Row rages over Facebook ‘bias’

What happens behind the scenes on social media? How do AI filters and human moderators work to keep hate out? Metrolife finds out

Facebook did not flag derogatory posts by BJP politicians as it was keen to be in the Modi government’s good books, according to a Wall Street Journal article. The article has kicked up a storm.

What do social media sites do to keep out hate speech? How do they balance the demands of free speech with the demands for moderation?

Spotting hate

Most social media companies use a combination of artificial intelligence and human moderation to clean up hate speech. What constitutes objectionable content is defined by each site’s exhaustive community standards, publicly available. But that isn’t the be-all and end-all of moderation, says Torsha Sarkar, policy officer at the The Centre for Internet and Society.

“Facebook takes down or moderates content after an outrage around a post. So the general policy seems inconsistent and pretty ad hoc,” she elaborates.

Legally the restrictions to free speech come under Article 19 (2). “This is extremely subjective, and based on what is influencing the judiciary,” says Benson Rajan, social media researcher and assistant professor, Jindal School of Journalism and Communication, Sonipat. He says lack of regulation has led to majoritarian, hate-mongering narrative dominating virtual and other media.

Context matters

Social media platforms have a tough time determining hate speech as their guidelines are more often than not agnostic to geographical terrain.

“What should be considered hate speech flies under the radar as the platform does not invest in resources or capacity in understanding the culture and context in which it exists,” says Torsha.

Benson believes that the contours of cyberspace and the lack of physical territory make it difficult to hold social media platforms liable.

No platform can claim to be content-neutral, says Torsha. “Platforms, their technological infrastructure, community standards, and enforcement, are all done by people. And it would be difficult to remove bias from people,” she says. She worries that a handful of entities are being allowed to determine what is seen by a large majority of the Internet.

FB reaction

On Tuesday, a Facebook spokesperson reacted to the allegations: “We prohibit hate speech and content that incites violence and we enforce these policies globally without regard to anyone’s political position or party affiliation. While we know there is more to do, we’re making progress on enforcement and conducting regular audits of our process to ensure fairness and accuracy.”.

Mark Zuckerberg, CEO of Facebook, had earlier said Facebook was not ‘an arbiter of truth’. Meanwhile Sheryl Sandberg, COO, has said that Facebook would not hesitate to take down posts by Donald Trump if they amount to hate speech or give false information.

However, according to a study by Equality Labs, 93 per cent of all hate speech posts reported to Facebook remain on Facebook.

Free speech vs hate speech

Free speech is free as defined by the Indian Constitution. All objections to this should be settled in faithful accordance with and under this rubric. Hate speech is also free speech, with Constitutionally prescribed constraints, explains Benson Rajan, who teaches journalism. “The press and the public, even politicians and party propaganda, may test the boundaries of this right,” he says, suggesting that only the courts have a right to adjudicate. Speech guidelines assume a preference for self-censorship and accountable autonomy, he adds.

Das auto

Ankhi Das, 49, the Facebook executive in the eye of the storm, has filed a police complaint in Delhi, saying she was facing threats, and five people are engaging in online and offline abuse.

What is hate speech?

Twitter’s policy on ‘Hateful conduct’ says that users may not “promote violence against, threaten, or harass other people on the basis of race, ethnicity, national origin, caste, sexual orientation, gender, gender identity, religious affiliation, age, disability, or serious disease.” Facebook defines hate speech as an attack on the basis of the above characteristics. It defines an attack as “violent or dehumanising speech, statements of inferiority, or calls for exclusion or segregation.” A study by Equality Labs says Facebook’s community guidelines have multiple shortcomings, some of which include the lack of their translation into regional languages, a lack of cultural competency with the moderators. Their reporting mechanisms are geared to the US and lack options to report caste- and religion-based hate speech.

If it’s hate speech, then what?

According to Twitter’s publicly available guidelines, an individual who violates content rules serves a period in read-only mode before they can Tweet again. Subsequent violations will lead to longer read-only periods and may eventually result in permanent account suspension. If an account is engaging primarily in abusive behavior, or is deemed to have shared a violent threat, Twitter can decide to permanently suspend the account. While Twitter outlines its policy on action against hate speech, Facebook does not.

Are they biased?

Internet policy expert Torsha Sarkar says people across the spectrum accuse Facebook and Twitter of bias. Many complain they allow posts that target minorities and women. While the Wall Street Journal article accuses Facebook of a BJP bias, Tejasvi Surya says many have complained the site “is unfairly censoring many nationalist, pro-India or pro-Hindu voices.”

What the law says...

The IT Act and the Intermediary Liability Guidelines govern content on social media. They only talk about ‘takedown’ of content pursuant to government orders, says Internet policy expert Torsha Sarkar. “For every other sort of ‘problematic’ content, the platform is immune from liability, provided certain conditions are fulfilled. So, there are no specific provisions related to content not flagged by the government. The platforms can moderate it such as they deem fit,” she explains.