned keywords, we extracted tweets from May 8–15. On May 15, we extracted 42,436 for Week 1. After waiting for one week, we extracted the next set of tweets, 36,845 in total on May 23, and the final set of 13,882 tweets on May 31 to give us tweet data sets for three different time-periods immediately after the train incident. MAXQDA only has the option of extracting text tweets and does not extract pictures or videos Therefore, we only analyzed the text in the tweets and did not include the graphics or images/videos in our analysis. Following this method, we had a total of 93,163 tweets from May 8 to May 31. Duplicate tweets were discarded. For the ease of analysis, we only included tweets that were either in English or Hindi language (the two coders were only familiar with these two languages). We used stratified random sampling to select 4598 tweets for manual coding and analysis. We pooled a representative sample from the three time-periods that we selected in this study. For week 1, we randomly selected 2288 tweets (5.4% of the tweet population), for week 2- 1398 tweets (3.8% of the tweet population) and for week 3- 912 (6.5% of the tweet population). This made sure that we have tweets from the three different weeks in our sample.

Coding Scheme

After randomly selecting 4598 tweets, we manually coded the tweets using our coding scheme that contained 13 mutually exclusive categories under four main variables: a) moral outrage, b) collective guilt and sympathy, c). collection action, and d). Pro-establishment/Status quo.

The unit of analysis was one tweet. For ease of analysis, we assigned only one main code to each tweet, and discarded any sub-themes in the same tweet. Thus, each tweet could fall only in one of the 13 mutually exclusive categories. Two coders coded the tweets. For intercoder reliability, one coder coded the entire sample and another coder coded 10% of the tweets (Intercoder reliability, Krippendorf alpha range 0.79- 0.90).

Moral Outrage (MO): Expression of moral outrage entails putting responsibility on the system (Batson et al., 2007; Krauth-Gruber & Bonnot, 2020), especially on the government and the authorities responsible for providing justice and an equitable environment to the citizens. In this study, we coded moral outrage in the following ways:

1. Anger against the central government including PM Modi, senior ministers (e.g., Railway minister Piyush Goyal and the ruling BJP party): Those tweets that clearly labeled the central government, PM Narendra Modi or any of the key ministers responsible for the crisis were coded in this category. For example, such tweets: “This BJP government is useless or else these migrants won't be on the roads.”

2. Anger against the structural economic divide (Rich Vs. Poor): The stark economic disparity in the country was highlighted by several tweeters. Tweeters expressed anger on how the government was treating its Non-resident Indians (NRIs) by arranging special flights but completely ignored its poor. Tweeters used the hashtags #RichVsPoor, #BharatVsIndia etc. Tweets included: “A fancy title & strategy to bring Indians stranded abroad thru #VandeBharathMission. Great work but, why not the same for the poor migrants?”

3. Anger against the state governments: In these tweets, the responsibility for the migrant crisis was fixed on the state governments and not on the central government. Most of the non-BJP government states were targeted in these tweets. E.g., “State govts should provide food to the migrants who are on foot walking to their villages.

4. Anger against the judiciary: The Supreme Court in India came under a sharp criticism after it refused to direct the central government to take steps to mitigate the crisis. Tweets included: “Governments are not for the poor, but the Supreme court did not look after the poor either.” “You might get some progressive judgements on gender and sexuality issues. But not on these. Shame!” and “#SupremeCourt, how low can you stoop?”

5. Anger against the systematic failure to reform labor laws: These tweets questioned the rights of migrants and implored the community to think about labor laws and working conditions of the workers. “Think about people who do our dirty work. They also have a right to live with dignity” and “Need for reforms in labor laws. These people need a voice.”

6. Anger against the previous governments (e.g., Left-liberal Indian National Congress party): Tweets in which the migrant crisis was presented as a failure of the government prior to the BJP-led government presently running the country.

Collective Guilt and Sympathy (CG & S): Advantaged Group's Emotions Toward Workers:

The data for collective guilt and sympathy of the ingroup for the migrants were coded in the following categories:

1. Community shame: We coded those tweets that highlighted collective shame, guilt, or failure of the community to help the workers. No responsibility was fixed on any authority, but a communal sense of failure was expressed. E.g “Not here to point finger at government may it be State govt. or Central govt. But I feel the rest of us should never be able to forgive ourselves for the pain we have allowed our brothers & sisters,” and “What makes us proud of being Indians? We've kept trying to look away from poverty. But now, there's a lockdown, and the poorest are on the streets. Where are you going to turn your gaze?”

