Elon Musk owned X’s crowdsourced fact-checking program Community Notes is struggling to limit misinformation on its platform in India’s highly polarised election where political rhetoric has turned into anti-Muslim hate speech.
Community Notes was rolled out to users in India on April 4, 2024, a fortnight ahead of the start of voting in a country known for having an acute misinformation problem.
The rollout in India, as expected, was bumpy at the outset.
The feature allows ‘contributors’ to write notes for tweets which they think are misleading or missing important context. Individuals with an X account can sign up to be a contributor upon meeting the eligibility criteria.
Previously named Birdwatch, the program piloted in the United States in January 2021 before Musk took ownership of the platform in 2022.
“It’s crowdsourced, which is a cool idea. I love the idea. But the idea that you could have this completely automated algorithm driven way of surfacing quote, unquote “truth” on a platform where truth is very hard to come by is just nonsensical,” Alex Mahadevan, Director of Mediawise at Poynter, told BOOM.
“My main issue with Community Notes is that you just cannot completely automate fact-checking, which is essentially what they are trying to do,” Mahadevan added.
The program uses a bridging algorithm to decide whether a note is shown or not. The data is public and can be downloaded and studied to analyse trends and limitations to improve the program.
“Community Notes doesn't work like many engagement-based ranking systems, where popular content gains the most visibility and people can coordinate to mass upvote or downvote content they don't like or agree with,” X’s guidelines state.
“Instead, Community Notes uses a bridging algorithm — for a note to be shown on a post, it needs to be found helpful by people who have tended to disagree in their past ratings,” it further states.
Therefore for a note to be shown on X, reviewers of diverse viewpoints need to agree that it is helpful.
And that has been the crux of the challenge in India where Community Notes contributors cannot seem to agree on what constitutes a fact, an opinion, or misinformation.
“My hypothesis is that in the more partisan polarised countries there’s a lower chance of notes becoming public and therefore a lower impact of a program like Community Notes,” Mediawise’s Alex Mahadevan said.
“I think that rolling it out in India ahead of the election is just a way for Twitter or X to look like they are doing something good and I just don’t know think it will have too much of an impact,” he said.
Ankita Deshkar, Fact Checker and Deputy copy editor, at The Indian Express, who is also a Community Notes contributor, said the feature has been helpful at countering misinformation but there has been misuse.
“I think Community Notes has been very helpful, even though less fact checkers have access to it, there are some vigilantes who really want to put the truth out and they are contributing to Community Notes. But on the other hand I see that there has also been a misuse of this feature,” Deshkar told BOOM.
“People literally label anything that they don't agree with and it goes up on the notes for rating. People fail to get it that Community Notes should be added only if the tweet is factually incorrect and it has nothing to do with the opinions of the people,” she added.
Contributors Disagreed On Posts That Required Nuance
Fact-checkers in India expected a surge in misinformation over the country’s six week long general election. And while the ongoing polls were unofficially dubbed as ‘India’s AI elections’, non-artificial intelligence based disinformation and particularly hate speech has complicated fact-checking efforts.
This reporter got limited access to Community Notes. Those who apply to join Community Notes can only start writing notes after rating a certain number of notes by other contributors as either helpful or not helpful.
Contributors mostly bickered over posts about Prime Minister Narendra Modi’s scathing attacks on the opposition Congress party and its leader Rahul Gandhi’s remarks about his manifesto promise of conducting a caste survey and plans for redistribution of wealth.
At a rally in Rajasthan in April, Modi attacked Congress’ manifesto and claimed former Prime Minister Manmohan Singh said Muslims had the first claim on the country’s resources.
A tweet by the Bharatiya Janata Party’s official handle posting a snippet of Singh’s speech in support of the claim was annotated with several Community Notes either supporting or disputing Modi’s remarks.
The claim made headlines, divided opinion and even split consensus among Indian fact-checkers. BOOM Live found that Modi misquoted Singh’s remarks.
When Mohammed Zubair, co-founder of fact-checking website Alt News, tweeted about the same speech and slammed the PM for using “infiltrators” and “those who produce more children” as synonyms for the Muslim community; several contributors wrote notes on his tweet either in support of or disputing Zubair’s post and accusing him of peddling misinformation.
We found that in many instances notes were not displayed on outrightly false tweets even though there were notes written debunking the misleading claims in the tweets. The notes did not receive enough ‘helpful’ ratings and thus were not displayed.
