In India, late May 2026 ‘National Eligibility cum Entrance Test, got leaked and in days anger over the leak grew into a nationwide student movement, supported by a self-mocking name, the Cockroach Janta Party, after a dismissive remark made comparing young people to pests. By June it had filled not only Jantar Mantar, a permitted area to protest, but spilled over to other parts of the country. Students went on hunger strike, an opposition MP was briefly detained, and more such events happened.
Police response to curb the alleged instability was tear gas and baton charges. What began as frustration and anger over an exam leak, turned into one of the sharpest democratic challenges that the government has faced in years. Most of us watched it unfold in the same manner we consume or watch everything now, through our phones!
(Sign up for THEdge, The Hindu’s weekly education newsletter.)
Two people, coming from the same socio-economic backgrounds, scroll the same social media platform at the same time, watching the same event unfold in real time. However, by the end of the scroll, they have different conclusions regarding the same event. One has concluded a story of injustice while the other has concluded disorder and threat. Neither is lying nor the content seen is fake. Each has simply consumed different slice of the same cake, one which had frost, the other which has bread.
Sadly, this is how a majority of people are forming opinions on issues, not by reading, but by feeding themselves. Feeding in such manner is not neutral, as recommendation engines are not built to inform, they are built to hold your attention for long, because ‘attention’ is the product being sold. So, the algorithmic system will decide the definition for what ‘everyone is talking about’ for ‘you’. As a domino effect your world is quietly also deciding what you are going to feel about it, and by extension what will you go on to believe and prefer to further consume.
The algorithm’s illusion
Part of what I have said above feels convincing to me is because it has a name in network science: ‘majority illusion’. It has been explained by Kristina Lerman, an academician, in her research regarding paradoxes of social perception and ‘majority illusion’. It says that in a network where a small group of highly connected accounts generate a lot of footfall, a view held by a few people seems to have an overwhelming consensus. Here, the opinion did not spread on merit, rather the network’s structure just made it look that way. Hence, the two people who scroll two different feeds feel sure and certain that the content in their feed is what everyone thinks, while seeing the two sides of the same coin.
It is important to discuss this, as it rides on top of two subtle issues. First is the automation bias i.e. human tendency to perceive a decision as more trustworthy simple because it was made by a machine. Researchers who study human-automation interactions have shown that people believe a computer’s recommendation over and above their own judgment, or over contradictory evidence, because of the assumption that machines are neutral, above the human bias and exhaustive. Similarly, a recommendation engine does not seem to have an agenda, it is just showing what is popular as per your definition and what will suit your intellectual palate.
That perceived neutrality is what makes it powerful. For example, if a friend curated your news feed and diet, you may still use your critical thinking to question and resist it, but when an algorithm is doing that same thing silently, at scale, every day, most people forget to question and read and rather take it face value.
Bias by design
The second issue is the biased dataset. The simple argument is that the machine is not neutral because it is trained on the data generated by humans, and humans are not neutral. A clear example of this is the Amazon’s experimental AI hiring tool (reported by Reuters in 2018), which was trained on ten years of the company’s actual hiring data and built to score resumes on a five-pointer scale. Because in the past tech-hiring had long skewed male, the model itself learnt that male resumes were a safer bet and thereby began downgrading applications that included the term women or women colleges. Nobody while coding gave a rule to prefer men, no bias was built into the code. It was in the history, that the code was asked to imitate.
Social media feeds also run on the same logic, aimed at a different target. While they are not trained to determine who is a good employee, they are trained to analyse and predict what will hold your attention based on what was held before, and what has held the attention of people who click and watch like you. Just as Amazon’s tool mistook what we did in the past for what is correct today, similarly your feed mistakes ‘what kept you here yesterday’ for ‘what you need to know today’.
Layer automation bias and the majority of illusion on top of that, you have the perfect recipe to make one believe and convinced that their political view is simply the obvious conclusion on reasonable feeding of events, because an object (i.e. algorithm) they do not question about having a bias or stance kept confirming it and a majority around them in the feed created the illusion to agree.
Who shapes truth?
This is exactly what happened with the protests I described at the start.
Depending on which feed suited your palate, you may have seen tear gas and baton charges, or videos framing the police response as necessary control of an unlawful mob blocking the capital. Whichever version reached your feed first was selected because it was more likely to keep your attention. It had already shaped your opinion. By then, it felt like the truth, not because you had verified it, but because your feed never gave you a reason to question it.
Both videos are real, but almost no one scrolling the feed asks the basic question of what happened first and how to chronologically verify the order? A few people have taken this basic question to court, where it sits sub-judice, waiting for a verified answer. But most cannot wait, as they prefer truth to arrive in same time as the reels or videos seen on social media that has trained them to expect instant answers, delivered at scale. In a true democracy, I often wonder is that the right way to arrive at the truth?
Well, it not just a theoretical mechanism now, as it is something a platform has admitted and apologised for. Meta chief executive officer Mark Zuckerberg conveyed his “apologies” for child sex abuse material (CSAM), “deepfake content and errors in operating the platform,” government sources said, after Joel Kaplan, the technology giant’s global head of public policy, met with officials from the Ministry of Electronics and Information Technology on August 5, 2026. So, it is safe to say the platforms that appear neutral, are not. The platform is not just informing you, it is hosting an opinion in your feed, made to match your palate to keep you engaged longer.
I do agree that it does not imply that both events are not real or one is fabricated. My point is that the version reaching you has already been filtered through a system optimised for something other than your understanding. Having said that, the responsibility is not only of the citizen here, but the Governmental authorities also have a strong responsibility too. They need to publish primary data, answer direct questions in a form of a dialogue and not one way of monologue, rather than leaving citizens to reconstruct the truth from soundbites, a spokesperson’s speech, or dueling clips online. A democratic society who is fed only curated content is just as vulnerable to distortion as one fed only on algorithm, hence accountability must run in both directions.
Learning to read
In my opinion the remedy is not cynicism about all information, rather it is the intake of information in a slow manner. Keep scrolling, watching, the feed is a good place to learn about what is happening around, but it is a poor place to firmly form your opinions. Treat that thirty-second clip as a headline, not a conclusion. Before you form or repeat an opinion, go find the full report from more than one outlet, read the past context and information, check what each side has said. It is the merely the difference of having an opinion and being served one, and it takes a little extra effort to find out which one you have got.
This is not only for the young scrollers, who are usually assumed to be the ones who need to be taught the skill of critical thinking. In reality, this habit of hosting an opinion rather than taking effort to form one, cuts across age. An anchor, newspaper, commentator, celebrity or an influencer says something once, which happens to resonate with our own gut instinct or beliefs, then usually from that point onwards we stop questioning that voice altogether irrespective of the issue, context, and simple agree and believe in advance. Reading fully, verifying facts from more than one source, is no longer a skill meant only for young generation, rather for every one of us on how we consumer our information today.
(Paavni Jain, Assistant Professor, Jindal Global Law School (JGLS) and Assistant Director, CSCAILR. Views are personal.)
