Instagram, X And The Art Of Changing Minds: Facebook Was Testing The Theory 14 Years Ago

Recently, it feels as though we have another conversation about how social media algorithms shape what we think, what we believe and who we trust every couple of months. The focus might be Elon Musk and X, Mark Zuckerberg and Meta, or the growing concern that a handful of extraordinarily wealthy people have enormous influence over the information millions of people see every day. And every time this happens, the concerns are equal parts valid and concerning.

It certainly feels as though this has hit a peak in recent years, and that may be somewhat accurate, but the underlying issue itself is nothing new. In fact, not only is the idea of individuals influencing popular opinion by means of social media a very familiar theme, it’s something that already attracted a great deal of controversy quite a while ago. And not just in the context of conversations among friends or even between industry professionals. It was actually a big deal, yet somehow our collective memory of this seems to have somewhat disappeared.

If you don’t know what I’m referring to, that’s okay, because up until quite recently, I didn’t really know much about the incident either. I think I vaguely remember the controversy, but I would’ve been a young adult with far more important things on my mind (or so I thought), so it really wasn’t a big deal in my world at all. Or, perhaps I’ve conjured up this memory while trying my best to think back to the 2010s. Either way, whether you were aware of it at the time or are only learning about it now, the fact that it isn’t spoken about frequently is quite bizarre, especially given the current world of social media and algorithms we’re currently drowning in.

So, what happened? Well, the year was 2012 and Facebook was booming. “The Social Network” had been released a year and a bit prior (bringing in a worldwide gross of $224,920,315), MySpace was gone with the wind and Facebook was at its absolute peak, with reports estimating that the platform had between 900 million and a whopping one billion active users throughout 2012.

Zuckerberg was flying high, and he and the team had bigger questions about the platform they had created and grown into the beast that it had become. What the platform could do for users was one thing, but what could Facebook learn from them in turn? And not only as individuals, but as a collective? How would users’ behaviour change if they altered what was shown on newsfeeds and timelines, and further to that, would it be possible to intentionally influence what people did as a result of what they were consuming?

Those sound like questions we’ve been hearing recently about how algorithms affect the content we’re shown, but it’s an issue that’s been alive in the zeitgeist for a far longer time than many may think. And it wasn’t just a hypothetical or a question Mark went to sleep thinking about every night. Indeed, in 2012, Facebook began testing whether changing the emotional content of users’ News Feeds could influence how they felt and what they posted. The results were published in 2014, and while it seemed as though the Facebook team was expecting a standing ovation, this prompted a significant wave of criticism and a clear apology from one of the researchers involved.

Regardless of whether or not your remember this happening, it’s quite incredible to look back at now. Today’s concerns about algorithmic influence are important and poignant, but they’re not new. We’ve known for years that social media feeds can easily affect users, but the really uncomfortable questions are why we still seem surprised by the idea, and why we allowed this news to fade into obscurity and be forgotten twelve years ago.

Facebook’s 2012 Experiment Tested Whether Emotions Could Spread Through a Feed

In 2012, Facebook conducted an experiment involving 689,003 users, manipulating the proportion of positive or negative emotional content appearing in their News Feeds. The research was published in the Proceedings of the National Academy of Sciences (PNAS) in June 2014 under the title, “Experimental Evidence of Massive-Scale Emotional Contagion Through Social Networks”.

The researchers were aiming to investigate whether exposure to emotional content affected users’ own emotional expression. In one group, Facebook reduced the amount of positive content users saw, and in another, it reduced negative content. The researchers then analysed the language users posted themselves. The findings suggested that emotions could indeed spread through social networks even without direct interaction between people. Users who were exposed to fewer positive posts subsequently used fewer positive emotional words in their own posts, while those exposed to fewer negative posts showed a corresponding reduction in negative emotional expression.

Of course this wasn’t a study proving that Facebook could make people change their political allegiance or rewrite their worldviews at will. It definitely was, however, evidence that the content appearing in a feed could very clearly influence how people expressed themselves. And it should have raised some very real, very loud alarm bells, because in this context, it was Facebook that was controlling what ended up in people’s feeds and their objective was purely analytical. But the results of the study revealed that it was possible that any other individual or group could potentially act in Facebook’s capacity and dictate what people felt and how they responded based on the type of information that was fed to them.

Of course, one of the main questions that was asked then and has continued to be raised since then is why Facebook was investigating this in the first place. It’s a good question and one worth answering, for sure. According to researchers, the study was designed to test a hypothesis about emotional contagion, rather than simply documenting something the company had already conclusively established.

Still, the very fact that Facebook could manipulate what users saw and measure the effects tells us something important about the platform’s power and just how long that power has been under investigation.

The Backlash Was About More Than the Results

The research might have been presented as a scientific contribution to understanding human behaviour, but many users and observers saw something rather different: a technology company had altered people’s feeds as part of an experiment without explicitly obtaining their informed consent for that specific manipulation. The immediate defence and explanation was that users had already provided consent by ticking the T’s & C’s box when they signed up, but unsurprisingly, that wasn’t particularly well received.

