In the Shadow of Outcry: The Disparate Online Harassment of Diaspora Female Voices in the Iranian Protests of 2022
The digital age has ushered in a plethora of opportunities for journalists and activists worldwide. However, with it comes a dark underbelly: online harassment, a tool used to silence voices that challenge prevailing narratives or established norms. Female journalists particularly confront extensive harassment regularly due to their profession, a situation that dramatically escalates on digital social media platforms. Our quantitative investigation explores this phenomenon in the context of the Iranian protests of 2022. In this article, we analyze how different subgroups of activists and journalists reporting on the Iranian protests faced harassment on Twitter during and after the protests. The three subgroups are as follows:
(1) Female journalists of the Iranian diaspora
(2) Non-Iranian journalists
(3) Male journalists of the Iranian diaspora.
We explored the experiences of the three groups by analyzing Tweets data mentioning these individuals and discerned the broader narrative surrounding each group. Our investigation revealed that (1) female journalists experienced a novel harassment campaign that was specifically designed to attack their credibility as non-biased journalists; (2) this campaign was primarily driven by bot accounts within the network; (3) threat actors are adapting to changing trust and safety policies and evolving their methodology to avoid being detected by traditional investigative methods.
Context: Background on the Protest and the Role of Female Journalists on the Ground:
Amnesty International reported that Mahsa Amini, an Iranian citizen was first taken into custody by Tehran's morality police on September 13, 2022, for not complying with the nation's dress regulations. [1] Allegations claim that Amini was tortured by the morality police which resulted in her suffering from a coma before passing away on 16th Sept 2022. [1] The protest began on the 17th of September 2022 at the funeral ceremony for Amini and soon spread across Iran. Protesters took to the streets chanting slogans against the Iranian regime and demanding action against the individuals involved. [2]
Throughout the protests, journalists, particularly female Iranian journalists, played an important role in making information related to the protests available to the public at large. In fact, one can argue that female journalists have been at the forefront of this movement since its inception. The story of Mahsa Amini’s torture was exposed by a female Iranian journalist, Nilufar Hamadi. [3] Hamedi shared a picture of Amini’s bruised body and her distressed parents which went viral both nationally and internationally. [4] Additionally, Yalda Moaiery covered the protest from the ground. Her photographs of the violence displayed by the Iranian forces quickly went viral too and provided evidence for the violent suppression of protesters. [4] Both Hamedi and Moaiery were arrested between the 19th and 21st of September 2022. According to the Coalition for Women Journalists (CFWIJ), 25 female journalists were arrested since the beginning of the arrests, of which 12 were arrested within 48 hours of Amini’s death.[5]
Context: Reaction Against Journalists Not Physically Present in Iran:
Female journalists and activists of the Iranian diaspora, who were not physically present in Iran, managed to evade arrest; however, a significant number have reported an uptick in online harassment and threats.[6] The individuals targeted believed they were the focus of a systematic effort intended to discredit their journalistic integrity. This consisted of an organized group on Twitter campaigning for the dismissal of these individuals by their employers, labeling them as regime representatives, spokespersons, and advocates.[7] Notably, the victims of this online campaign were often well-regarded female professionals who had no prior record of biased journalism. The BBC voiced its concerns to the UN about the intimidation faced by its Persian service journalists in light of these events.[8] Furthermore, The Coalition for Women In Journalism highlighted this growing concern through several press statements.[9]
In this article, we will explore how these journalists/activists were targeted and determine if these claims of harassment are substantiated when we look at the data granularly.
Our Sample of Different Subgroups to Explore:
Female Journalists/Activists of the Iranian Diaspora:
Control Groups:
· Time period of Tweets collected: 1st Sept 2022 – 1st Oct 2022
· # of Tweets: 77,226
· # of Agents in the network: 51,992
Analysis of Abusive Tweets Experienced by These Subgroups:
Tweets that are abusive often include offensive language characterized by insults, curse words, and crude language. For our study, we deemed a tweet as abusive towards a journalist or activist if it specifically mentioned an individual from our samples and contained any derogatory terms from our curated list, which encompasses abusive words from 40 different languages. We employed NLP techniques to detect such tweets by matching them with our list of offensive terms.
