Comparing the #StopTheSteal Movement across Multiple Platforms: Differentiating Discourse on Facebook,Twitter,and Parler

Abstract

Given that political groups are dispersed across platforms, resulting in different discourses, there is a need for more studies comparing communication across platforms. In this study, we compared posts about #StopTheSteal from three social media platforms after the 2020 US Presidential election and preceding the January 6 Capitol Riot. To do so, we utilized Snow and Benford’s typology of social movement frames—diagnostic, prognostic, and motivational frames—in the context of far-right movements and an additional frame device: violence cues. This study focused on the following three social media platforms: Facebook, Twitter, and Parler. We built three corpora of social media data: 26,093 Facebook posts, 248,643 tweets, and 400,600 Parler posts. Using Bidirectional Encoder Representations from Transformers (BERT) classifiers, dictionary methods, and qualitative text analysis, we find that the use of these frames varies by platform, with users on the alt-tech platform Parler using violence cues such as “smash” and “combat,” suggesting a greater call to action relative to the mainstream platforms.

Keywords

social media digital platforms political communication framing far-right social movement collective action frames

Following the defeat of former President Donald Trump in the 2020 U.S. Presidential election, Trump and his supporters disputed Joe Biden’s victory, alleging that the Democrats engaged in voting fraud. Efforts to overturn this outcome, described as the “Stop the Steal” (StS) movement, culminated in the storming of Capitol Hill on 6 January 2021, endangering the democratic process. Among the many factors that contributed to the alternative right (alt-right) StS movement, this study highlights their use of different social media platforms to disseminate election-related disinformation, mobilize supporters, and reinforce their in-group members’ political identities and nationalism.

We compared discourse from three social media platforms about the StS movement after the 2020 U.S. Presidential election and leading up to the January 6 Capitol Riots. Specifically, we compared social media discourse related to the StS movement on two mainstream platforms—Facebook and Twitter—and an alt-right platform Parler.

To analyze this discourse, we gathered 675,336 social media posts published during a 5-month period (1 September 2020 to 1 February 2021) and applied a mixed-methods (computational and qualitative) content analysis to identify framing activities employed when discussing far-right social movements. Applying Snow and Benford’s (1988) three collective action frames (diagnostic framing, prognostic framing, and motivational framing), we first investigated the prevalence of three frames on Facebook, Twitter, and Parler. In doing so, we demonstrate the nature of discourse in online activism across various social media platforms. The findings will help us understand how various social media affordances and content moderation practices might have facilitated collective action in a social movement.

Although mobilizing discourse on social media have been studied with regard to disenfranchised populations (e.g., Enjolras et al., 2013), it is also worth considering the ugly side of mobilizing discourse on social media; that is, how extremist movements utilize mobilizing discourse to advocate for violent solutions (Koopmans & Olzak, 2004). Our analysis accounts for this by studying an additional prominent framing device in the StS discourse: grievance-driven, pro-violence language features, which we conceptualize as a frame cue (McLeod & Shah, 2015). We compared the prevalence of violence cues in the discourse among three platforms, as well as the specific linguistic elements utilized by social media users to incite violence.

Literature Review

Political Discourse Across Different Social Media Platforms

The social media ecosystem is composed of many platforms, varying in modality (e.g., text, image, video), size, affordances, and moderating tools (Xu et al., 2022), and most US citizens use a myriad of platforms (Auxier & Anderson, 2021). For this reason, researchers have described social media consumption and production as “platformized” (Helmond, 2015), requiring social movements and political groups to construct messages that could be both homogeneous across platforms and tailored for a platform’s affordances. However, most social media research has overwhelmingly been single-platform analyses (Bode & Vraga, 2018). Addressing this gap with comparative analyses of platforms is critical to understand how discourses may vary across platform contexts (Matassi & Boczkowski, 2021).

One mechanism for contrasting social media is to consider whether they are distinct in their affordances. For example, users may use private or encrypted communication for talking to certain people but might opt for public-facing platforms to reach a wider audience. Different affordances of platforms can also contribute to a greater degree of the disclosure; for example, if users feel as if what they say will not be shared beyond that group (O’Leary et al., 2020). This can be good, as platforms can specialize in different types of content and build both mass and niche user bases. However, this can also allow groups to organize malicious online and offline activities on private or less popular social media platforms.

Alternatively, content across platforms can share some similarity, particularly when strategically published together. This is common in the case of advertising, where one advertising campaign is expected to have a unified message that works across multiple platforms (Laurie & Mortimer, 2019). Drawing from these techniques, organized groups, including state-sponsored trolls, have also built multi-platform political campaigns (Kreiss et al., 2018; Wagner & Boatright, 2019), producing shared narratives and claims that reach different audiences across the platforms (Shrestha et al., 2020).

The multi-platform structure of the social media ecosystem raises questions about how content across different social media may vary as a consequence of the platform’s affordances and its user base. While social media platforms support a range of communities, both political and non-political, we focus this question on one particular group, US far-right movements (and specifically the “Stop the Steal” effort) that are especially successful at using social media to advance their agenda (Schroeder, 2019).

Far-Right Movements on Social Media

Scholars have used social movement theory to understand the nature and context of extremist groups, including White supremacists and neo-Nazis (Bhat & Klein, 2020; Caiani & Della Porta, 2018). One subgenre in this scholarship focuses on far-right movements, defined as ultra-nationalists that reject out-groups and advocate for radical, sometimes violent actions to achieve political, economic, social, or cultural goals (Crosset et al., 2019). Studies examining far-right groups on social media show how they mobilize support and coordinate action online (Castelli Gattinara & Pirro, 2019). To be clear, we are not saying that social media causes far-right extremist beliefs.¹ Rather, social media enables the communication between far-right activists and prospective supporters (Mamié et al., 2021).

As a result, far-right groups have proliferated online, both on traditional social media (Winter, 2019) and on smaller, niche platforms known as “alt-tech” platforms, since they collectively frame themselves as alternatives to larger platforms (Wilson & Starbird, 2021). Within alt-tech platforms, far-right movements organize and mobilize offline events (Ekman, 2018), discuss and reframe mainstream narratives (Peucker & Fisher, 2023), and facilitate more amenable spaces for far-right rhetoric (Urman & Katz, 2022).

These platforms can be distinguished from their mainstream counterparts in two ways. First, some alt-tech platforms, such as “Gab,” emphasize their encrypted services, facilitating more private conversations (Shehabat et al., 2017). Second, alt-tech platforms often brand themselves as “free speech” havens (Munn, 2021), inviting anti-democratic and sometimes pro-violence discourse that would otherwise be routinely removed from mainstream platforms (Dori-Hacohen et al., 2021). Both the encrypted structure of these platforms and their ideological stance allow extremist communities and hate groups to build more resilient social media spaces wherein extremist narratives and conspiracy theories proliferate (Wilson & Starbird, 2021).

