Discrepancy in implementations of social media election information policies in English and Spanish in-platform searches

2024 U.S. ELECTIONS RAPID RESEARCH BLOG

By Nina Lutz, Mert Can Bayar, Danielle Lee Tomson, and Kate Starbird
University of Washington
Center for an Informed Public

This is part of an ongoing series of rapid research blog posts and rapid research analysis about the 2024 U.S. elections from the University of Washington’s Center for an Informed Public.

Takeaways

• Many social media platforms have policies that address election rumors.
• However, our researchers found a notable disparity in how those policies were implemented when comparing Spanish and English search results for queries related to rumors about election procedures and election integrity.
• This linguistic discrepancy has been observed previously and with other languages. 

Introduction

In the wake of widespread misinformation — and public discourse around that misinformation — about the 2016 U.S. election, several social media platforms, including Twitter (now X), Facebook, and YouTube put into place policies about election-related rumors. 

These policies included provisions that attempt to limit or problematize the spread of election rumors by applying labels, quarantines, downranking, and content removal. Many platforms made commitments to extend these policies to other languages. However, researchers and experts have demonstrated that platform policies on misinformation are often not consistently enforced within non-English communities and content. 

Heading into the 2024 U.S. presidential election, around 24 million voters in the U.S. are expected to rely upon language translations of voting materials while casting their votes. These numbers suggest that language differences in policy enforcement around voting and election rumors will impact millions of U.S. citizens. 

In this blog post, we explore policy discrepancies across language as they surface within in-platform searches related to prominent election rumors.

Background: Election Rumors, Platforms Policies, Search, and Language Discrepancies

Election misinformation can be harmful to democracies. False claims about when and where to vote can disenfranchise voters, and unsubstantiated allegations about widespread voter fraud can diminish trust in election results and democracy more broadly. However, rumors can also provide a signal about real issues with voting and sincere confusions about election processes. 

Social media platforms, which increasingly mediate real-time information exchange, play a critical role in surfacing election-related problems, questions, and concerns, but they also serve as vectors for manipulating voters, delegitimizing elections, and sowing mistrust in our democracy. As such, finding appropriate ways of addressing rapidly spreading rumors on platforms is a matter critical to democracy. 

Platform policies around election-related rumors are nascent and dynamic. Despite repeated updates, they remain imperfect and have been criticized both by journalists and political leaders on both sides of the political spectrum.  Recognizing an ongoing discussion and debate — including an upcoming U.S. Supreme Court decision — regarding the legality, efficacy, and impact of platform interventions addressing misinformation, this blog post does not delve into discussions about what the platform policies should be. 

Instead, the investigation in this blog post is meant to draw attention to the fact that existing policies are not enforced equally across languages. In 2020, our research team’s collaborators observed inconsistencies in the search moderation of non-English language posts with election rumors. In the 2022 midterms, non-English fact-checkers reported seeing disproportionate misinformation in their communities and raised concerns about election falsehoods.  Here, we look at in-platform searches in Spanish versus English to further interrogate this alleged discrepancy in the lead up to the 2024 U.S. general election.

Why in-app searches? With more users turning to in-app searches on social media platforms, there is a growing concern about misinformation in search results, in part due to the lack of enforcement of existing policies. 

In this rapid response post, we examine search results and policy implementations for queries related to election integrity rumors and investigate differences between English and Spanish experiences. We compare the search results of three social media platforms for both English and Spanish queries. 

This study is not peer-reviewed and does not aim to replace one. Nor is it attempting to assess the efficacy of a given policy. Instead, it demonstrates a phenomenon of inconsistent policy implementation we observed in real-time. We hope this study will raise awareness among platforms, election officials, fact-checkers, citizens, and other researchers.

Method and Analysis

A substantial amount of misinformation around the 2020 election took the form of false, misleading, or unsubstantiated allegations of “election fraud.” With this election season underway, our team set out to compare the platform policies of Spanish and English specifically around situations where users are searching terms that are or could be associated with misleading claims of election fraud. 

It is important to note that there are instances where voting irregularities do reflect fraud, both in the U.S. context — for example, a recent mayoral election in Connecticut — and around the world. However, there are also, increasingly, examples of wholly unsubstantiated claims of fraud as well as cases where small but incidental and inconsequential issues with the voting process are misinterpreted or mischaracterized to sow distrust in the election procedures and results. Given the pace, nature, and dynamics of election rumors, there is an ongoing debate for how platforms should handle real-time election-related information, ranging from educating users or slowing down the release of information to users of platforms. 