A few examples of tweets that were considered abusive are as below:
The counts of abusive tweets for each subgroup are as following:
The provided table above indicates that the Iranian female journalists and activists of the diaspora encountered a notably higher percentage of abusive Tweets (19.93% of Tweets were abusive) than the control group of male journalists from the Iranian diaspora (4.74% of Tweets were abusive). This subgroup of female Iranians experienced abuse on Twitter nearly 8.5 times more frequently than male journalists of the diaspora (by absolute numbers).
Additionally, our sample of non-Iranian journalists received a comparable abuse rate at 17.96%. However, the absolute number of abusive tweets was only 122 Tweets for this group compared to 8,97 Tweets experienced by female journalists of the Iranian diaspora. This prompted us to look at the data more granularly to identify if there were specific individuals who were targeted more than others. The results are as follows:
From our female-Iranian dataset, Negar Mortazavi emerged as the primary recipient of online abuse, being cited in 6,526 such Tweets, which accounts for 34% of the mentions she received. Following her, Hoda Katebi faced 1,772 abusive mentions, which represent 8.98% of the Tweets in which her name appeared.
Harassment directed at Negar wasn't limited to social media. It manifested in real life when a talk she was scheduled to give at the University of Chicago was canceled because of bomb threats.
In our dataset of non-Iranian journalists, which also received a similar proportion of abusive tweets (although not by absolute number), Cora Engelbrecht, a reporter with the NY Times, stands out. She received most of the abuse in this subgroup of journalists. Notably, Cora frequently collaborates with Farnaz Fasih, an Iranian diaspora female journalist featured in our test sample. It could be that Cora's association with Farnaz led to the spike in negative Tweets directed at her as well.
We further explored instances of abusive tweets more granularly and investigated what the abuse was directed at. We explored the language in each message and found evidence of association with 6 categories (political, gender, religion, race/nationality, job, violence).
The data shows that for female Iranian individuals, the abuse was targeted to three main categories, namely gender, job, and violence. What this means is that most abusive words for the individual were used in relation to their profession (i.e., journalist), their gender (i.e., being female), and suggested propagation of violence against them. This further supports our initial assumption that harassment of female journalists is deeply rooted and targeted at their gender and choice of profession.
The trend of these abusive tweets for each female of the diaspora is as below:
The chart above shows that Negar Mortazavi remained the predominant focus of abuse. Abusive Tweets spiked for her right after she started reporting news of Amini’s imprisonment. We see that Negar experienced two cycles of abusive tweets, with the first peaking on September 16th and the latter peaking on September 26th. Abusive Tweets for Hoda peaked on September 18th.
It's crucial to highlight that there doesn't appear to be coordination among the abusive tweets targeting different individuals. We infer this because we don't observe simultaneous spikes in abusive tweets for everyone in our sample around the same days. In a coordinated harassment campaign, we'd typically see a surge in tweet counts targeting multiple individuals at roughly the same time – a pattern we didn't find in our data.
A Novel Credibility-Attacking Campaign that Only Targeted Female Journalists of the Iranian Diaspora:
In our investigation, we also found a substantially large number of tweets that did not necessarily use “abusive language” but still harassed these journalists and activists. This set of Tweets tried to create a sentiment that these individuals were not credible professionals, and their message did not represent the current progressive movement in Iran. These Tweets also tried to paint the journalists as representatives of the government in order to urge Western media not to support or promote them. A sample of such tweets is below.
The tweets seemed quite peculiar and caught our attention for three main reasons:
Firstly, they frequently contained very specific derogatory terms like "apologists," "puppets," "mouthpieces," "lobbyists," etc. repeatedly (as shown in the provided visualization). The consistent use of certain terms was a cause for concern, as it often indicates campaigns aimed at crafting a narrative by associating figures with potent words to emphasize certain concepts.
Secondly, although the tweets portrayed the journalists as supporters of the Iranian government, an examination of their work showed the opposite. Most of their articles offered critical insights into the government's actions during the protests.
Lastly, the collective referencing of these authors in tweets was peculiar. While we initially presumed that joint publications might have led to shared criticism, we found no evidence of such collaborative works. This anomaly indicated a potential coordinated online campaign targeting researchers who covered the protest.