It is not clear the extent to which far-right discourse may be similar or different across these social media spaces. Literature on multi-platform social media dynamics broadly highlights the value of cross-platform coordination for genuine actors and disinformation agents (Kreiss et al., 2018). While we expect such groups to be somewhat aligned in their messaging (Mundt et al., 2018), the less formal hierarchical structure may produce disagreements about how a group manages their discourse (Gerbaudo, 2017). For example, members of a group may frame a topic using a unified critique; however, different factions of a group could advocate for different solutions across multiple platforms. This may be especially true if some users promote more heterodoxical actions (e.g., violence).

As mainstream social media platforms continue to suspend users posting or sharing far-right misinformation or calls to violence, particularly after the Capitol was attacked on 6 January (Aliapoulios et al., 2021), much of the far-right media ecosystem has relocated to alt-tech platforms (Zuckerman & Rajendra-Nicolucci, 2021). However, during the lead up to the 2020 election, far-right content proliferated on mainstream platforms like Twitter and Facebook (Muis et al., 2020). We focus on discourse about “Stop the Steal,” a far-right movement that promotes the false claim that the outcome of the 2020 US Presidential election was a result of electoral fraud (Homans & Peterson, 2022).

“Stop the Steal”

Building on the popularity of Trump’s claims of election fraud, which were further amplified across the right-wing media ecology (Bump, 2021), the “StS” movement is commonly understood as an extremist, far-right subset of Trump supporters, some of which have also advocated for other Trump-endorsed conspiracy theories (MacFarquhar, 2021). However, it is worth noting that the false claims had reached wider support among conservatives: post-election polls suggested 64% of Republicans believed that the election results were unreliable (C. Kim, 2020).

Attempts to reinstate Trump or stop the official vote count reached an apex when, on 6 January 2020, Trump supporters gathered in Washington D.C. to participate in several loosely related rallies and marches to “Stop the Steal.” In a bombastic speech, Trump urged his supporters to “fight like hell” (Andersen, 2021). During and after his speech, over 2,000 attendees marched to the Capitol Building and broke into the building, breaching the police perimeter and resulting in the death of five people.² Most Americans viewed the storming of the Capitol Building as an attack on US democracy, though the public remains split as to whether Trump bears direct responsibility for the attacks (Quinnipiac University Poll, 2022).

As a result of these criticisms, many mainstream social media platforms suspended users espousing far-right beliefs. Facebook focused on removing StS content or groups espousing the election fraud conspiracy, whereas Twitter primarily suspended users affiliated with the false QAnon conspiracy (a related, but distinct, far-right conspiracy group, see Booker, 2021). Trump himself was suspended permanently from Twitter and Facebook (Delkic, 2022), although his Facebook suspension was later changed to a temporary ban.

A substantial amount of policy and media attention has focused on how the StS movement utilized social media platforms to organize the 6 January rallies and marches (Timberg et al., 2021). Internal documents suggest that, in the days after the 2020 election, platforms like Facebook struggled to detect and moderate false election fraud claims (Levine, 2021) and, frustratingly, had few guidelines regarding how to moderate this content. As a result, the StS movement presents a unique opportunity to study how discourse in far-right movements varied across mainstream and alt-tech platforms.

Our analysis will focus on StS discourse in three platforms: Facebook, Twitter, and Parler. The first two are large, mainstream platforms with active StS discourse in the lead up to the election. The third is an alt-tech platform used by far-right protesters to organize the 6 January rallies (Munn, 2021). To study the discourse across these three platforms, we will use the concept of collective action frames. To ground our work, we ask the following:

RQ1. To what extent do Facebook, Twitter, and Parler vary in the prevalence of different collective action frames in #StopTheSteal discourse?

Collective Action Frames within StS Discourse

While media framing theory can be traced back to sociology and psychology research about print and broadcast news (Lecheler & De Vreese, 2019), the study of frames—defined as elements within a message that encourage a specific perspective or understanding of an event or issue (Matthes & Kohring, 2008)—in social media is understandably newer. This work has found that, because frames can emerge bottom-up, social media users are able to spread their political beliefs with relatively few barriers to entry (J. Kim, 2017). Thus, frames from social media content can promote alternative viewpoints on a political issue (Hamdy & Gomaa, 2012). Owing to the ephemeral attention of online audiences, social media frames can create “ad hoc issue publics” (Meraz & Papacharissi, 2013, p. 144) of which members may have multiple, scenario-specific affiliations (Bennett & Segerberg, 2012).

Drawing from the scholarship that has applied framing theory to study social media discourse (including its relationship to news media, see Aslett et al., 2022), our analysis will focus on collective action frames, which have been traditionally used to study social movements. We draw from this scholarly tradition to define framing as “the signifying work or meaning construction engaged in by movement adherents [. . .] and other actors [. . .] relevant to the interests of movements and the challenges they mount in pursuit of those interests” (Snow, 2013).

We define collective action frames as, “action-oriented sets of beliefs and meanings that inspire and legitimize the activities and campaigns of a social movement organization” (Benford & Snow, 2000). Social movements use three different types of collective action frames (Snow & Benford, 1988): diagnostic, prognostic, and motivational frames. The diagnostic frame identifies social problems (or victims of social problems), and attributes blame and responsibility, often delineating an “us” and “them” boundary (Gamson, 1995). The prognostic frame suggests proposed solutions, plans, or strategies in response to the identified problem. The motivational frame refers to a motivational drive to participate, such as by calling to arms or providing a rationale for joining in ameliorative or corrective action (Snow & Benford, 1988).

Empirical studies have identified and analyzed various types of diagnostic, prognostic, and motivational frames in right-wing social movements. In explaining the rise of right-wing parties in Denmark, Rydgren (2018) found these parties tended to blame specific ethnic groups for social problems (diagnostic frame) and to suggest stricter immigration policies and more law and order as solutions (prognostic frame). When studying an anti-Islamic social movement in Norway, Berntzen and Sandberg (2014) found that the social movement organizations diagnostically framed Islam as an existential threat, identified the prognosis as mostly non-violent and democratic (e.g., advocating a complete halt to non-Western immigration), and used a slogan to manifest the motivational frame, “fight for what is yours” to encourage activism.

We anticipate that Parler, our alt-tech case, will likely be more amenable to far-right conspiratorial misinformation and pro-violence discourse (Wahlström et al., 2021). Thus, we expect to see that the diagnostic frame, focusing on election fraud as the problem; the prognostic frame, focusing on political action; and the motivational frame, focusing on right-wing populism (defined as political rhetoric and ideology that argues society is divided into the pure people and the corrupt elite; Mudde, 2004)³ will be most prevalent on Parler. We hypothesize that

H1. The diagnostic (H1a), prognostic (H1b), and motivational (H1c) frame will be most prevalent on Parler.