A tweet from a verified X user peddling a fake letter claiming to show Dubai-based Association of Sunni Muslims offering financial support to Muslims travelling to Karnataka to oust the incumbent Bharatiya Janata Party (BJP) and Prime Minister Modi, did not have any note on it. The fake letter was debunked by fact-checkers.
Similarly a photoshopped image purporting to show a person in a crowd flashing their middle finger to Modi, did not see any note attached to it on X despite the notes being rated as helpful. The note was visible to only a select test group of contributors and not visible for other X users.
Misleading political content that was altered through traditional or old school editing techniques did not have notes on them either or if they did they were not applied consistently to other posts that shared the original tweet.
Megh Updates, a verified X user with a history of posting pro-right misinformation, posted a cropped video of Rahul Gandhi with a false claim that the Congress leader promised to credit an annual sum of 1 lakh rupees to accounts of youth, who spent time browsing Instagram and Facebook.
The original tweet by the social media head of BJP Uttar Pradesh’s youth wing, included a Community Note that stated the video was presented out of context.
But Megh Updates’ tweet, which has been viewed over 1.4 million times and reshared over 5K times, had no such note for lack of helpful ratings on notes that fact-checked the tweet.
A lot of times by the time a note appeared on X, the misinformation had spread to other platforms in India.
Most social media platforms design systems to tackle misinformation on their platforms only and do not take into account how it travels to other platforms or messaging apps.
Community Notes Appeared On Old, Manipulated Content
Community Notes were visible on certain types of content such as old videos passed off as recent. These included videos of voting irregularities at polling stations, old videos showing a ‘strong room’ where electronic voting machines (EVMs) are stored and, old videos of violence falsely linking them to the ongoing election.
Community Notes also appeared on tweets featuring manipulated content or videos doctored with generative artificial intelligence (AI).
Notes appeared on tweets of a few Congress social media functionaries who posted videos using AI voice clones of actor Aamir Khan and Ranveer Singh to target Modi and the BJP.
X User Base In India Small But Sets Political Discourse
X had about 25 million users in India behind the United States (106.23 million) and Japan (69 million) as of April 2024, according to Statista.
The user base is negligible in comparison to WhatsApp, Facebook and YouTube users individually, in India.
However, the micro-blogging platform is hugely influential in setting the political discourse and increasingly where political misinformation originates.
Prime Minister Narendra Modi, who has a whopping 97.7 million followers on X currently, has often used the platform to bypass journalists and reach his followers directly.
The site is also a public battleground for the BJP and its infamous IT cell, after WhatsApp which is end-to-end encrypted and where its network is well entrenched over several years.
The Indian right-wing, like its US counterpart, alleged for a long time that Twitter, under co-founder Jack Dorsey, censored conservative accounts. They threatened to abandon it for an Indian copycat app called Koo.
The Modi government too kept threatening to scrap Twitter’s safe harbour protection from liability under Indian law, from time to time.
But all talk of abandoning Twitter for Koo and government censure died down when Elon Musk took over the platform.
Musk’s decision to cull Twitter’s trust and safety teams and allow anyone to buy a verified badge have led to a deluge of misinformation on the platform that was once an imperfect but useful resource for journalists but now is a place to find false claims and conspiracy theories.
Under Musk, X has been criticised of democratic backsliding by censoring accounts of protesting Indian farmers, critics, and complying with legal requests from New Delhi.
The platform told Delhi Police that a doctored video of Home Minister and BJP leader Amit Shah which appeared to show him promising to end reservations for marginalised castes and classes in India, was first posted from an IP address in Telangana.
The doctored video was also tweeted by Telangana Congress and Jharkhand Congress’ X handles. X withheld Jharkhand Congress’ account following a ‘legal demand’.
While the prime minister has referred to the video as a ‘deepfake,’ the video was actually not altered using AI or deep learning algorithms.
“During investigation, police summoned around 24 persons, including the Telangana Chief Minister, Revanth Reddy, asking them to join the probe. Among those summoned, five were from Telangana, one from Jharkhand, one from Nagaland Congress party, and another from Samajwadi Party,” according to the Indian Express.
Separately, the Election Commission of India directed X to take down a tweet by BJP Karnataka that posted an animated video demonising Muslims while targeting the Congress party.
The tweet appeared deleted after voting in the state was over.
Note, BOOM is a part of Meta’s third-party fact-checking program.