In a press interview, former Vermont state house representative Kiah Morris said she reported at least 26 incidents to the local police where she and her family felt threatened between 2016 and 2018. The severity of the targeted abuse both on and offline ultimately led Rep. Morris to a premature resignation three months before the end of her second term in office in 2018. In the same story, U.S. Rep. Nikema Williams expressed the view that the onslaught of mis- and disinformation and abuse seemed to be to designed to intimidate women of color out of government: “Early on, when we were getting the list of credible threats coming in for members of Congress, they were centered around members of color and there are only 25 black women that serve in the United States…there’s not that many of us…which I think is part of the whole thing of people trying to scare people [black women] into silence.”
As more women[1] have sought political representation by running for elected office, we have seen demonstrated increases in online harassment and abuse, including targeted mis- and disinformation campaigns. Civil society and human rights groups have been at the forefront of documenting and characterizing these forms of “gendered disinformation” and online gender-based violence (GBV). Researchers argue that these attacks are attempts to limit women’s ability to participate in electoral politics and suppress their voices in a variety of settings.
While women in general may be subject to significant mis- and disinformation and abuse online, an intersectional approach, which recognizes that women of color have to contend with multiple sources of oppression at the same time and that this impact is unique, can better illuminate the additional challenges faced by women of color in general, and women of color political candidates specifically. By not sufficiently examining the intersectionality of women of color’s experiences, we may miss the point of some mis- and disinformation campaigns, which are intentionally designed to exploit existing forms of discrimination by targeting people based on both race and gender identity.
In this report, we use the phrase “mis- and disinformation” to encompass three categories of information that researchers have developed: false information shared without the intent to cause harm (misinformation); false information shared with the intent to cause harm, often for some political, social, or other goal (disinformation); and accurate information shared in a misleading context (malinformation). As we do not attempt to determine the intent of the person sharing information online in this research, we decided to use the phrase “mis- and disinformation” to refer to both categories of false information regardless of intent.
We also examine online abuse in this report. More specifically, in our study, we examined 15 types of abusive content (see Appendix C) including categories such as sexism or misogyny, doxing, threats of violence, racism, the use of offensive language, and attempts at demeaning the person. A subset of these categories constitutes a form of abuse that is referred to as online GBV – harmful acts directed at an individual because of their gender. GBV results from differences in access to power between people with different gender identities. Given the societal roles and powers ascribed to men in many societies, most experiences of GBV around the world are directed against women and girls.
As Sinders argued, the harms that stem from mis- and disinformation and online GBV are similar and include perpetuating bias and falsehoods, psychological abuse, and real world impacts. However, she also notes that, with a few exceptions many researchers and most policymakers and social media platforms overlook these similarities, in particular the coordination around the creation and distribution of harmful content. Previous work at CDT examined the problem of election related mis- and disinformation. Building on this research and, after consultations with partners and assessing scholarship on the problems of race, gender, and mis- and disinformation, we identified two key research questions:
Are women of color political candidates more likely to be subject to mis- and disinformation and online abuse compared to other types of candidates?
What are the impacts of mis- and disinformation and online abuse of women of color political candidates?
We focus on both mis- and disinformation and online abuse because they are part of the larger problem of violence against women in politics and both are often aimed at undermining the political efficacy of women in public spaces.
These questions are particularly important given that women of color face significant barriers of entry into representative politics (only 10% of candidates that ran for Congress in 2020). Some researchers suggest that what the candidates face online, including online abuse, could contribute to this. A lack of representation of a significant part of our population among our elected officials is extremely problematic for our democracy.
This research began in late 2021 and, given the scale of the work, took place over several months. Our focus was on the 2020 Congressional elections in the U.S. We chose not to examine state and local elections to maintain a feasible project and so only make inferences from our findings based on candidates in Congressional elections. We also note that campaigns during the 2020 elections took place during the Covid-19 pandemic (2020) and relied heavily on these social media platforms to operate.
We present our research results in two parts. In Part I, we conducted a content analysis of over 100,000 posts on Twitter during the 2020 election that were targeted at or were about a random selection of candidates that ran for Congress. Our objective was to compare the levels of mis- and disinformation and abuse targeted at or about different groups of candidates.
We found that during the 2020 U.S. Congressional election:
Women of color candidates were twice as likely as other candidates to be targeted with or the subject of mis- and disinformation.
Although women of color candidates are not the most likely target of online abuse overall—white men are—they are the most likely to be the target of particular forms of online abuse, including sexist abuse (as compared to white women), racist abuse (as compared to men of color), and violent abuse (four times more than white candidates and two times more than men of color.)
Women of color candidates were the most likely to be targeted with or the subject of posts that combined mis- and disinformation and abuse.
Women of color candidates were at least five times more likely than other candidates to be targeted with tweets related to their identity that focused specifically on their gender and race.
Women of color candidates were less likely to be the subject of tweets that were positive compared to white women candidates, but more likely to be the subject of positive tweets compared to white men candidates.
