Recently, there have been many spelling mistakes in the tweets of Twitter users debating controversial subjects like religion, terrorism, crime, and yes, even Indian history. These are not errors made in the heat of the moment, but careful distortions meant to keep the tweet from being flagged or deleted for hateful content. For example, a tweet posted in November 2022 that singled out people with Muslim names from a series of random crime news reports did not refer to the perpetrators as “Muslims” but instead used the phrase “Ola ke Bande” (Ola’s group/Ola’s gang). The word “Ola,” which refers to the Indian ride-hailing service, was used in place of “Allah.” The tweet was posted after CEO Elon Musk’s $44 billion takeover of Twitter. The tweet author currently has close to 7,00,000 followers.
Reporting this tweet for targeting a group of people based on their religion is now more difficult than reporting a hateful tweet simply referring to “Muslims,” as the Twitter moderator viewing the complaint would have to be familiar with not only the Hindi phrase being used, but also the double meaning of “Ola.” Moreover, if the moderation process is automated, the machine would most likely see a user verbally abusing a taxi service, which does not constitute hateful conduct.
Distortions of hate
Some of the common hate speech distortions are — ‘beeph eaters’: a derogatory term used to refer to Muslims, by distorting the word ‘beef’; ‘muzlim’: a distortion of the word ‘Muslim’; ‘Is!am’: a distortion of ‘Islam,’ which fools Twitter’s search filter by bringing up results for “isam” or “is am”; ‘peaceful/peacefuls’: a word used in place of ‘Muslim,’ to avoid being flagged for hate speech; and ‘rice bag’: a derogatory phrase referring to India’s religious minorities, referencing the bigoted stereotype that they left Hinduism in exchange for a “rice bag” or economic benefits.
A lack of moderation
There are fears that Twitter no longer has the resources or the staff it needs to adequately combat hateful conduct, spam, criminal content on the site, and misinformation in not just English but global languages. In an interview with a BBC journalist on April 12, Mr. Musk shared that there were around 1,500 employees still at Twitter, down from slightly under 8,000. The months after his takeover saw employees being fired in multiple rounds, as well as mass resignations. Fired employees included those involved in content moderation worldwide, reported Bloomberg.
The Center for Countering Digital Hate (NGO) said on February 9 that “Elon Musk has reinstated tens of thousands of accounts, including neo-Nazis, white supremacists, misogynists and spreaders of dangerous conspiracy theories.” It added that Twitter could rake in up to $19 million a year from advertising on just ten such reinstated accounts, which included figures such as Andrew Tate and Robert Malone. Mr. Musk denied that hate speech was rising on Twitter and claimed that the BBC journalist was lying to suggest otherwise.
However, it is a mistake to assume that content moderation on Twitter before Mr. Musk’s takeover was effective. Former CEO Jack Dorsey has admitted to making “mistakes.” Yet, the key difference now is that Twitter does not even pretend to let users hold it accountable, as its press@twitter.com email ID automatically responds to all media queries and emails with a graphic emoji.
Twitter’s safety policy
Furthermore, on April 17, the Twitter Safety team announced its new approach to content moderation, called “Freedom of Speech, Not Reach.”
This policy would leave up select tweets that violate Twitter’s Hateful Conduct Policy but filter their visibility by adding a label flagging them as potential violations. Ads will also not be placed near them. Twitter’s Hateful Conduct policy says that users cannot “directly attack” others based on “race, ethnicity, national origin, caste, sexual orientation, gender, gender identity, religious affiliation, age, disability, or serious disease.” However, the new “freedom of speech, not reach” policy would mean that some tweets violating these principles will be allowed to remain on Twitter.
Hate speech on other platforms
The Hindu reached out to Facebook, Instagram, and WhatsApp-owner Meta to learn the rival social media company’s approach to hate speech.
In an emailed statement, a Meta spokesperson said the company takes a zero-tolerance approach to hate speech on Facebook and Instagram, and had pulled down such content as soon as they were made aware of it. “Since 2016, we have invested more than $16 billion in teams and technologies to keep hate speech, and other forms of harmful content off the platform and to enhance safety and security of users on our apps. As a result of our efforts, we’ve reduced the prevalence of hate speech on our platforms down to .02% which means for every 10,000 pieces of content, two were violating,” said the spokesperson. Meta explained it used artificial intelligence (AI) to identify images or text identical to hate speech content it had already removed. This technology helps it accurately detect hate speech even when the meaning is not obvious or the content is changed to avoid detection, the company said.
Additionally, Meta said it had human content reviewers in 20 Indian languages including Hindi, Bengali, Tamil, Malayalam, Punjabi, Urdu, Kannada, Marathi, Gujarati, Assamese, Telugu, Oriya, Sindhi, Mizo, Marwari, Chhattisgarhi, Tulu, Maithili/Bhojpuri, Konkani, and Meitei. The social media company also claimed it took down around 16 million pieces of content from Facebook and Instagram in the last quarter of 2022.
However, as Meta confirmed yet another round of layoffs in late April that impacted its technical employees, reduced staff could translate to less effective content moderation on Facebook and Instagram. Tech outlet The Verge further reported that most of Meta’s misinformation engineering team was fired as part of April’s cuts. Meta is due to fire more workers in late May and aims to reduce its workforce by around 10,000 people in 2023.
The Hindu also reached out to Mastodon and Reddit to learn their approaches to content moderation but did not receive a response.
As Twitter’s roughly 1,500 employees put in the hours to take the social media company forward, hateful content is finding new ways to evolve on the platform.
COMMents
SHARE