What people are saying
Nobody has said anything yet.
Brand New Thoughts/Media Literacy
Your Feed Is Not Reality: How Social Media Algorithms Shape What You Believe
A personalized feed can make a narrow slice of the world feel like public consensus. The problem begins when we forget that the timeline is a selection, not a census.
By ElleJanelle|July 30, 2026|4 min read
Everybody Is Not Talking About It
One of the easiest sentences to believe online is also one of the hardest to prove:
Everybody is talking about this.
Usually, we do not actually mean everybody. We mean the people, posts, clips and reactions that appeared in front of us.
That distinction became a Brand New Thoughts question during the episode Live News Dashboard. ElleJanelle noticed how often heavily online people describe an idea as being "everywhere" when the evidence they are looking at is their own personalized timeline. She pushed the thought one step further: if the algorithm determines so much of what appears in front of you, how much can it influence the world you think exists?
A feed is not a census.
Brand New Thoughts
The question is bigger than whether algorithms are good or bad. It is about confusing a ranked sample with the population itself.
Your Feed Was Chosen
Social platforms openly describe their feeds and recommendation systems as personalized.
YouTube says its homepage combines personalized recommendations with subscriptions, news and other information. Meta describes AI systems that rank Feed, Reels and Stories based on predictions about which content will be most relevant to a particular user.
That means two people can open the same platform at the same moment and encounter dramatically different worlds.
One sees a political scandal.
Another sees a celebrity breakup.
Another sees labor organizing.
Another sees sneaker releases, philosophy lectures and cooking videos.
None of those feeds is necessarily fake. The mistake is treating any one of them as an unbiased measurement of what the public thinks.
The missing denominator: why a timeline is not a poll
This is partly a data problem.
Imagine surveying 30 people about a political issue, except instead of selecting them randomly, you choose people because they resemble people you already agree with, interact with or spend time around.
You might collect 30 authentic opinions. You would still have a terrible national poll.
A personalized feed creates a similar danger when we use it to make claims such as:
- "Nobody agrees with that anymore."
- "Everyone is mad about this."
- "This is the biggest story in the country."
- "People have finally realized..."
- "Nobody even cares about that."
The posts may be real. The reactions may be real. What is missing is the denominator: compared with whom?
That is why a timeline can be emotionally convincing while remaining statistically useless as evidence of public consensus.
You train the algorithm too
There is another complication.
Algorithms do not create our information environments alone. We train them.
We follow people. We mute people. We click certain headlines and scroll past others. We watch one creator for 40 minutes and abandon another after five seconds. Our friends send us clips. We join communities. We search for subjects we already care about.
Then the platform observes those choices and makes more choices.
The resulting feed is not simply "what the algorithm did to us." It is a feedback loop between platform design, commercial incentives, social networks and our own behavior.
That makes the phrase "the algorithm made me believe it" too simple.
But it also makes "I chose everything I saw" impossible to defend.
ElleJanelle Tried Building Her Own Algorithm
The Brand New Thoughts conversation did not stop with criticizing recommendation systems.
ElleJanelle's response was to experiment with building a different information pipeline: the Ellements Dashboard, a custom news aggregator organized around subjects she deliberately wanted to hear more about, including federal, state, local and hyperlocal politics, technology, culture, philosophy, international news and Southern California stories.
The experiment matters even if no custom dashboard can eliminate bias.
Choosing your own categories still involves choices. Sources still make editorial judgments. Search engines still rank results. A person can intentionally build a news diet and still miss something important.
But there is a meaningful difference between having an information diet happen to you invisibly and periodically asking what ingredients are in it.
Audit your own feed: five questions
Is your algorithm biased? Of course it is.
Every selection process includes some things and excludes others.
A more useful audit is:
- What kinds of information reliably reach me?
- What kinds almost never reach me?
- Which claims am I treating as popular because I have seen them repeatedly?
- What evidence would I need before turning "my feed says" into "the public thinks"?
- Which sources do I consult when a story matters enough to leave the timeline?
That last question may be the most important.
Social media can be where you encounter a claim without being where you finish investigating it.
Reality Is Bigger Than the Timeline
A personalized feed can be useful, entertaining and even educational.
It can also make a tiny corner of the world feel enormous.
The danger is not merely that an algorithm might show us something false. It is that repeated exposure can change our sense of scale. A niche idea can feel universal. A major issue can become invisible. A loud group can look like a majority.
That is why media literacy increasingly requires a simple mental distinction:
I am seeing this a lot is not the same claim as this is happening a lot.
And "everyone on my timeline agrees" is still a statement about your timeline.
Watch the episode
What does your feed leave out?
Add to the DiscussionWhat topic do you suspect your own algorithm is making look bigger, smaller or more unanimous than it really is?