2. Plight of migrants: These emotional tweets described the sufferings of the migrants using their personal narratives, stories of endurance, etc. For example: “So, this's a photo I took from my house. It is so hot here in #Mumbai but still these #migrants are desperate to leave the city. They are walking on the railway tracks. Hard to see their plight,” and “Saw this disturbing scene in the park across from my house: 2 migrants passed away with 2 infants playing nearby,”

Collective action (CA): We coded CA in the following two categories:

1. Pleas to help migrants: In this category, we coded tweets that urged the community to take actions to help the migrants in need. These also included tweets that shared how some people helped the returning migrant workers. “Please donate shoes to migrants who are walking miles to reach their home,” “At this point of time if you find migrant laborers, provide them with anything that you have. They need us now,”

2. Appreciation of those who helped the stranded migrant workers: Tweets included gratitude for those who helped the migrants, including public figures. E.g., “#SonuSood needs to be awarded the highest civilian award.” “Only if others did half of what #SonuSood is doing, the migrant's lives would have been so much better and some wudnt have lost their lives. Some people walk but some people only talk. #MigrantLivesMatter.”

3. Urging others in the advantaged group to help the workers: Tweets urging industrialists and the rich to come forward and extend help.

Pro-Establishment/Status-quo (PRO):

In our sample, we also found tweets that blamed the migrant workers for spreading the disease (by traveling to different places despite the lockdown) or by highlighting the positive steps taken by the government. Jost et al. (2017) have argued that the advantaged group can advocate for the established system. Additionally, not everyone on Twitter can be outraged by migrants’ crisis and may hold opposite views. We coded under the following categories:

1. Support for government policies: any positive role played by the institutions. For example: “Jharkhand is the only state which brought #MigrantWorkers back home in chartered flights.”

2. Blame on migrant laborers: Tweets in which workers themselves were blamed for the crisis (their inability to follow the lockdown) and, thus, their irresponsibility of spreading the coronavirus. For example: “These villagers will never understand the concept of social distance, we should use force on them”, and “Sleeping on the tracks. That's the most absurd thing. We can’t justify sleeping on the tracks.”

Results

In this study, moral outrage was the most common emotion expressed during the selected period (Table 1). Out of the 4598 tweets, 44.54% (n  =  2048) were coded under one of the six expressions of moral outrage. Collective guilt and sympathy, 32.21% (n  =  1481), was the second most common expression, collection action accounted for 20.47% (n  =  941) and pro-establishment tweets were the least in numbers, with only 2.78% (n = 128) tweets coded under this category.

Table 1.

Distribution of Group-Based Emotions and Actions from Week 1 to Week 3.

Moral Outrage	Week 1	Week 2	Week 3	Total/variable
Anger against central govt/Modi	538	351	161	1050 (51.27%)
Anger against economic divide	213	85	21	319 (15.58%)
Anger against state govts	112	52	37	201(9.81%)
Anger against judiciary	73	33	20	126 (6.15%)
Anger against lack of labor reforms	232	16	83	331(16.16%)
Anger against previous govts	9	8	3	20 (0.98%)
Total	1177 (51.44%)	545 (38.98%)	326 (35.75%)	2048 (44.54%)
Collective Guilt & Sympathy	Week 1	Week 2	Week 3
Sympathy for migrants	457	263	90	810 (54.70%)
Community guilt/shame	391	157	123	671(45.30%)
Total	848 (37.06%)	420 (30.04%)	213 (23.36%)	1481 (32.21%)
Collective action	Week 1	Week 2	Week 3
Pleas to help migrants	185	146	125	456 (48.46%)
Appreciation for those who helped	14	180	159	353 (37.51%)
Urge elites to Speak up	2	60	80	142 (15.09%)
Total	201 (8.78%)	386 (27.61%)	354 (38.82%)	941 (20.47%)
Pro-establishment tweets
Blame migrants	10	7	3	20 (15.63%)
Pro-govt statements	52	40	16	108 (84.37%)
Total	62 (2.71%)	47 (3.36%)	19 (2.08%)	128 (2.78%)
Total/week	2288 (100%)	1398 (100%)	912 (100%)	4598

Within moral outrage, outrage against the central government, its ministers including PM Modi, was the highest (n  =  1050, 51.27%). Tweets mentioned the PM's name directly and put blame on him and his ministers for mismanaging the crisis. This was followed by anger against lack of reforms in labor laws (n  =  331, 16.61%), anger against structural economic divide (rich Vs Poor) (n = 319, 15.58%), anger against the state governments (n  =  201, 9.81%), anger against the judiciary (n  =  126, 6.15%), and anger against the previous government (n  =  20, 0.98%). In the collective guilt and sympathy category, sympathetic tweets highlighting the plight of migrants were the most prevalent (n  =  810, 54.70%). These were followed by tweets expressing community shame and guilt, n  =  671 (45.30%).