An important feature of these tweets was that they did not use any derogatory language. Therefore, since this set of tweets was concerned with maligning the reputation of the journalists, we decided to label this category as credibility-attacking tweets.
The counts for credibility-attacking tweets for each subgroup are as following:
The provided table indicates that the female journalists of the Iranian diaspora clearly and undeniably encountered a much higher rate of credibility-undermining Tweets (16%) compared to both control groups. The difference in the proportion of tweets experienced by each group hints at the female Iranians being the unique target of this campaign. This group faced credibility-related harassment at a rate 11 times higher than the non-Iranian journalists' control set and nearly 9 times more than the male Iranian journalists' sample.
In our data analysis, Hoda Katebi was the most commonly targeted individual with tweets questioning her credibility. From September 15th to September 30th, 4,197 tweets targeted her credibility, accounting for 21.3% of all mentions about her. Next in line was Tara Far, a researcher for Human Rights Watch, who was mentioned in 1,054 derogatory tweets, making up 33.9% of all tweets about her. In terms of proportion, Azadeh Moaveni, a journalism professor at NYU, faced the harshest attacks on her credibility, with a staggering 69% of tweets referencing her containing words that undermined her trustworthiness.
In the group of non-Iranian journalists, Cora Engelbrecht topped the list again, although there were only 5 Tweets. Among male journalists from the Iranian diaspora, Parham Ghobadi was the most targeted, receiving 280 such Tweets, however, these constituted just 2.7% of the total Tweets referencing him.
On average females experienced 895 more Tweets of credibility attacking compared to male reporters. Our results also show that being Iranian was associated with experiencing 833 more Tweets of credibility attacking compared to being non-Iranian.
The graph shown underscores a crucial element of this campaign: the simultaneous surge in these types of tweets targeting all female journalists of Iranian descent in our sample. From the 25th to the 27th, nearly every individual in our sample experienced a simultaneous uptick in such tweets. This was then followed by a sharp decline, with no subsequent resurgence.
This concurrent mention of these authors in tweets was atypical. Contrary to our initial hypothesis that joint publications might lead to collective criticism, we found no evidence of any collaborative work between them. This anomaly hints at a possible coordinated online effort targeting researchers who reported on the protest.
Presence of Bot Accounts in Credibility Attacking Tweets:
The possibility of coordinated inauthentic activity in this category of tweets forced us to investigate the presence of bot accounts in the network as they are commonly associated with such campaigns. We used Beskow and Carley’s Tier 1 bot hunter algorithm to identify the bot probability of each account in our dataset. Tier-1 bot hunter is a random forest model that is trained on a dataset of known bot accounts to make a prediction based on node-level and network-level features of the account in question. If an agent had a bot prediction score higher than 0.7 we labeled it as a possible bot account. The results for bot presence in each group are as follows:
Astonishingly, 3,796 accounts in this network (or 79.3%) were suggested to be bots by our model. These bots were behind 78.3% of all tweets aiming to discredit these female journalists of Iranian descent. In contrast, in the male Iranian diaspora journalist sample, 323 of the accounts in the credibility-undermining tweet network were suggested to be bots as well. These bot accounts were responsible for 77.4% of the tweets that attacked the credibility of these male journalists. For non-Iranian journalists, the bot presence was minimal, with only 2 bot accounts identified, making up 20% of the total tweets in the network.
BEND Analysis of Credibility Attacking Tweets:
Through BEND Framework analysis we were able to divide these credibility-attacking Tweets by the bot accounts into 16 different categories of narrative and network maneuvers. Our BEND analysis model of maneuvers provides a granular understanding of how the accounts were trying to manipulate the conversation that was taking place in the network and the structure of the network itself.
The analysis revealed the following key points about the nature and categorization of tweets undermining the credibility of female Iranian journalists:
Dismiss Tweets: A significant portion of the bot-generated tweets that attacked credibility were labeled as "Dismiss" by our model. These made up 80.4% (equivalent to 4,500 tweets) of all the Tweets. As per the BEND framework, "Dismiss" tactics aim to nullify the message conveyed by the person being targeted. We witnessed a common pattern within this category was responses to the journalist's original tweet, intending to discredit the journalist's perspective by branding them as apologists or spokespeople for the Iranian government
Dismay Tweets: This was the second most prevalent bot-driven narrative maneuver, accounting for 66.27% of tweets. These tweets aim to provoke negative feelings towards the targeted individual by criticizing them as government agents and questioning the sincerity of their work.
