Political Astroturfing on Twitter: How to Coordinate a Disinformation Campaign

“…level have become more influential, political actors such as governments (King, Pan, & Roberts, 2017; Lukito et al., 2018), hyperpartisan media (Allcott & Gentzkow, 2017; Faris et al., 2017), far-right groups (Marwick & Lewis, 2017) or actors with economic motives (Silverman & Alexander, 2016)…”

“…a “social movement grows when people with grievances meet, agree on a common agenda, and organize for political action” (Howard, 2006, p. 86; see also Bennett & Segerberg, 2013), an astroturfing campaign tries to appear like such an organic expression of public opinion, but is actually centrally…”

“, they tweet on the same days – and groups them together (Hartigan, 1972).”

“Tweets posted by known and suspect NIS accounts should reflect these goals and strategies, and therefore differ significantly from those posted by regular users, political users, and opinion leaders (Broniatowski et al., 2018).”

“In order to prevent this type of shirking, the principal may opt for constant and direct supervision by physically locating all agents in the same office – as done by the “Russian troll factories”(3) – or at least hire them to work during office hours instead of paying by post, as the eponymous “50 cent party” alleged…”

“Social media also have the potential to set the news agenda (Chadwick, 2013; Neuman, Guggenheim, Jang, & Bae, 2014) as journalists pay close attention to social media buzz (Lukito et al.”

“To better understand how the patterns of the two types of campaigns should differ, we turn to principal-agent theory (Miller, 2005; Ross, 1973).”

“The focus is on the UK because whilst the UK and USA both have large numbers of Tweeters, extensive astroturfing (Keller et al., 2020) seems likely in the USA in 2020 in the run up to the presidential election.”

“Having said this, the risk of emergence of some power law discrepancy seems larger if the algorithm is also learning from the future inputs given by the environment, especially if such inputs are adversarial (see [KSSY19] for an example of attacks on algorithms).”

“In both cases we are confident that the collected samples are nearly complete within a specific time window.”

“Uma parte significativa desses esforços se concentra em analisar a disseminação de conteúdo polı́tico no Twitter [Pacheco et al. 2020, Keller et al. 2020, Grimminger and Klinger 2021], identificando e caracterizando ações coordenadas para promover conteúdo polı́tico por meio de modelos de…”

“For example, beyond orchestrated political disinformation campaigns (e.g., Keller et al., 2019; Krafft & Donovan, 2020), AI-disinformation may be used in interpersonal communication (i.e., WhatsApp ) and used to harm specific individuals in one’s own network.”

“Information warfare often involves dis-information campaigns, which can be carried out by very different means – from automation to (state) coordination of human actors (Metaxas and Mustafaraj 2012; Grimme et al. 2017; Keller et al. 2020).”

“Adversaries may employ fake accounts to interact with some content to give the impression that the content is popular, a deceptive strategy that is similar to political astroturfing [16].”

“This conceptual framework synthesizes and aligns with a diverse collection of research strands which examine the framing strategies of information operations and their concerted control of information flow to shape who hears what message and from whom [5, 14, 36, 38].”

“De hecho, Twitter ha sido objeto de varias campañas de desinformación durante períodos electorales, que se caracterizaron por el uso masivo de bots (Keller et al., 2020).”

“Disinformation campaigns on social media employ accounts with coordinated behaviour (Keller et al. 2020; Elmas et al. 2021).”

“Researchers have looked at disinformation in the 2012 Korean Presidential Elections [137], 2016 US Elections [39, 193, 265], the French National Elections [92], the Brexit Vote [130] as well as other events.”

“CSMM campaigns are often organized centrally, ‘initiated by a principal directly instructing a group of users who respond to extrinsic rewards—the agents’ (Keller et al., 2020: 259).”

“As part of the opposition efforts to undermine health campaigns, the principals of anti-tobacco control astroturfing (e.g., the tobacco industry) can rely on automated or human agents (e.g., advocates) to disseminate anti-tobacco control messages (Keller et al., 2019).”