For our study, we first reviewed the policies on TikTok and Meta (the parent company of Instagram and Facebook, which both use the same search policies) regarding search queries related to voting, election administration, and election integrity. We summarize them below:

Platform	Search Policies	Platform Guidelines Around Election Misinformation Pertaining to English and Spanish Languages
TikTok	Searches for queries related to election misinformation will be redirected to authoritative sources, and informational banners will be shown.	English Election Policy as of 3.31.24 Spanish Election Policy as of 3.31.24
Meta (Instagram and Facebook)	When searching for terms “related to the 2024 elections”, users will see official information about how to vote, powered by the Voting Information Center.	Overall Preparing For Elections at Meta Specific 2024 sheet as of 3.30.24

Next, using a set of similar voting-related queries in Spanish and English, on TikTok, Instagram, and Facebook, we analyzed the results to assess the implementation of existing policies on those platforms. We initially conducted this analysis on March 30 and 31, 2024 and then reran it on May 28, 2024 to confirm that the results are still accurate. We used the personal machines and in-platform accounts of members of our Hispanic and Latine research team, who regularly receive content in English and Spanish. These individuals collaborated with the National Hispanic Media Coalition to design queries about election fraud to observe how these policies were enforced in Spanish and English. Results were done on devices set to both Spanish and English.

From these searches, we observed discrepancies in enforcing policies between English and Spanish versions, particularly in how the content is displayed in search results and classified as potentially false information. In English, search results for certain terms may be restricted, with users either being unable to access the content or redirected to accurate election information.

We have marked the platforms with a ✅ if we observed the interventions described in the table above from searching these queries, such as labeling, adding links to authoritative content, and redirection. We denoted with an ❌ if no intervention was apparent.

Search Terms: English	TikTok	Facebook	Instagram	Search Terms: Spanish	TikTok	Facebook	Instagram
Election fraud	✅	✅	✅	Fraude electoral	❌	✅	❌
Election stolen	✅	✅	✅	Elecciones robadas	❌	❌	❌
Rigged election	✅	✅	✅	Elección amañada	❌	❌	❌
US election fraud	✅	❌	❌	fraude electoral  Estados Unidos*	❌	❌	❌

_{*  After discussing this term with our community partner, the National Hispanic Media Coalition, we also looked it up with EEUU, America, and América. For all variations, we did not see interventions.}

Below is an example of results for “election fraud” where users are either unable to access the information on TikTok or information is obscured with a mandatory banner above any results on Meta platforms (Facebook and Instagram). This implementation matches the above-linked policies for these platforms.

 

 

This is not the case in Spanish. Searching for “fraude electoral” returns results on TikTok and does not trigger the voting information banner on Instagram, showing a discrepancy between the search terms in English and Spanish. This trend continues with other terms, revealing the policies to be consistently enforced in English for queries like “Election stolen” or “Rigged election,” but not in Spanish.

 

The results returned for the Spanish language searches may not all be false claims of election fraud, and likely include factual reports of election concerns and even corrections of false claims of fraud. It is worth noting the above screenshot features content made regarding the 2024 Mexican election, given the timing of this work. However, Spanish language users in the United States were able to retrieve this information without the interventions that contextualize this type of information the way that English language users received.

Conclusion and Recommendations 

Election rumoring is a multifaceted issue that cannot be addressed solely through platform interventions — it is a social phenomenon that needs more than just technical solutions. There is an ongoing debate on how platforms should handle searches for content related to election fraud, especially given the possibility that some allegations of voter or election fraud may turn out to be accurate. Regardless, as more and more U.S. voters rely on non-English sources and social media for political news and as we head into the 2024 U.S. general elections, it is important to have consistent policies across languages.

Election fraud is a real problem, with many countries having experienced it, and trust in elections vary in different global and diasporic contexts. In the U.S. context, platforms should be mindful of the harms of previous election integrity misinformation in 2020 and how non-English speaking communities may be particularly vulnerable to election misinformation campaigns about the U.S. 2024 Presidential Elections. Even as discussion continues on the legality and responsibility of moderating searches, it is crucial to have fair, consistent policies and interventions that consider non-English language users and fact-checking entities.

For those who want to learn more about misinformation in Hispanic communities in the United States, we recommend the following resources from organizations in this space:

• Metodología · Factchequeado.com 
• NYU’s Center for Social Media and Politics 
• Voto Latino 
• Countering Disinformation | Wearemasfl 
• Digital Democracy Institute of the Americas (DDIA) 
• Equis – EquisLabs 501(c)(4) 
• National Hispanic Media Coalition 
• Spanish Language Media | Media Matters 

Acknowledgements

At the Center for an Informed Public (CIP), we have collaborated with various community and non-English fact checking organizations, including the National Hispanic Media Coalition, Equis Labs, Factchequeado, Asian American Disinformation Table, and Viet Fact Check, who have also observed this phenomenon in U.S. elections. Researchers Sarah Nguyễn and Rachel Moran have written about Vietnamese diasporic case studies highlighting the unique mechanics and extent of information disorder in non-English speaking diasporas and the need for platforms and researchers to investigate non-English cases of mis- and disinformation.  We want to thank them for their recommendations and help with this work and their former work in this space! Thank you to Osiris Cruz-Antonio for your review of queries!

Thank you to Randy Abreu from the National Hispanic Media Coalition for chatting through this with us! 

---

• Nina Lutz is a University of Washington Center for an Informed Public graduate research assistant and doctoral student in the UW Department Human Centered Design & Engineering. 
• Mert Can Bayar is a CIP postdoctoral scholar.
• Danielle Lee Tomson is a CIP research project manager.
• CIP director and co-founder Kate Starbird is a UW HCDE professor.