The controversy was widely reported, including by The New York Times and Forbes. The central concern was not just whether the experiment had produced interesting results, but rather whether Facebook should have been able to conduct it on users who had not knowingly agreed to participate in that particular study at all. Adam Kramer, a Facebook data scientist and co-author of the paper, subsequently apologised for the way the research was conducted. In a Facebook post reported at the time, he acknowledged that the study’s benefits might not justify the anxiety it caused.

But, perhaps the scientific question and ethical questions ought to be separated here. Researching how people respond to online content is not inherently sinister, but when a platform has the ability to change what hundreds of thousands of people see, the question of who gives permission (and who gets to decide what counts as an acceptable experiment) becomes considerably more complicated. For those affected by the experiment, it was deeply unsettling to learn how they’d been manipulated without having any idea that it was happening at all.

Facebook had demonstrated that its feed could be used to test the influence of content on people’s emotional expression, and the public reaction made clear that users were not necessarily comfortable with that power being exercised without their knowledge.

Fast-Forward To Musk, Zuckerberg and the Great Algorithm Debate

Today, the conversation has moved beyond emotional language in Facebook posts. Social media algorithms are now being scrutinised for how they rank news, promote political content and shape public debate, and the question of who controls these systems has become difficult to separate from the question of what influence they might have. The issue is far more complicated and nuanced now, but there’s no doubt about the fact that it has its roots in the same problems that were so clearly highlighted in 2014.

In March 2026, Phys.org reported on research examining how Twitter’s transformation into X under Elon Musk had changed the way the platform was governed. The study, published in New Media & Society, argued that changes to content moderation and platform rules had helped create a model in which the owner could exert substantial influence over the online environment.

And it’s safe to say that that concern is no longer confined to academics. According to The Guardian, in October 2026, French foreign minister Jean-Noël Barrot accused Musk of using X to facilitate online harassment campaigns against European leaders and destabilise democratic institutions ahead of France’s 2027 presidential election. These are serious allegations, not independently established proof that Musk has successfully manipulated an election, but they illustrate exactly how the power of platform owners has become a live political concern.

A study published in Nature in February 2026 examined the political effects of X’s feed algorithm: researchers randomly assigned active US-based users to either an algorithmic or chronological feed for seven weeks during 2023. They found that switching from a chronological feed to X’s algorithmic feed increased engagement and shifted some political attitudes in a more conservative direction. Users exposed to the algorithm were also more likely to follow conservative political activist accounts, and those following patterns persisted after the algorithm was switched off.

The results came with important points to note: that is, the study found no significant effects on users’ self-reported partisanship or affective polarisation, and its findings relate to the platform and period studied. Nevertheless, the research offers evidence that the way a feed selects and ranks content can influence political attitudes and online behaviour.

Sure, this isn’t the same as Facebook’s emotional contagion experiment, but the underlying question is very much connected. That is, what happens when a platform decides which information people see, how frequently they see it and what’s placed in front of them first?

We’re Still Debating The Power of Social Media Many Years Down the Line

There’s a tendency to treat every new revelation about social media algorithms as if it marks the beginning of a completely new problem. Of course, the technology changes, the people in charge come and go and the political stakes evolve. But the fundamental concern – int his case, that platforms can influence users by controlling what appears in their feeds – has been around for a long time.

Facebook’s experiment didn’t prove that social media companies can simply dictate what people believe, nor does it mean that every algorithmic recommendation is an attempt to manipulate users. Feeds are designed for several reasons, including personalisation, relevance and engagement, and it’s important to remember all dimensions of the situation.

But, that’s precisely why the issue deserves attention, both in the current context as well as in memory of the past. The ability to influence what people see doesn’t require a grand conspiracy or a company explicitly setting out to change their minds. It can emerge from decisions about what keeps people scrolling, what content gets ranked highly and what posts are quietly pushed further down the feed. And when those decisions are made by a small number of powerful companies, the consequences extend beyond individual users. They can affect the visibility of news, the reach of political messages and the information people draw on when forming opinions. And if we ignore the early warning signs, what happens if we only wake up when it’s already too late?

For startups and technology companies, there’s also another elephant in the room which is the question of accountability. The algorithms they build and deploy don’t operate in a vacuum: their design choices can influence behaviour, sometimes in ways that are difficult for users to see or understand. As AI-driven personalisation becomes more sophisticated, transparency and user control are likely to become even more important, both ethically and with regards to corporate regulations.

Facebook was testing the relationship between feed content and emotional expression in 2012, and fourteen years later, we’re still arguing about what happens when technology companies have so much control over the information we see and how that information shapes the way we understand the world. Perhaps the most surprising thing is not that these concerns have resurfaced, but rather how easily we’ve come to treat them as new.