“Similar observations were reported by Keller et al. (2020), who attributed this tendency to the division of astroturfing agents into smaller teams operating in various Internet cafes to orchestrate a disinformation campaign.”

“…CIB, CB, and social bots (i.e., software agents that communicate autonomously on social media and other platforms, having different uses and purposes [20–23]) may have played a role in relevant events [24] such as political elections [25–27], and in disseminating false information [28–32].”

“Identifying astroturfing campaigns is challenging without ground truth information; Keller et al. (2020) provide evidence that coordinated group activity is a robust hallmark of astroturfing.”

“Keller et al. [25] suggested that misinformation may be propagated as a “Centrally Organized Message Coordination” in which fake or biased digital news can be amplified on the social media by different methods such as maneuvered passive re-sharing or fabricated comments by opinion leaders [19].”

“First, we selected a 60-second time window 4 to define rapid actions, allowing us to identify potential coordination by users acting simultaneously (Keller et al. 2020).”

“However, it has become evident that benign actors, such as activists, are using similar techniques and malicious actors are engaged in political astroturfing [24] and the dissemination of inappropriate content.”

“Keller et al. [25], in turn, explored the network of retweets to study ad coordination during the presidential election in South Korea, whereas Nizzoli et al. [26], proposed a network-based framework to detect various degrees of coordinated social media efforts during the 2019 UK general election.”

“tweet [11,19,24], using the same URL [3,14,20,24], similar text [9], and similar media [19,26], and any other ‘behavioural trace’ [19] that could link accounts (e.”

“This shift in perspective makes it possible to expose co-ordinated efforts by identifying an anomalous degree of consistency within a group of accounts [10].”

“As an artificial network that eyes a collective goal, the Russian network bears similarities to an astroturf, a fake social movement that promotes paid agenda in concert (Mix and Waldo 2015; Keller et al. 2020).”

“As others (Keller et al., 2020; Weeks & Gil de Zúñiga, 2019) have noted, scholarship can become bogged down in definitional work and creating taxonomies for false information.”

“…seen an explosion of interest in how disinformation interfaces with democratic outcomes, ranging from studies of disinformation campaign anato-mies (Keller et al. 2020) and external election interference (Lysenko and Brooks 2018) to the role of disinformation in creating alternative ontologies…”

“Most studies on coordination detection use text similarity features [2, 4, 25], co-retweeting (or repeatedly retweeting specific accounts or tweets to amplify messaging) [19, 20, 46], synchronicity (temporally correlated activity) [10, 48], or some combination [24, 44, 45, 47].”

“More concretely, this situation can be related to astroturf-ing campaigns as explained by Keller et al. (2020), when in the flood phase there is a significant increase in tweets with the help of automated bots (Arce-Garcia et al. 2022; Arce-Garcia et al. 2023).”

“Such phenomena are termed ‘astroturfing’ in the political science literature, whereby firms or entities create multiple social media accounts to simulate grassroots support or opposition to influence public opinion (Keller et al. 2020).”

“These vulnerable users then begin sharing the disinformation to increase its reach online (Woolley 2020; Keller et al. 2020; Pierri et al. 2020; Yao et al. 2017; Zannettou et al. 2019).”

“This situation reveals the point that the issue of disinformation on Twitter (Keller et al., 2020; Linvill & Warren, 2020) should not be ignored.”

“It is designed to manufacture public support (Keller et al., 2020; Linvill and Warren, 2020), intensify ideological, racial, and ethnic divisions (Marwick and Lewis, 2017; Yadav et al., 2023), and legitimize undemocratic policies (Sinpeng et al., 2020; Soares and Recuero, 2021).”

“We did not direct the study towards a content analysis of fact‐checkers’ websites and so we are unable to determine the proportion of Covid‐19 vaccine content that was shared between fact‐checkers’ websites and their watchdog role performances, as the scrutiny of top‐down actors, did not account for disinformation cam…”

“Some methods focus on specific components of the content, such as embedded links, hashtags, and media [39, 28, 20, 21, 51].”