Violence Frame Cues

As far-right groups are themselves not a monolith, it’s worth considering the extent to which far-right users will advocate for violence. Notably, not all far-right social movements advocate for violence. However, pro-violence discourse and violent actions may be especially prevalent in communities with hate groups and far-right organizations (Gaudette et al., 2022). Given the use of violence as a response to perceived election fraud, both in the United States and internationally, we explore whether pro-violence frame cues will vary across platforms.

Unlike the larger, more extensive claims that ground the aforementioned collective action frames (diagnostic, prognostic, and motivational), we anticipate that calls for violence will manifest as cues that guide information processing (Igartua & Cheng, 2009). Cues and frames are two components of discourse that are said to shape individuals’ cognitive processing and social judgments of issues, groups, and figures (Cho et al., 2006). Whereas frames refer to the underlying principles that organize and structure discourse about a topic, cues are the labels and terms utilized to highlight aspects of that issue (Shah et al., 2002). As such, while frames may exhibit consistency across different issues, cues often vary from issue to issue. For example, it is conceivable to observe the presence of diagnostic or prognostic frames in discourses of the #MeToo and #StopTheSteal movements, while the cues employed in the discourses are distinct. Despite being relatively small in linguistic size, a few choice words, cues can nevertheless affect people’s perception of political issues (Mondak, 1993).

We focus here on violence cues—cues that advocate for or talk about violence—because of far-right groups’ propensity to use more negative (Ceron & d’Adda, 2016), violent, dehumanizing (Wahlström et al., 2021), and exclusionary discourse (Krämer, 2017). The presence of far-right discourse across multiple platforms may contribute to the use of violence cues, albeit for different reasons. For example, alt-tech platforms may create discursive opportunities for extremists to advocate for pro-violent actions (Wahlström & Törnberg, 2019). In the context of the StS movement, we anticipate that violence cues may be especially prevalent in the alt-tech platform Parler, as there are fewer moderation policies that would limit or remove this content:

H2. The violence cues will be most prevalent on Parler.

However, violence cues may manifest differently across platforms. On mainstream platforms, with users across the political spectrum, online discursive arguments may contribute to a greater willingness to use violence offline (Gallacher et al., 2021). Thus, users on these platforms may use violence cues reactively, rather than consistently. In more heavily moderated social media platforms, users may resort to dog-whistling and subtle cues to signal their more extremist beliefs (Weimann & Am, 2020). We, therefore, look at the various ways in which users on different platforms have instigated violence:

RQ2. Which words or phrases did social media users on three platforms use to incite violence?

Method

Data Collection

We collected social media posts engaged in the StS movement from three social media platforms: Facebook, Twitter, and Parler. For each platform, we queried posts that contained the hashtag “#StoptheSteal.” We adopted a conservative keyword-searching strategy to reduce the noise brought by using multiple keywords. For example, if we query posts containing “Stop the Steal” or “sts,” we would get millions of posts that are actually criticizing, not promoting, the movement. However, posts that use the hashtag “#StoptheSteal” are often meant to amplify the movement, which makes our data collection more relevant for testing our hypotheses. In addition, we collected posts that were created from 1 September 2020 to 1 February 2021; this time frame allows us to look at the complete narrative of the #StoptheSteal discourse during the 2020 presidential election (Atlantic Council’s DFRLab, 2021).

Facebook

To gather Facebook data, we used CrowdTangle, which archives public content on Facebook. CrowdTangle currently covers more than 7 million Facebook pages, groups, and verified profiles.⁴

Twitter

We used the Twitter 2.0 application programming interface (API) to collect tweets. We only collected original tweets (excluding retweets). To ensure consistency across the platforms, only English tweets were collected.

Parler

To gather the Parler data, we drew from a publicly available archive of Parler data (Aliapoulios et al., 2021). This dataset includes 183 million posts made by 4 million users between 1 August 2018 and 11 January 2021. It does not contain information after January 2021, when the Parler website lost support from Amazon Web Services (AWS).

This collection yielded about 26,093 Facebook posts, 248,643 tweets, and 400,600 Parler posts.

Labeling Frames

We used a Bidirectional Encoder Representations from Transformers (BERT) model to train a classifier for coding the frames for each social media post. BERT is a deep learning technique that has been shown to outperform other classification algorithms and language models in common NLP (Natural Language Processing) tasks such as text classification (Devlin et al., 2019). Because of its proven performance in classification and its adaptability to different domains, BERT has been increasingly used in communication research (e.g., Lu et al., 2021; Moffitt et al., 2021). In this study, we used the pre-trained “DistilBERT” model—a lighter version of the standard “bert-base-uncased” model while preserving over 95% of its performance, and then fine-tuned the model with human-annotated data to perform the labeling.

Two human coders (two of three authors) produced a labeled dataset for whether a post contains the following collective action frames: diagnostic, prognostic, and motivational frames. We defined the diagnostic frame as focusing on the roots of the #StoptheSteal movement (e.g., “election fraud”). The prognosis frame referred to posts suggesting solutions and calling for actions, in other words, mobilizing collective action. Finally, the motivational frame spoke to the usage of motivational vocabularies such as “patriots,” “Americans,” “Republicans,” and “Trump supporters” to resonate with people and recruit participants for the #StoptheSteal movement. For each post, we coded 1 if it contains a certain frame and 0 otherwise. A single post could contain multiple frames or none of the three frames. After two rounds of coding, a sufficient inter-coder reliability level was reached between the two coders with an average Cohen’s Kappa is 0.88 (diagnostic: 0.91, prognostic: 0.88, motivational: 0.86). Finally, a total of 1,500 randomly sampled posts were labeled by the coders.

Three BERT models were trained to perform the classification tasks on the three labels: diagnostic frame, prognosis frame, and motivational frame. Model parameters such as batch size and learning rates were adjusted to optimize the models’ performance. The operationalization for each frame and evaluation metrics of the three models are reported in Table 1.

Table 1.

Operationalization of Frame Classification and Metrics of BERT Models.

Frames	Operationalization	Example	Cohen’s K	BERT models
				Precision	Recall	F1
Diagnostic frame	Problem statement and responsibility attribution, or what is going on and why	“ . . . this election is stolen, #voter fraud”	0.91	0.96	0.87	0.91
Prognostic frame	Suggesting solution to the problem—“what should we do”	“ . . . we must fight back . . . ”	0.88	0.95	1	0.97
Motivational frame	Usage of certain vocabularies to motivate people (e.g., “patriots”)	“ . . . every patriot should join us . . . ”	0.86	0.99	0.93	0.96

Note. BERT: bidirectional encoder representations from transformers.