Nuke Tweets: At 65.5%, the "Nuke" maneuver was the third most frequent, aiming to alter the community structure around a specific topic. These tweets discouraged readers from regarding the journalists as movement representatives and called out institutions employing them, such as the NY Times and Human Rights Watch, suggesting they should regret hiring these journalists.
So What Do We Learn from This Investigation?
From our thorough analysis of Tweets targeting female journalists of the Iranian diaspora, several critical insights have emerged regarding the landscape of online abuse:
Differential Targeting: Journalists and activists, though commonly facing online threats, are not equally targeted. Our findings highlight that certain subgroups, notably female Iranian diaspora journalists, encounter distinct, credibility-based attacks. This differential pattern suggests that broad solutions may inadequately address specific vulnerabilities inherent to certain communities.
Adaptive Inauthentic Behavior: The orchestrators of these campaigns are evolving, deploying more subtle tactics. By strategically avoiding blatant derogatory language, they attempt to undermine their targets, making detection by conventional trust and safety measures of calculating sentiment by words (severeness of negativity) challenging. This underscores the need for constantly evolving counterstrategies.
Relevance of Advanced Methodologies: The intricate nature of online threats calls for the application of network science and machine learning models. As demonstrated by our use of the Beskow and Carley’s Tier 1 bot hunter algorithm and the BEND Framework, these tools provide nuanced insights into online discourse structures, especially for underrepresented subgroups in Trust and Safety Research. Embracing such methodologies isn't just beneficial; it's crucial.
Sources:
[1] “Iran: Deadly Crackdown on Protests against Mahsa Amini's Death in Custody Needs Urgent Global Action.” Amnesty International, September 23, 2022. https://www.amnesty.org/en/latest/news/2022/09/iran-deadly-crackdown-on-protests-against-mahsa-aminis-death-in-custody-needs-urgent-global-action/.
[2] “Iran: Mahsa Amini's Father Accuses Authorities of a Cover-Up.” BBC News. BBC, September 22, 2022. https://www.bbc.com/news/world-middle-east-62998231.
[3] “In Iran, Crackdowns on Dissent and Journalism.” Columbia Journalism Review. Accessed March 18, 2023. https://www.cjr.org/the_media_today/iran-protests-niloufar-hamedi-mahsa-amini.php.
[4] “Witch Hunt in Iran: Grave Concerns for Journalists Niloofar Hamedi and Elahe Mohammadi.” Center for Human Rights in Iran, November 18, 2022. https://iranhumanrights.org/2022/11/witch-hunt-in-iran-grave-concerns-for-journalists-niloufar-hamedi-and-elahe-mohammadi/.
[5] Khan, Sahar. “Women Journalists Face Increasing Danger in Iran.” Inkstick. Inkstick Media, September 30, 2022. https://inkstickmedia.com/women-journalists-face-increasing-danger-in-iran/.
[6] Khan, Sahar. “Women Journalists Face Increasing Danger in Iran.” Inkstick. Inkstick Media, September 30, 2022. https://inkstickmedia.com/women-journalists-face-increasing-danger-in-iran/.
[7] “Online Harassment against Women Journalists in the Iranian Diaspora.” ARTICLE 19, October 19, 2021. https://www.article19.org/resources/online-harassment-against-women-journalists-in-the-iranian-diaspora/.
[8] Ghobadi, Parham. “Why Reporting on Iran Comes at a Heavy Price.” BBC News. BBC, January 12, 2023. https://www.bbc.com/news/world-middle-east-64222261.
[11] Keskin, Enes. “United States-Iran: CFIWJ Stands in Solidarity with Farnaz Fasihi Who Was Targeted through Vicious Online Trolling.” Coalition For Women in Journalism. Coalition For Women in Journalism, December 8, 2021. https://www.womeninjournalism.org/threats-all/us-iran-cfwij-stands-in-solidarity-with-farnaz-fasihi-who-was-targeted-through-vicious-online-trolling.