In the collective action category, tweets seeking help for migrants dominated this category (n  =  456, 48.46%), followed by tweets showing gratitude to those who had helped, including public figures (n  =  353, 37.51%). Among these were tweets about Bollywood actor Sonu Sood, who was among the first to arrange private transportation for the migrant laborers, Amitabh Bachchan, and celebrity chef Vikas Khanna were the most prominent. These were followed by tweets urging other celebrities, industrialists, etc. to come forward and help the suffering people (n  =  142, 15.09%).

In the pro-establishment tweets, 20 tweets (15.63%) blamed the migrants for travelling and spreading the virus, and 108 (84.37%) tweets mentioned the positive role of the authorities for managing the crisis.

In RQ1, we proposed to find the most frequently occurring group-based emotions and their break-up. Moral outrage, thus, was the most common emotion and the anger against the ruling government was most common. In RQ2, we wanted to learn how these group-based emotions behave (change or remain consistent) over the three-week period. In our analysis, we find that even though moral outrage was the most frequently expressed emotion during this study's time frame, the movement of outrage is insightful. As the time progressed since the tragic train incident, the proportional number of tweets expressing moral outrage decreased gradually, while the collective action tweets increased proportionally from week 1 to week 3. In week 1, moral outrage constituted 51.44% (n  =  1177) of the total tweets. It then went to 38.98% (n  =  545) and in week 3 it reached 35.75% (n  =  326). Though this drop is not very huge, and moral outrage remained the most frequent emotional expression throughout, we observe that collective action jumped from week 1 to week 3. In week 1, collective action comprised only 8.78% (n  =  201) of the total tweets, it reached 38.82% (n  =  354) in week 3 in our sample. The jump from week 1 to week 2 was substantial too; it was 27.61% (n  =  386) of the total tweets in week 2. The movement in collective guilt and sympathy also changed, though not as substantially as collective action. The total number of collective guilt tweets in Week 1 were 848 (37.06%) reduced a little to 30.04% (n  =  420) in Week 2 and went down to 23.36% (n  =  213) in Week 3. Pro-establishment tweets and tweets blaming the migrant laborers for not following the stringent lockdown rules remained largely consistent over the three-week period. Such tweets were 2.71% (n  =  62) in week 1, slightly rose to 3.36% (n  =  47) in week 2 and went down to 2.08% (n  =  19) in week 3, indicating a range-bound movement of such tweets on the social media platform.

In RQ3, we proposed to find an association between the group-based emotions. The codes’ proximity analysis revealed (Figure 1) that the main codes pertaining to moral outrage and community guilt and shame occurred were closely related. In Figure 1, “Negative Central Government and Modi (Anger against the Central Government) appears at a close distance with “Plight of migrants” and “Community guilt and shame” and are strongly connected. These are the main nodes from which the other codes such as “Rich Vs. Poor (anger against economic divide) ``anger against the role of judiciary”, and “Pleas to help,” emerge, that are further connected to “Positive celebrity messages” and “Speakup elites,” etc.

Figure 1.

The network of the proximity of codes and their relationship.

Thus, it is demonstrated in this study that one aspect of moral outrage (here, anger against the Central government) is strongly connected to other emotions such as guilt and shame, and these group-based emotions are further linked to other types of moral outrages (such as economic divide, judiciary's role, etc.) and can then connect to collective action in the form of resource mobilization, including appreciation, to help.

Discussion

By focusing on India's 2020 migrant crisis related to the COVID-19 lockdown, this study extends our understanding of the expression of group-based emotions on digital platforms. By analyzing tweets at different points after the stimulus incident, we provide insights into how the expressions of moral outrage, collective guilt and collective action flow and intersected on social media platforms as India grappled with an unprecedented domestic migrant crisis during the pandemic. The expression of moral outrage was the most common emotional reaction on Twitter during India's domestic migrant crisis. Digital platforms can increase the outrage volume disproportionately (Crockett, 2017) and people, in general, find putting blame on the authorities easier than feeling guilty for the ingroup's actions or inactions (Krauter-Gruber & Bonnot, 2020). Furthermore, Elsa (2020) stated that anti-Modi tweets were the most trending hashtag in India in June 2020 as the migrant workers toiled on roads. This study corroborates these findings as tweeters in our sample expressed moral outrage the most, and within moral outrage, the blame was mostly put on the Modi-led government and the PM. As Crockett (2017) has argued, online platforms give the convenience of expressing moral outrage without having to deal with any offline confrontation. Apparently, it was convenient for tweeters to tag Modi, who is very active on social media platforms, and his political party on Twitter to express their outrage. Twitter provided a platform for people to express their anguish and disappointment toward the judiciary, previous governments, and other structural inequalities in India, e.g., the rich and poor divide and the lack of labor reforms. Tweeters, thus, directed their anger at not just one authority; their outrage was directed at the present government, and the prevailing structural differences that have existed for the past several decades. This study thus highlights that the expression of moral outrage can take several forms and shapes and individuals bring in their own understanding of history and culture to direct their anger at an authority they deem responsible for the crisis.