Violence Cues

As cue frames can be as small as word choices (McLeod & Shah, 2015), we employed the Grievance Dictionary (van der Vegt et al., 2021) to detect violence cues. The Grievance Dictionary is a psycholinguistics dictionary that identifies 20,502 words signaling “grievance-fuelled violent language”; it includes 22 sub-categories, including “threat,” “violence,” or “paranoia.” For the purpose of this study, we only focused on the “violence” category, which includes keywords such as “bloodshed,” “fight,” and “bullet” (van der Vegt et al., 2021). The Grievance Dictionary provides a more contextualized list of language for violence/threat assessment than other dictionaries, such as the Linguistic Inquiry and Word Count (LIWC; van der Vegt et al., 2021).⁵ We applied the Grievance Dictionary to the Twitter, Facebook, and Parler datasets to calculate proportional scores for each and ran an analysis of variance (ANOVA) to examine if the proportional differences are statistically significant. In addition, we conducted a qualitative analysis of social media posts to supplement our findings and gain a more nuanced understanding of the context in which violence frame cues were employed.

Results

Descriptive Results

Figure 1 presents the weekly number of #StoptheSteal posts created between September 2020 and February 2021. The first wave of #StoptheSteal discourse was observed on Twitter around 7 September 2020, when far-right One America News correspondent Jack Posobiec tweeted, “#StoptheSteal 2020 is coming” (Atlantic Council’s DFRLab, 2021). The volume of #StoptheSteal discourse dramatically increased at the end of October 2020 and peaked around election day (3 November). Within the peak week, about 80,000 #StoptheSteal posts were generated on Parler and about 50,000 tweets were posted on Twitter. After the election day, there was a steady decrease in the #StoptheSteal posts, until another peak around 6 January 2021—the day when the vote results were to be verified by Congress, and the date of the Capitol Insurrection.

Figure 1.

Weekly number of #StoptheSteal posts on Facebook, Twitter, and Parler.

There are two noteworthy patterns. First, Twitter is the platform where the #StoptheSteal discourse gained its first wave of attention in early September 2020. Second, Parler generated the most #StoptheSteal posts between November 2020 to January 2021, followed by Twitter and Facebook. At its peak time, the number of #StoptheSteal posts on Parler was almost 1.5 times more than that on Twitter. The number of posts by frames and violence cues for each platform are presented in Appendix A.⁶

Notably, the spike in content, seen most clearly in Parler and Twitter, occurred soon after the election, with Parler producing far more content at its peak compared to Twitter. Activity persisted until 6 January. Thus, the “time range” of activity for #StoptheSteal posts, as expected, would be from the end of the 2020 election day to 6 January 2021.

Comparing Frames Across Platforms

To answer RQ1, which asked to what extent will Facebook, Twitter, and Parler vary in the prevalence of different collective action frames (diagnostic, prognosis, motivational), we calculated the percentage of posts with each frame for each platform, and the results were summarized in Table 2. The results show that for all three platforms, posts with a diagnostic or motivational frame are present, while the posts with a prognosis frame were less present (13% for Facebook, 21% for Parler, and 21% for Twitter). The most present collective action frame on Twitter is diagnostic (48%), while on Parler, the most present frame is motivational (53%), the same as that on Facebook (21%).

Table 2.

Proportions of Frames and Violence Cues across Facebook, Twitter, and Parler.

	Diagnostic frame	Motivational frame	Prognosis frame	Violence cue
Facebook	0.23	0.25	0.13	0.13
Twitter	0.48	0.46	0.21	0.21
Parler	0.49	0.53	0.21	0.2

Note. Numbers above the bars are proportions of posts containing three main collective action frames or violence cues. The sum of the proportions can exceed, or be under, 1 because one post may contain multiple frames. The results of proportion tests show that all pairwise comparisons are significant (p < .001) except differences between Twitter and Parler in prevalence of “prognostic frame” and “violence cues.”

To test H1a, H1b, and H1c, which hypothesized that the three types of collective action frames—diagnostic, prognostic, and motivational frames would be more prevalent on Parler than on Facebook and Twitter, we conducted a series of pairwise proportion tests. The results show that the proportion of diagnostic framed posts on Parler is significantly higher than that on Facebook (χ² = 6,998.5, p < .001) and Twitter (χ² = 61.35, p < .001). The proportion of prognosis frames on Parler is higher than on Facebook (χ² = 968.17, p < .001), but the difference is not significant between Parler and Twitter (χ² = 33.22, p = 1.00). The proportion of motivational frames on Parler is higher than on Facebook (χ² = 7,909.1, p < .001) and Twitter (χ² = 7,909.1, p < .001). Therefore, H1a and H1c were fully supported while H1b was partially supported.

To test H2, which hypothesized that the violence cues would be most prevalent on Parler, we compared the proportions of posts containing violence cues between three platforms. The results show that the posts containing violence cues are more prevalent on Parler than those on Facebook (χ² = 959.03, p < .001). However, this difference in proportions is not significant between Parler and Twitter (χ² = 18.9, p = 1.00). Therefore, H2 was not supported.

RQ2 asked which words or phrases social media users on three platforms use to incite violence. As shown in Table 3, “fight,” “law,” “dead,” “war,” and “power” were some of the most commonly used words that signaled violence across three platforms, according to the Grievance dictionary. When comparing the most frequently used words on each platform, some words were more present in one social media platform than in others. For instance, Facebook users used the words “racist,” “stress,” and “violent” more frequently than Parler and Twitter users. Parler users frequently and uniquely used the word “smash” and “combat,” whereas Twitter users used “push” and “murder.” “Jail” and “death” were prevalent on Twitter and Parler but not on Facebook. Selected examples of violent posts with unique violence cues are presented in Appendix B.⁷

Table 3.

The Top 15 Most Violence Cue Words Based on the Grievance Dictionary.

Facebook (N = 26,093)		Twitter (N = 248,643)		Parler (N = 400,600)
Words	Frequency	Words	Frequency	Words	Frequency
Fight	578	fight	7,586	fight	11,814
Power	236	dead	2,179	law	6,854
Law	209	law	2,103	power	4,436
war	185	push	2,141	dead	4,354
attack	157	power	1,723	war	3,282
dead	144	war	1,357	smash	2,038
racist	128	bad	1,100	bad	1,798
stress	117	hate	765	destroy	1,712
violent	80	destroy	725	hate	1,486
crime	85	crime	700	death	1,324
die	58	jail	606	back	1,552
riot	57	death	600	jail	1,352
arrest	52	murder	557	fear	1,318
destroy	69	attack	505	combat	1,426
prison	69	prison	500	attack	1,234

Note. Frequency number in the table shows the number of unique posts containing a violence cue word.

Qualitative Analysis of Violence Content

To further explore the RQ2, we conducted a qualitative discourse analysis of the violence frame cues to better understand how these cues were employed. In this analysis, we were especially interested in the semantic context of how cues were used and whether that varied across platforms. We used a grounded theory approach to observe the data, uncover patterns, and generate concepts (Glaser & Strauss, 2017). First, we compared the use of the top violence-related keywords on Twitter, Facebook, and Parler. Our findings reveal both similarities and differences. For example, across the three platforms, the word “fight” was used similarly, with social media users encouraging others to fight against election fraud and for Trump, justice, country, and democracy.