Of note throughout all our findings is the specific targeting of women of African descent or African American women. This group of candidates was subject to the highest levels of mis- and disinformation, certain forms of abuse, and tweets with both mis- and disinformation and abuse compared to other women of color and most other candidates. Indeed, they were also more likely to receive tweets where the main focus was their gender and racial identities.
In Part II, we conducted a series of interviews with women of color candidates that ran for Congress in 2020 and their teams. These interviews were conducted on the basis of informed consent and anonymity. In all, we interviewed 13 women candidates and 7 staffers (from a total of 14 different campaigns). From the interviewees we learned that:
In their view, the aim of the people behind the mis- and disinformation and abusive attacks was to destroy the candidates’ resolve. They believed the purpose of the attacks was to get them to internalize the abuse directed toward them, to accept the oppression they face as women of color, and to drop out of politics.
The mis- and disinformation and abuse they encountered was not only about challenging their electoral prospects by damaging their reputations with voters, but also about attacking them based on their identity as women of color.
Identity-based online GBV targeted at women of color candidates focused on the transgressiveness of running for office (i.e. a woman seeking power, as someone presumed unworthy or unsuited for power or authority).
In many cases the attacks were focused on the fact that the candidates identified as women, and were often intensified by referring to the candidates’ other identities or attributes, such as their race, age, marital and parental status.
While many of the attacks interviewees described were severe, we also learned about the degree of resolve and coping that the participants displayed as they encountered these attacks. Together with their campaign teams and a broader community of support, some candidates employed several resilience and coping strategies and, in most cases, they continued their campaigns through the election. Many remain in representative politics.
Recommendations
Based on our findings, we make the following recommendations for social media platforms, other political candidates (particularly women of color), their parties, and researchers working in the field.
To combat abuse and mis- and disinformation targeted at women of color political candidates, social media companies should:
Clearly articulate policies that prohibit content that harasses or abuses someone on the basis of gender or race.
Offer training for political campaigns on how to use their platforms and specifically on tools that are available to users to address online abuse and mis- and disinformation.
Publicly provide information about how they consider gender and race in their policies and enforcement processes against mis- and disinformation and abuse.
Provide publicly available transparency reports around election mis- and disinformation and abuse before, during, and after an election.
Make data available to independent researchers that enables them to study the impact of mis- and disinformation and online abuse, including GBV, on political candidates.
Take additional steps to protect and prevent abuse and mis- and disinformation from reaching women of color candidates. They should:
Conduct risk assessments of their ranking and recommendation systems to evaluate their impact on women of color candidates and what abuse mitigation measures the service provider can implement.
Offer tools that allow users to report content that violates the companies’ policies against abuse or mis- and disinformation and to control who can interact with their accounts.
Invest additional resources into enforcement of content policies prohibiting abuse and mis- and disinformation in the run up to and after elections, including a necessary increase in responding to appeals.
Ensure that content moderation systems, including human moderators and algorithmic systems, are attuned to the needs of and the threats faced by women of color political candidates, in particular.
Understand that not all candidates require the same type or degree of support from a social media service provider to address these problems.
Scrutinize the role of political advertising in spreading mis- and disinformation and abuse on their services.
Campaigns, political organizations, and other initiatives supporting candidates should:
Offer free or low-cost campaign training designed to prepare women of color candidates for the social media landscape.
Create additional toolkits to inform candidates of digital security best practices. Existing toolkits should be better promoted to address the needs of women of color candidates.
Researchers should:
Pursue research analyzing the problem of online abuse and mis- and disinformation with an intersectional lens; we hope that this report can serve as a point of reference for future research.
Repeat this or a similar study for the 2022 U.S. elections and use a longer period (i.e., more than two months) for data collection from Twitter.
Expand research to other platforms, especially Facebook.
Focus on abuse or mis- and disinformation from political candidates targeting women of color candidates.
Examine posts where women of color candidates are not explicitly tagged or named but still referred to in other ways (e.g., by another name).
[1] In this report we focus on people who identified as women. We do not address the experiences of people who identify as non-binary, although those are also important.
Not All Guardrails Are Created Equal: Comparing Content Safety and Copyright Filtering
As courts and policymakers work through questions about chatbot liability, they should be wary of analogies that flatten meaningful technical differences. Copyright filtering and safety intervention share real challenges around ambiguity and evasion, but they diverge in what each control must assess, how each manifests over the course of a conversation, and how much can be verified from the outside.
Op-Ed: France’s Constitutional Council Ruling Tests the Limits of Social Media Bans
Read our analysis the legal and policy significance of the Constitutional Council’s decision on ongoing social media age restriction and age assurance debates ahead of the most-anticipated EU-wide approach on child online safety due by December 2026.
The brief explains that Section 230’s liability protections are essential to enable free expression online and they extend to the use of automated systems to engage to rank and order content as part of traditional publishing activities.