“For instance, accounts created within narrow time windows may signal bot networks (Abrahams & Leber, 2021; Jones, 2019), and bots may also be identified through reposting patterns shared with known bot accounts (Keller et al., 2020).”

“Conversely, coordination can also be employed for goals widely recognized as problematic, such as disinformation campaigns [8, 9, 10], election interference [3, 11], or astroturfing [12].”

“The speed and virality of tweets, combined with the limited character count, can make it easier for disinformation to spread before it is verified or corrected (Keller et al., 2020).”

“Nevertheless, the spreading of fake news in social discourse due to these organised political activities by troll farms, interest groups or the bots they design is minimal (Badawy et al., 2018; Keller et al., 2020; Zannettou et al., 2019).”

“Astroturfing is a global phenomenon that recently became a serious problem for online environments (Chan 2022), the consequences of which are being studied on examples (apart from the IRA mentioned above) from China (Han 2015), the Philippines (Ong and Cabañes 2019), Turkey (Bulut and Yörük 2017), South Korea (Keller…”

“For example, Retweet and Co-Retweet networks have been used to identify and characterize coordinated astroturf-ing (Keller et al. 2020), while others have explored Co-Tweet or Co-Reply signals to uncover coordinated behavior (Graham et al. 2024).”

“However, a minority of users produce highly repetitive texts and exhibit low complexity, which may indicate the presence of automated or coordinated accounts [10,14,29,30,38,41].”

“These artificial grassroots movements (so-called astroturfing) are intended to infiltrate the discourse with political content, identifiable via keyword hashtags, across platforms and to drive the public in front of them (Keller et al. 2020).”

“Previous studies have proposed different frameworks to identify coordination among actors by examining platform-specific features such as co-retweet or co-tweet networks [9], or the retweet network [8].”

“…made these platforms vulnerable to manipulation of discourse by bad actors who utilize automated accounts (Ferrara et al., 2016; Varol et al., 2017), spread disinformation (Mosleh and Rand, 2022; Keller et al., 2020), and coordinate targeted attacks (Shao et al., 2018; Varol and Uluturk, 2020).”

“, 2017), disinformation campaigns (Keller et al., 2020), as well as elite communication with citizens (e.”

“Recent studies have explored political astroturfing and its effects on online engagement (e.g., Arce‐García et al., 2022; Keller et al., 2020), yet little is known about how such practices integrate disinformation narratives for mobilization and negative campaigning in influencer‐driven…”

“…prioritization means that messages heavily laden with public opinion cues are more likely to surface in social media users’ feeds (Twitter, no date; Bucher, 2012; Backstrom, 2013; Facebook, 2014, 2016; Westerwick, Johnson and Knobloch-Westerwick, 2017; Owen, 2019; Ho, 2020; Keller et al. , 2020).”

“To complicate things further, many bots are partiallymanaged “cyborgs” that disseminate content created manually by teams [3][16].”

“…Roth 2019), we first map the Twittersphere revolving around the Iran protests and detect key clusters (RQ1), and, second, focus on the evidence of the presence of bots, bot networks, and coordinated misinformation campaigns and political astroturfing (Keller et al. 2019) in these clusters (RQ2).”

“Notable examples of this civic engagement include the 2008 can-dlelight protests against U.S. beef imports (Yun and Chang, 2011) and the 2016–2017 protests that resulted in the impeachment of the former President Park Geun-hye over abuse of power and corruption (Keller et al., 2020).”

“…Rid, 2020), “astroturfing” which describes the wide use of inauthentic, often bot, accounts that pretend to be real users supporting a message (Keller et al., 2020), and efforts to prompt users to interpret digital affordances, perceived pathways for acting in a digital environment (e.g., a…”

“…text-template campaigns (a.k.a. “copy-pastas”) linked to astroturfing behavior that simulates grassroots consensus (Vishnuprasad et al., 2024; Keller et al., 2020) and concentrated retweeting (Elmas et al., 2022; Schoch et al., 2022) remain among the simplest yet most persistent…”