However, there were differences in how users utilized the term “law” in the three platforms. Twitter and Facebook users often used it as part of the noun phrase, “law and order,” a common policing refrain among conservatives (Wozniak et al., 2019), which was used in this context to argue that policing was necessary to reverse perceived election fraud. The word “laws” was also used to refer to mail-in ballot laws that were passed in the eleventh hour. For example, one Facebook user posted, “@ABC @GStephanopoulos You need to look into North Carolina because the election laws were changed at the last minute in that state like they were in Wisconsin.” In this post, the user raised suspected cases of voter fraud and demanded an investigation by the news media. Owing to its user base, users on Twitter were also especially likely to tag Trump, Republican politicians, and journalists in their tweets, urging them to take legal action to criticize voter fraud.

When Parler users used the word “law,” by comparison, they were urging Trump to declare martial law to stop voter fraud. This suggests greater advocacy for military involvement. Parler users also highlighted the laws that gave people the right to keep and bear arms: “#realdonaldjtrump should declare #martiallaw and order military to hold a complete election for POTUS or invoke the #12amendment immediately! #stopthesteal.”

We also looked into keywords that appeared uniquely on each platform. On Facebook, alt-right communities used the term “racist” to accuse “leftists”—particularly Black Lives Matter (BLM) demonstrators—of being racists or arguing that conservatives and alt-right users were not racist. BLM was frequently used as a negative foil in the StS discourse. For example, one Facebook post lamented, “Big Tech is censoring this peaceful rally! When was the last time Big Tech censored a violent rally for Black Lives Matter?” In this example, Facebook users frame BLM events as “violent,” compared to the “peaceful” #StoptheSteal “rally.” Furthermore, the user blames “Big Tech” as allies of BLM. Similar narratives, placing blame on social media platforms or Democratic and liberal groups, were pervasive across platforms.

Similarly, on Twitter, violence cues were often used to frame opponents as aggressive. For example, the term “murder” was frequently used to cast blame on liberal protesters. One tweet claimed, “#BBC hack Katy Kay described Trump supporters as a Mob. Never described #BLM a mob while they were Burning Looting and Murdering #LiberalHandwringingVermin #STOPTHESTEAL.” In this example, StS activists describe liberals as “vermin,” dehumanizing political opponents, and use terms like “looting and murdering,” which has historically been used to degenerate Black Americans (Johnson et al., 2011). Again, violence cues are not used to discuss their own actions, but the actions of BLM.

Compared to Twitter and Facebook, Parler discourse was far more aggressive. Parler users often used military metaphors to frame the Democratic party as enemies. This is especially evident when the term “combat” is used, as seen below:

Throw out any ballot not for Trump as it is ILLEGAL BY LAW!!! It’s time for Trump to take EMERGENCY action to #stopthesteal! REPOST THIS!! 1. Declare a NATIONAL EMERGENCY because of the treason from the Democrats. 2. Executive order to nullify the results of this FRAUD election! 3. Fire ANYONE who disagrees with this course of action, including in the military brass. Hire only those who will protect the President! 4. Declare the Democratic Party a TERRORIST organization—enemy combatants! All patriots who love liberty must defend President Trump and ensure at least another 4 years!!

In this example, the user goes so far as to describe the Democratic Party as a “terrorist organization—enemy combatants” and recommends military action to overturn the election.

While the overall prevalence of collective action frames and violence cues may not have varied greatly between Twitter and Parler, our qualitative analysis reveals a heightened level of absolutism and panic in posts from the alt-tech platform. Similarly, other violence cues such as “smash” and “jail” were used to advocate for “smashing” or “jailing” “the left” (encompassing Democratic politicians, Democratic voters, and BLM) or “Big Tech.” Parler users also frequently referenced being in “Twitter/Facebook jail” after their accounts or posts were taken down on other platforms. Despite criticizing these platforms for their suspensions, Parler users would simultaneously ask other users to share their posts on the mainstream platforms or would tag people they perceived as being able to influence their suspension: “This needs to get to @CodeMonkeyZ and I am in twitter jail! Please save and send!#MarchForTrump #stopthesteal.”

It is worth emphasizing that alt-tech platforms such as Parler make it easy for such discourses to proliferate. Unlike mainstream platforms that have some standards against hate speech, mis/disinformation, and other harmful content, Parler’s community guidelines (during and after 6 January) emphasized their desire to create a space “in the spirit of the First Amendment” (Parler, 2021), striving to minimize content moderation. This might explain why the discourse on Parler is substantively different from its mainstream counterparts, where such content would either be removed or flagged by users.⁸ As a consequence, while we see some similarity in the use of the collective action frames and violence frame cue, the level of extremity on Parler appears to be heightened, and perhaps more mobilizing.

Discussion

Social media platforms can be pivotal for groups to mobilize collective action, but different platforms may be more suited to different types of discourses. In this study, we focused on discourse about StS, examining how the movement and its claims were framed differently across three platforms. Specifically, we identified the use of diagnostic, prognostic, and motivational framing in this discourse, as well as violence cues. In studying these collective action frames across platforms in the context of alt-right communities, our results yield several curious findings. We find that, on the whole, Parler users evoke these collective action frames more than on mainstream platforms. However, Twitter discourse appeared to use the prognostic frame as much as Parler users.

These findings can be interpreted through the lens of platform affordance, specifically in relation to content moderation and anonymity. Facebook and Twitter’s content moderation policies, for example, may have encouraged or forced users to find alternative platforms with fewer content moderation policies, such as Parler. In these alternative spaces, users could garner a sustainable audience within a smaller, closed network, leading to the circulation of disinformation and posts inciting violence. This suggests that barring or removing content on one platform is insufficient to curb online political extremism or hate speech (Zuckerman & Rajendra-Nicolucci, 2021).

The varying levels of anonymity provided by the three platforms can explain why Parler and Twitter exhibited more prognostic and violent content than Facebook. Unlike Facebook users, who use real names, many Twitter and Parler users used pseudonyms. Political extremists have been found to leverage this anonymity to spread hate and White supremacism while making the far-right community more cohesive (Bhat & Klein, 2020; Crosset et al., 2019).

These findings advance three areas of scholarship. Methodologically, we present a mixed-methods approach for comparing frames across multiple social media platforms. Our approach to building a classifier suggests that bespoke, platform-specific classifiers still yield the most consistent results. Furthermore, our qualitative analysis allows us to assess the rich contexts of these frames in use.

Our results suggest that different platforms serve different organizing roles for far-right extremists, particularly when we compare mainstream platforms such as Twitter and Facebook to alt-tech platforms such as Parler. More specifically, if far-right extremist discourse is systematically removed from mainstream platforms, alt-tech platforms not only facilitate the spread of this content; they can also help translate it into offline form.

