The internet is a confirmation bias machine

Over the past several months, I keep coming back to this thought: The internet is a confirmation bias machine.

I once believed that because the internet exposes people to new ideas, it helps us to think more carefully. I still believe it can, but the vast majority of the time, the internet paves the way to motivated reasoning.

One of the best descriptions I’ve read of motivated reasoning comes from Jonathan Haidt’s 2012 book The Righteous Mind (p. 98-99):

The social psychologist Tom Gilovich studies the cognitive mechanisms of strange beliefs. His simple formulation is that when we want to believe something, we ask ourselves, “Can I believe it?” Then… we search for supporting evidence, and if we find even a single piece of pseudo-evidence, we can stop thinking. We now have permission to believe. We have justification, in case anyone asks.

In contrast, when we don’t want to believe something, we ask ourselves, “Must I believe it?” Then we search for contrary evidence, and if we find a single reason to doubt the claim, we can dismiss it. You only need one key to unlock the handcuffs of must.

Psychologists now have file cabinets full of findings on “motivated reasoning,” showing the many tricks people use to reach the conclusions they want to reach. When subjects are told that an intelligence test gave them a low score, they choose to read articles criticizing (rather than supporting) the validity of IQ tests. When people read a (fictitious) scientific study that reports a link between caffeine consumption and breast cancer, women who are heavy coffee drinkers find more flaws in the study than do men and less caffeinated women. […] The difference between a mind asking “Must I believe it?” versus “Can I believe it?” is so profound that it even influences visual perception. Subjects who thought that they’d get something good if a computer flashed up a letter rather than a number were more likely to see the ambiguous figure as the letter B, rather than the number 13.

We ask, “can I believe this?” if we want to believe something, and “must I believe it?” if we want to disbelieve. When we frame the question like this, we only need one piece of evidence to support our desired conclusion.

The heading for this section of The Righteous Mind is “Reasoning (and Google) can take you wherever you want to go,” because all you need to do to find that one reason is to search the web for it. Anything that brain of yours can think can be found on the internet, so when we go looking for a reason, we will find one. The path to motivated reasoning is paved online.

It’s even worse though in the current era of consolidated corporate social media, where just a few private algorithms control the online public square. We don’t even need to go looking for the reasons to support our pre-determined conclusions. The reasons come looking for us.

When Charlie Kirk was murdered, I started to really notice the pattern. I saw people on the right thinking, “I bet the left is celebrating this,” and people would go looking and it wouldn’t be hard to find examples (“Can I believe this?”). I saw people on the left saying, “I bet the right is calling for violent retribution,” and people would go looking and it wouldn’t be hard to find examples (“Can I believe this?”). The people who go looking share their examples widely, and many other people see them without trying to search for them. More recently, I’ve seen the same phenomenon with the killing of Alex Pretti. Frankly, you see it on any big story, and in little ways every day. I’m not saying it’s never valid to criticize what your political opponents are saying or doing, but people go looking for examples — and of course they will find them somewhere. If you believe your opponents are evil, and you go looking for some examples from someone on the other side (“Can I believe this?”), you’re going to find it on the internet.

I see this more and more in our balkanized social media landscape. People on the service formerly known as Twitter go to Bluesky just to search for examples of silly things they think people will be pearl clutching or sanctimonious about, and of course you’ll find some examples of it. People on Bluesky share screenshots from “the bird site” of the bottom of the barrel stuff they expect to find there, and of course the worst screenshots will circulate. (And people in the fediverse mock the trashiest stuff from Twitter/X and the self-righteousness smug from Bluesky on Mastodon.)

I think it’s perfectly valid to check what leaders on the other side of an issue are saying, or to try to get a sense of what the average person who disagrees with you thinks. I’m not criticizing that. Rather, I’m criticizing the tendency that I see far more often — maybe that gets spread more through the algorithmic filter because of its rage-bait potential — to cherry-pick examples of the awful things you imagine your opponents to be saying as if it’s representative, in order to validate predetermined conclusions that they’re obviously evil.

If you imagine someone who disagrees with you probably believes EvilThingX, and then you go looking for examples of that on the internet, you will find examples. The internet is a confirmation bias machine.

Even if you don’t go looking but you follow people who do, you will see it. And, increasingly, even if you don’t follow the people who go looking and post it, you’ll still probably see it in your feed on mainstream social media, because that’s what generates views and engagement. The internet is a confirmation bias machine.

I’m writing this post as much for myself as for anyone else who cares about thinking carefully and finding the truth, rather than just dunking on the other guys. It’s so easy to fall into this way of thinking. The path to motivated reasoning is paved in the current social media landscape.

What can we do about it? First, we need to notice the pattern so we can be conscious of it. Am I falling into the can / must method of motivating reasoning? Am I asking myself the right question? Or just the question that will lead to the conclusion I want?

Second, as we notice the pattern, I think we need the counterweight of steelmanning our opponents arguments. In a way, this is all just the age-old phenomenon of strawmanning, which is obviously not unique to the internet. But the internet makes it so easy. In fact, it has come to seem like the internet makes it the default — fuelling polarization, conspiracy thinking, eisegesis, contempt for the “other side,” and the worse demons of our nature.

But it doesn’t have to be our default. You can choose steelmanning over strawmanning. If you find yourself looking for examples of the worst from your opponents, maybe check to see if there are also examples of the best — or go looking for something that might surprise you or challenge your assumptions. If you find yourself asking, “can I believe this?”, also challenge yourself with “must I believe this?” — or vice versa. Reverse the question on yourself to challenge your predetermined conclusion before you make your mind up for good.

The internet is a confirmation bias machine. We need to understand and be conscious of that if we want to avoid constant confirmation bias and motivated reasoning in our own thinking.