The Covid-19 related lockdown also presents an unusual scenario in a democracy during which people could not gather in-person, even if they wanted to, as the pandemic gave the government power to restrict people's physical movement. Social media, thus, became the only viable tool for people to participate in the public sphere and emerged as a sustainable medium for people to express their anger and mobilize resources when their right to congregate was suspended. Future studies must investigate the role of social media in collective action, especially when the offline offers little opportunities to congregate on the ground.

Spring et al. (2019) assert that online moral outrage has the potential to provide a forum for exchange of information that can potentially result in some meaningful collective action. Our analysis demonstrates that Twitter became a platform for exchanging information as well, in addition to emerging as a medium to channelize people's anger. In this study, we analyzed the use of social media in arranging help, showing gratitude, and amplifying the Samaritans efforts, including celebrities like Sonu Sood, of reaching out to the workers in need. Thus, we agree with Spring et al. (2019) that social media platforms can become media of knowledge exchange even though digital platforms can increase the outrage volume disproportionately. Future studies must examine digital moral outrage's influence on collective action and investigate how resources are mobilized online when the platforms are primarily used to express anger. This analysis is especially meaningful for countries like India where access to the Internet is still a privilege (only 50% of the country has internet access) and Twitter is mostly used by the urbanized elites. In such a situation, the way the privileged group used the platform to highlight the plight of the poor laborers amplifies the role social media outrage, and guilt and sympathy can play in mainstreaming the issue.

We also acknowledge Crockett; (2017) digital outrage fatigue as moral outrage gradually decreased in our three-week sample and tweeters started tweeting more messages of hope and gratitude. This slightly shifted the narrative on social media, even as the migrant crisis continued. We find merit in Crockett's argument and urge future studies to investigate the onset of fatigue on these platforms and what are the people's motivations for persisting with expressing outrage. In this age of hashtag activism and social media outrage, some online movements survive more than the others. The cycle of digital moral outrage, thus, needs to be investigated as to when outrage begins to ebb, and which other group-based emotions gradually begin to replace it.

Previous studies find that digital platforms become stages for virtue signaling and people condemn behavior online to reduce their own guilt (Jordan & Rand, 2020; Puryear, 2020). We found ample evidence of people's expressions of their guilt and a feeling of shame in our study. Interestingly, we find that public figures (actors, celebrity chefs, wealthy businessmen, etc.) got more active after Week 1 and started offering help for the stranded migrants. Actor Sonu Sood was among the first celebrities to arrange private transportation for the workers. People directly connected with him on Twitter, and Sood posted about each person whom he had helped. While the use of Twitter by Sood during this crisis in an example of resource mobilization via digital platforms, the discussion around moral compass and virtue signaling is relevant here as Twitter also became an effective platform for virtue signaling for the public figures. It was an opportune moment for the advantaged group, especially the celebrities, to show their solidarity with the suffering group. After Sood came forward to help, many other celebrities including chef Vikas Khanna, actor Amitabh Bachchan, etc. joined in the cause and the story of migrant crisis also became stories of celebrities and their compassion. We are not dismissing the efforts put in by these people; many might have acted out of genuine compassion. However, the public figures’ use of social media during humanitarian crises should be studied using the concepts of virtue signaling, moral outrage, guilt, and sympathy.

Furthermore, we find very few pro-establishment or anti-migrant tweets in our sample. Virtue signaling can explain the overwhelming anti-establishment and pro-migrant sentiment on Twitter. While the digital platforms can tremendously increase the volume of outrage, it is plausible that people with pro-establishment or anti-migrant views did not express themselves as frequently to avoid the risk of appearing non-virtuous or receiving a backlash on Twitter. It is plausible that the volume of digital moral outrage and guilt painted a one-sided narrative on Twitter, and it deterred the opposing group from expressing their views. Brady and Crockett (2017) observed that the ease of expression of moral outrage on digital platforms presented the risk of amplification of dominant voices where the non-dominant voices can be easily silenced. Thus, to sound virtuous, digital media users can join the dominant narrative, amplify the majority voice on social media but this may not result in any real ground action. The link between virtue-signaling, moral outrage and dominant narrative on digital platforms need further exploration.

Limitations

We developed a coding scheme to manually code the complex constructs of digital moral outrage, collective guilt and sympathy, collective action, and pro-establishment expressions. Additionally, we only assigned one code to each tweet. It is possible that one tweet can contain multiple themes. This study, therefore, may not have captured the complexity of the tweet data and relies on a manual coding scheme that is susceptible to human judgment. Further studies can refine this coding scheme.