Finally, our study contributes to the ongoing use of social movements literature to understand far-right movements. While far-right extremism has gained more scholarly attention, only a few scholars have applied the social movement framework to understand these groups (Castelli Gattinara & Pirro, 2019). Our work builds on this by exploring the discourse of far-right movements through the lens of collective action frames, revealing that prognostic, diagnostic, and motivational frames used by far-right movements are conceptually distinct from those used by leftist social movements. Furthermore, our findings suggest that the use of violence cues is not inherently pro-violence; rather, these cues can frame opponents as enemy combatants. Thus, StS activists do not use violence cues to frame their own actions, but rather to portray political opponents as the “enemy.” This aligns with previous literature on the use of fear to evoke action among far-right communities (Scrivens & Amarasingam, 2020) and shows how far-right activists can further prime such fears through collective action frames.

Of course, this study is not without limitations. First, the representativeness of the data is constrained due to limitations in data collection. Facebook data were collected post hoc, and there might be missing data that was deleted after the January 6 Capitol attack, which cannot be retrieved. We highlight the challenges in accessing data when studying content such as mis/disinformation, conspiratorial beliefs, hate speech, and calls for violence. It is also worth mentioning that mainstream platforms, such as Facebook or Twitter, have posts containing the hashtag #StoptheSteal that opposed the StS narrative; however, our analysis did not differentiate between posts that supported the StS discourse and those that presented counter-arguments. We suggest future research to examine how Stop the Steal narratives varied on mainstream platforms. In addition, while we do study multiple platforms, the social media ecology in the United States comprises many more platforms, catering to different communities and offering different affordances (including, but not limited to, multi-modal functionalities). Finally, our analysis is comparative. The goal of this work is not to study how content from one platform moves to another, but rather to compare and contrast frames across these platforms. Future research can build on these findings by studying temporal relationships or the mobilizing effectiveness of these framing devices. The scholarship on cross-platform literature is nascent, and it is worth considering how descriptive analyses such as ours can lay the foundation for future causal work.

Nevertheless, we argue that our results have important implications for both understanding discourse about and from far-right group activity and for those interested in comparing alt-tech and mainstream social media platforms.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Bin Chen

Gyo Hyun Koo

Notes

Author Biographies

Bin Chen is a doctoral candidate in the School of Journalism and Media at the University of Texas at Austin. His research focuses on digital platforms and political communication.

Josephine Lukito is an Assistant Professor at the University of Texas at Austin’s School of Journalism and Media. She is also the Director of the Media & Democracy Data Cooperative and a Senior Faculty Research Affiliate for the Center for Media Engagement.

Gyo Hyun Koo (PhD University of Texas at Austin) is an assistant professor at Howard University’s Department of Communication, Culture and Media Studies. Her research focuses on individuals’ perceptions and engagement with online information and news.

References

Aliapoulios

Bevensee

Blackburn

Bradlyn

De Cristofaro

Stringhini

Zannettou

. (2021). An early look at the Parler online social network. arXiv. https://doi.org/10.48550/arXiv.2101.03820

Andersen

(2021, January 6). Before mob stormed US Capitol, Trump told them to ‘fight like hell.’ The Boston Globe. https://www.bostonglobe.com/2021/01/06/metro/heres-what-trump-told-his-supporters-before-many-them-stormed-capitol-wednesday/

Aslett

Webb Williams

Casas

Zuidema

Wilkerson

. (2022). What was the problem in parkland? Using social media to measure the effectiveness of issue frames. Policy Studies Journal, 50(1), 266–289. https://doi.org/10.1111/psj.12410

Atlantic Council’s DFRLab. (2021). #StopTheSteal: Timeline of social media and extremist activities leading to 1/6 insurrection. https://www.justsecurity.org/74622/stopthesteal-timeline-of-social-media-and-extremist-activities-leading-to-1-6-insurrection/

Auxier

Anderson

(2021). Social media use in 2021. Pew Research Center, 1, 1–4.

Benford

Snow

. (2000). Framing processes and social movements: An overview and assessment. Annual Review of Sociology, 26, 611–639.

Bennett

Segerberg

. (2012). The logic of connective action: Digital media and the personalization of contentious politics. Information, Communication & Society, 15(5), 739–768. https://doi.org/10.1080/1369118X.2012.670661

Berntzen

Sandberg

. (2014). The collective nature of lone wolf terrorism: Anders Behring Breivik and the anti-Islamic social movement. Terrorism and Political Violence, 26, 759–779. https://doi.org/10.1080/09546553.2013.767245

Bhat

Klein

(2020). Covert hate speech: White nationalists and dog whistle communication on twitter. In Bouvier

Rosenbaum

J. E.

(Eds.), Twitter, the Public Sphere, and the Chaos of Online Deliberation, (pp. 151–172). Macmillan Cham.

10.

Bode

Vraga

. (2018). Studying politics across media. Political Communication, 35, 1–7.

11.

Booker

. (2021, January 12). Facebook removes “Stop The Steal” content; Twitter suspends QAnon accounts. NPR. https://www.npr.org/sections/insurrection-at-the-capitol/2021/01/12/956003580/facebook-removes-stop-the-steal-content-twitter-suspends-qanon-accounts

12.

Boulianne

Koc-Michalska

Bimber

. (2020). Right-wing populism, social media and echo chambers in Western democracies. New Media & Society, 22(4), 683–699. https://doi.org/10.1177/1461444819893983

13.

Bump

. (2021, October 22). The voter-fraud allegation that perfectly captures the post-2020 Republican Party. The Wash-ington Post. https://www.washingtonpost.com/politics/2021/10/22/voter-fraud-allegation-that-perfectly-captures-post-2020-republican-party/

14.

Caiani

Della Porta

(2018). The radical right as social movement organizations. In Rydgren

(Ed,), The Oxford Handbook of the Radical Right, (pp. 327–347). Oxford University Press.

15.

Castelli Gattinara

Pirro

. (2019). The far right as social movement. European Societies, 21(4), 447–462. https://doi.org/10.1080/14616696.2018.1494301

16.

Ceron

d’Adda

. (2016). E-campaigning on Twitter: The effectiveness of distributive promises and negative campaign in the 2013 Italian election. New Media & Society, 18(9), 1935–1955.

17.

Cho

de Zuniga

H. G.

Shah

McLeod

. (2006). Cue con-vergence: Associative effects on social intolerance. Commu-nication Research, 33(3), 136–154. https://doi.org/10.1177/0093650206287075

18.

Crosset

Tanner

Campana

. (2019). Researching far right groups on Twitter: Methodological challenges 2.0. New Media & Society, 21(4), 939–961. https://doi.org/10.1177/1461444818817306

19.

Culliford

Nellis

. (2021). Google suspends Parler social networking app from Play Store; Apple gives 24-hour warning. Reuters. https://www.reuters.com/article/us-apple-parler-idUKKBN29D34N

20.

