Nobody opening an app has ever been handed an instruction about what to believe.
And yet the question keeps returning: how free is a thought built on material somebody else selected?
The best evidence says "less than we assume"
In 2020 the largest experiment ever permitted to outside researchers on this ran on Facebook and Instagram. Around 23,000 users spent three months without the algorithm: they saw posts in chronological order, the old way.
The result was not what anyone expected. Polarisation didn't move. Neither did their views on immigration or the pandemic, their political knowledge, or their trust in the media.
Two things have to be said alongside that. The researchers themselves noted that three months is long for an experiment but short for changing a belief. And that Meta had a say in how the questions were framed.
Even so: the simple accusation — "the algorithm makes you believe what it wants" — doesn't hold up the way it is usually stated.
Except that wasn't the question
Free thought isn't judged at the conclusion. It is judged at the material.
The algorithm doesn't choose what you think about something. It chooses what you think about — which subject appears nine times in a day and which appears not at all.
The conclusion stays yours. The agenda doesn't.
What you weren't shown leaves no gap
This is the insidious part.
When content is cut from a feed, no hole is left behind. Nothing says "47 posts were hidden". A feed always looks complete, however much is missing from it.
So there is no way to audit what you didn't see. And a judgement whose inputs can't be audited isn't quite your judgement.
There is something quieter too: your sense of what people believe. That doesn't come from people, it comes from ranking. The algorithm may not change your opinion — but it changes your estimate of how many share it.
So where is the freedom
This isn't mind control. Nobody is being manipulated in the way we tend to fear.
It is something narrower and harder to see: the feed isn't optimised to be true, or representative. It is optimised to hold attention. That is its purpose, written into it.
So the freedom that remains is real, but smaller than it looks: the freedom to judge — not to choose what you judge.
And the question isn't whether somebody is telling you what to think. It is whether you know what they didn't show you.
Source: Guess et al., "How do social media feed algorithms affect attitudes and behavior in an election campaign?", Science (2023) — part of the Meta 2020 US Election Research Project.