Misinformation online is bad in English. But it’s far worse in Spanish. WaPo
molly...@gmail.com
The release of internal Facebook documents showing that the platform isn’t doing enough to stop a flood of lies and misinformation has sparked outrage nationwide. As bad as these problems are in English, though, they are even worse in other languages: Facebook has admitted its platform was used to incite violence against the Rohingya in Myanmar, and in the Philippines, the site helped fuel a vicious drug war and attacks on dissident journalists. Social media platforms are allowing far more misinformation to spread in other languages than they are in English.
But some of the scariest misinformation online is spreading right here in the United States — in Spanish.
I worked at Google from 2015 to 2018, and I saw the power the Internet has to foster community, keep family connected and help small businesses. It can even fuel social movements — from Cuba’s most recent protests to police accountability. Yet I also saw firsthand how many of the platforms use the shiny possibility of the Internet as a shield to hide the depths of what happens on the dark side. We are living with the consequences of years of inaction, which have yielded a mass shooting of Latinos in El Paso, a literal insurrection and deaths from anti-vaccine misinformation.
Latino communities maintain strong connections across Latin America; the result is an entire continent of Spanish-language misinformation largely unchecked by the platforms. Latinos are more susceptible to misinformation simply because of how much time we are spending online — twice as much on YouTube as non-Latino adults, for example, according to the latest research by Equis, the organization where I work, which is dedicated to researching the Latino electorate. Two-thirds of Latinos treat YouTube as a primary source for their news and information about politics and elections. Half of Latinos in the United States use WhatsApp, the Facebook-owned messaging platform, more than any other ethnic or racial group in the country.
Spanish-language misinformation narratives often start on Facebook or YouTube, but then conversations or viral content move to closed WhatsApp groups where there’s less of a chance for fact-checkers to intervene. Even with all the misinformation spreading on WhatsApp, Facebook founder and chief executive Mark Zuckerberg still opposed having a “voting information center” for Spanish speakers on the platform ahead of the 2020 election because he thought it was not “politically neutral,” according to The Washington Post. (Facebook spokesman Andy Stone said this week that “this is false” and that WhatsApp had launched bilingual campaigns last year about voter registration information and fact-checking. The company did not answer a Washington Post editor’s request for comment for this story.)
Many Spanish-language social media pages and groups are cesspools, enabling smugglers to target desperate migrants and refugees and spreading harmful covid-19 and vaccine misinformation as fast as your tia’s “Dios Te Bendiga” meme. These tech platforms don’t just spread racist hate speech targeting Latinos; they’re also frequently spreading racial tropes that perpetuate colorism and anti-blackness, which help drive a wedge between Latino and Black communities.
Our research shows that social media networks are doing a poor job of addressing Spanish misinformation, with less moderation and harmful posts left up longer than in English. Facebook still has Spanish-language posts active today from November 2020 that promote election lies with no warning labels. Facebook and YouTube both announced policies to remove or restrict QAnon content, but it continued to spread in Spanish. The platforms allowed content to stay up for weeks until we flagged it for them — and they still refused to take some down.
Facebook pages we were tracking last year spread the lie that dead people voted in Nevada in the 2020 election, a claim Facebook’s fact-checking partners rated false multiple times. The pages posting the claim in English had the posts labeled as “false information.” One Spanish post with hundreds of shares still has no label.
More recently, we’ve seen that Facebook will flag vaccine misinformation content in English, but the same content in Spanish takes days to get flagged, if it ever does. The online activist group Avaaz found Facebook failed to issue warning labels on 70 percent of misinformation in Spanish, compared to only 29 percent in English. It isn’t just Spanish, either: In whistleblower Frances Haugen’s testimony before Congress, she revealed that 87 percent of misinformation spending is on English, but only about 9 percent of the users are English speakers. Haugen exposed how the company’s profit incentives prompt it not to offer the same safety systems for every language on the platform or every country where Facebook is used: “Every time Facebook expands to a new one of these linguistic areas, it costs just as much, if not more, to make the safety systems for that language as it did to make English or French,” she told “60 Minutes.” “Because each new language costs more money but there’s fewer and fewer customers. And so, the economics just doesn’t make sense for Facebook to be safe in a lot of these parts of the world."
When a soccer player collapsed in cardiac arrest during an international match this year, anti-vaccine social media accounts jumped on it, falsely claiming the incident was related to the coronavirus vaccine (the player hadn’t even been vaccinated, his coach said). Several Facebook pages we found sharing the lie in English were almost immediately labeled as false. But the exact same lie posted on a prominent disinformation account in Spanish was left up for days — an endless amount of time for disinformation — receiving hundreds of shares before being labeled.
Late last year, Facebook said it was banning content promoting false claims that the vaccine contained a microchip. Since then, we have seen Facebook label several posts in English promoting this narrative as false, but similar posts we have tracked in Spanish still have no label.
The platforms don’t handle false information in English and in other languages in the same way. Social media companies haven’t just underestimated what they were up against. They have decided to let their algorithms promote hate and misinformation because division leads to more clicks, traffic and profit. At the bare minimum, the platforms need to make the investment to take the same actions on Spanish content as they do for English. They have consciously chosen not to direct resources to address these problems. Just look at how effectively YouTube identifies and quickly removes content for copyright infringements to protect profits.
Advocates have been pushing for a set of solutions that the platforms can take — including hiring a C-suite position to oversee Spanish-language content moderation, expanding Spanish language moderation capacity and being more transparent about their moderation systems and processes — but with little to no success. And Facebook and the other platforms have repeatedly shown that they won’t solve this problem. The lack of self correction by the platforms has many advocates now calling on Congress and the Biden administration to not only demand answers but regulate the platforms or, in the case of Facebook and WhatsApp, break them up. A Senate hearing on Spanish-language disinformation specifically is likely.
Social media companies hide behind meaningless marketing terms like “connecting” and “community” because their only real goal is higher profits. Lies, hate and even insurrection and deaths are not accidental byproducts of the way these platforms operate — they are the growth strategy. These massive companies have proven they have the ability to disincentivize hate, but they choose instead to profit from stoking fear, anger, teen depression, even violence. And for all the harm Facebook and other companies allow to flourish in English, their handling of Spanish content has been even worse.