Delkic

. (2022, May 10). Trump’s banishment from Facebook and Twitter: A timeline. The New York Times. https://www.nytimes.com/2022/05/10/technology/trump-social-media-ban-timeline.html

21.

Devlin

Chang

M.-W.

Lee

Toutanova

. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (arXiv: 1810.04805). arXiv. https://doi.org/10.48550/arXiv.1810.04805

22.

Dori-Hacohen

Sung

Chou

Lustig-Gonzalez

. (2021). Restoring healthy online discourse by detecting and reducing controversy, misinformation, and toxicity online. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (pp. 2627–2628). ACM. https://doi.org/10.1145/3404835.3464926

23.

Ekman

. (2018). Anti-refugee mobilization in social media: The case of soldiers of Odin. Social Media + Society, 4(1), 2056305118764431. https://doi.org/10.1177/2056305118764431

24.

Engesser

Ernst

Esser

Büchel

. (2017). Populism and social media: How politicians spread a fragmented ideology. Information, Communication & Society, 20(8), 1109–1126. https://doi.org/10.1080/1369118X.2016.1207697

25.

Enjolras

Steen-Johnsen

Wollebæk

. (2013). Social media and mobilization to offline demonstrations: Transcending participatory divides? New Media & Society, 15(6), 890–908. https://doi.org/10.1177/1461444812462844

26.

Farley

(2021, November 1). How many died as a result of capitol riot? FactCheck.Org. https://www.factcheck.org/2021/11/how-many-died-as-a-result-of-capitol-riot/

27.

Gallacher

Heerdink

Hewstone

. (2021). Online engagement between opposing political protest groups via social media is linked to physical violence of offline encounters. Social Media + Society, 7(1), 2056305120984445.

28.

Gamson

. (1995). Must identity movements self-destruct? A queer dilemma. Social Problems, 42(3), 390–407. https://doi.org/10.2307/3096854

29.

Gaudette

Scrivens

Venkatesh

. (2022). The role of the internet in facilitating violent extremism: Insights from former right-wing extremists. Terrorism and Political Violence, 34(7), 1339–1356. https://doi.org/10.1080/09546553.2020.1784147

30.

Gerbaudo

. (2017). Social media teams as digital vanguards: The question of leadership in the management of key Facebook and Twitter accounts of Occupy Wall Street, Indignados and UK Uncut. Information, Communication & Society, 20(2), 185–202. https://doi.org/10.1080/1369118X.2016.1161817

31.

Glaser

Strauss

(2017). Discovery of grounded theory: Strategies for qualitative research. Routledge.

32.

Hamdy

Gomaa

. (2012). Framing the Egyptian uprising in Arabic language newspapers and social media. Journal of Communication, 62(2), 195–211. https://doi.org/10.1111/j.1460-2466.2012.01637.x

33.

Helmond

. (2015). The platformization of the web: Making web data platform ready. Social Media + Society, 1(2), 2056305115603080. https://doi.org/10.1177/205630511560-3080

34.

Homans

Peterson

. (2022, July 19). How “Stop the Steal” captured the American right. The New York Times. https://www.nytimes.com/2022/07/19/magazine/stop-the-steal.html

35.

Igartua

J.-J.

Cheng

. (2009). Moderating effect of group cue while processing news on immigration: Is the framing effect a heuristic process? Journal of Communication, 59(4), 726–749. https://doi.org/10.1111/j.1460-2466.2009.01454.x

36.

Johnson

Dolan

M. K.

Sonnett

. (2011). Speaking of looting: An analysis of racial propaganda in national television coverage of Hurricane Katrina. Howard Journal of Communications, 22(3), 302–318.

37.

Kim

. (2020, November 9). Poll: 70 percent of Republicans don’t think the election was free and fair. POLITICO. https://www.politico.com/news/2020/11/09/republicans-free-fair-elections-435488

38.

Kim

. (2017). #iamafeminist as the “mother tag”: Feminist identification and activism against misogyny on Twitter in South Korea. Feminist Media Studies, 17(5), 804–820. https://doi.org/10.1080/14680777.2017.1283343

39.

Koopmans

Olzak

. (2004). Discursive opportunities and the evolution of right-wing violence in Germany. American Journal of Sociology, 110(1), 198–230. https://doi.org/10.1086/386271

40.

Krämer

. (2017). Populist online practices: The function of the Internet in right-wing populism. Information, Communication & Society, 20(9), 1293–1309.

41.

Kreiss

Lawrence

R. G.

McGregor

. (2018). In their own words: Political practitioner accounts of candidates, audiences, affordances, genres, and timing in strategic social media use. Political Communication, 35(1), 8–31. https://doi.org/10.1080/10584609.2017.1334727

42.

Laurie

Mortimer

. (2019). How to achieve true integration: The impact of integrated marketing communication on the client/agency relationship. Journal of Marketing Management, 35(3–4), 231–252. https://doi.org/10.1080/0267257X.2019.1576755

43.

Lecheler

De Vreese

. (2019). News framing effects: The-ory and practice. Taylor & Francis. https://doi.org/10.4324/9781315208077

44.

Levine

. (2021, October 25). Inside Facebook’s struggle to contain insurrectionists’ posts. POLITICO. https://www.politico.com/news/2021/10/25/facebook-jan-6-election-claims-516997

45.

Pan

. (2021). Public sentiment on Chinese social media during the emergence of COVID19. Journal of Quantitative Description: Digital Media, 1, 1–47. https://doi.org/10.51685/jqd.2021.013

46.

MacFarquhar

. (2021, March 26). Far-right extremists move from “Stop the Steal” to stop the vaccine. The New York Times. https://www.nytimes.com/2021/03/26/us/far-right-extremism-anti-vaccine.html

47.

Mamié

Horta Ribeiro

West

. (2021). Are anti-feminist communities gateways to the far right? Evidence from Reddit and YouTube. In 13th ACM Web Science Conference (pp. 139–147). ACM. https://doi.org/10.1145/3447535.3462504

48.

Matassi

Boczkowski

. (2021). An agenda for comparative social media studies: The value of understanding practices from cross-national, cross-media, and cross-platform perspectives. International Journal of Communication, 15, 207–228.

49.

Matthes

Kohring

. (2008). The content analysis of media frames: Toward improving reliability and validity. Journal of Communication, 58(2), 258–279. https://doi.org/10.1111/j.1460-2466.2008.00384.x

50.

McLeod

Shah

. (2015). News frames and national security. Cambridge University Press.

51.

Meraz

Papacharissi

. (2013). Networked gatekeeping and networked framing on #Egypt. The International Jour-nal of Press/Politics, 18(2), 138–166. https://doi.org/10.1177/1940161212474472

52.

Moffitt

King

Carley

. (2021). Hunting conspiracy theories during the COVID-19 Pandemic. Social Media + Society, 7(3), 20563051211043212. https://doi.org/10.1177/20563051211043212

53.

Mondak

J. J

. (1993). Source cues and policy approval: The cognitive dynamics of public support for the Reagan agenda. American Journal of Political Science, 37, 186–212.

54.

Mudde

(2004). The populist zeitgeist. Government and Opposition, 39(4), 541–563.

55.

Muis

Klein

Dijkstra

. (2020). Challenges and opportunities of social media research: Using Twitter and Facebook to investigate far right discourses. In Ashe

S. D.

Busher

Macklin

Winter

(Eds.), Researching the far right: Theory, method and practice (pp. 147–163). Routledge.

56.

Mundt

Ross

Burnett

C. M.

(2018). Scaling social movements through social media: The case of Black Lives Matter. Social Media+ Society, 4(4), 2056305118807911.

57.

Munn

. (2021). More than a mob: Parler as preparatory media for the U.S. Capitol storming. First Monday, 26(3). https://doi.org/10.5210/fm.v26i3.11574

58.

O’Harrow

Jr. Tran

A. B.

Hawkins

(2021, April 12). The rise of domestic extremism in America. The Washington Post. https://www.washingtonpost.com/investigations/interactive/2021/domestic-terrorism-data/

59.

O’Leary

Coulson

Perez-Vallejos

McAuley

. (2020). Towards understanding how individuals with inflammatory bowel disease use contemporary social media platforms for health-related discourse. Computers in Human Behavior, 112, Article 106463. https://doi.org/10.1016/j.chb.2020.106463

60.

Oliver

Rahn

. (2016). Rise of the Trumpenvolk: Populism in the 2016 election. The ANNALS of the American Academy of Political and Social Science, 667(1), 189–206. https://doi.org/10.1177/0002716216662639

61.

Parler. (2021). Community guidelines. https://nationalfile.com/wp-content/uploads/2021/02/Parler-Community-Guidelines.pdf

62.

Pennebaker

Boyd

Jordan

Blackburn

. (2015). The development and psychometric properties of LIWC2015. https://repositories.lib.utexas.edu/handle/2152/31333

63.

Peucker

Fisher

T. J

. (2023). Mainstream media use for far-right mobilisation on the alt-tech online platform Gab. Media, Culture & Society, 45(2), 354–372. https://doi.org/10.1177/01634437221111943

64.

Quinnipiac University Poll. (2022). 64% think the January 6 attack on the capitol was planned, Quinnipiac university national poll finds; number of Americans viewing rising prices as a crisis climbs. https://poll.qu.edu/poll-release?releaseid=3850

65.

Rydgren

. (2018). The Oxford handbook of the radical right. Oxford University Press.

66.

Schroeder

. (2019). Digital media and the entrenchment of right-wing populist agendas. Social Media + Society, 5(4), 2056305119885328.

67.

Scrivens

Amarasingam

. (2020). Haters gonna “Like”: Exploring Canadian far-right extremism on Facebook. In Littler

Lee

(Eds.), Digital extremisms: Readings in violence, radicalisation and extremism in the online space (pp. 63–89). Springer. https://doi.org/10.1007/978-3-030-30138-5_4

68.

Shah

D. V.

Watts

M. D.

Domke

Fan

D. P.

(2002). News framing and cueing of issue regimes: Explaining Clinton’s public approval in spite of scandal. Public Opinion Quarterly, 66(3), 339–370.

69.

Shehabat

Mitew

Alzoubi

. (2017). Encrypted Jihad: Investigating the role of Telegram app in lone wolf attacks in the West. Journal of Strategic Security, 10, 27–53.

70.

Shrestha

Sathanur

Maharjan

Saldanha

Arendt

Volkova

. (2020). Multiple social platforms reveal actionable signals for software vulnerability awareness: A study of GitHub, Twitter and Reddit. PLOS ONE, 15(3), Article e0230250. https://doi.org/10.1371/journal.pone.0230250

71.

Snow

. (2013). Framing and social movements. In D. McAdam, D. Della Porta, B. Klandermans, & D. A. Snow (Eds.), The Wiley-Blackwell Encyclopedia of Social and Political Movements. John Wiley & Sons. https://doi.org/10.1002/9780470674871.wbespm434

72.

Snow

Benford

. (1988). Ideology, frame resonance and participant mobilization. International Social Movement Research, 1, 197–217.

73.

Timberg

Dwoskin

Albergotti

. (2021, October 22). Inside Facebook, Jan. 6 violence fueled anger, regret over missed warning signs. The Washington Post. https://www.washingtonpost.com/technology/2021/10/22/jan-6-capitol-riot-facebook/

74.

Urman

Katz

. (2022). What they do in the shadows: Examining the far-right networks on Telegram. Information, Communication & Society, 25(7), 904–923. https://doi.org/10.1080/1369118X.2020.1803946

75.

van der Vegt

Mozes

Kleinberg

Gill

. (2021). The grievance dictionary: Understanding threatening language use. Behavior Research Methods, 53(5), 2105–2119. https://doi.org/10.3758/s13428-021-01536-2

76.

Wagner

Boatright

. (2019). Generating and framing content: Strategic multi-platform content marketing in academic libraries. Public Services Quarterly, 15(1), 59–67. https://doi.org/10.1080/15228959.2018.1555074

77.

Wahlström

Törnberg

. (2019). Social media mechanisms for right-wing political violence in the 21st century: Discursive opportunities, group dynamics, and co-ordination. Terrorism and Political Violence, 33(4), 766–787.

78.

Wahlström

Törnberg

Ekbrand

. (2021). Dynamics of violent and dehumanizing rhetoric in far-right social media. New Media & Society, 23(11), 3290–3311. https://doi.org/10.1177/1461444820952795

79.

Weimann

. (2020). Digital dog whistles: The new online language of extremism. International Journal of Security Studies, 2(1), Article 4.

80.

Wilson

Starbird

. (2021). Cross-platform information operations: Mobilizing narratives & building resilience through both “big” & “alt” Tech. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW2), 1–32. https://doi.org/10.1145/3476086

81.

Winter

. (2019). Online hate: From the far-right to the “alt-right” and from the margins to the mainstream. In Lumsden

Harmer

(Eds.), Online othering: Exploring digital violence and discrimination on the web (pp. 39–63). Springer. https://doi.org/10.1007/978-3-030-12633-9_2

82.

Wozniak

Calfano

Drakulich

. (2019). A “Ferguson effect” on 2016 presidential vote preference? Findings from a framing experiment examining “shy voters” and cues related to policing and social unrest. Social Science Quarterly, 100(4), 1023–1038.

83.

Hui

P. M.

Jha

O. K.

Xia

Johnson

. (2022). Preventing the spread of online harms: Physics of contagion across multi-platform social media and metaverses. http://arxiv.org/abs/2201.04249

84.

Zuckerman

Rajendra-Nicolucci

. (2021). Deplatforming our way to the alt-tech ecosystem. http://knightcolumbia.org/blog/deplatforming-our-way-to-the-alt-tech-ecosystem