Judging people by their beliefs is not scientific.

Nassim Taleb

Why Believe?

Beliefs play several roles.

Some beliefs are substantive. They possess action-guiding content. They concern mundane physical facts and models of people. We have beliefs about how best to get to work, how to make various foods, what gifts our friends would appreciate whether the people around us are trustworthy, and so on. These beliefs represent the way things are and, as such, are relevant for action. We take substantive beliefs literally - it's the substance or content of the belief that matters.

Other beliefs play a social role. Here, the content of the beliefs is not as important. Instead, it's what we say to others that matters. Stealing economist Robin Hanson's metaphor, these beliefs are like clothes. We communicate and express ourselves through them. My beliefs say something about me. If you share the same beliefs, that says something about us.

Substantive beliefs typically have tight feedback loops. Reality rewards us when we act on true beliefs and punishes us when we're mistaken. If I mistakenly believe that there is no car coming my way while I cross an intersection, that can result in injury or worse. Similarly, with less drastic consequences, mistakes about a meeting's location or a friend's dietary restrictions will make a difference.

Different incentives drive social beliefs. Here other people play the role of praisers and punishers. We're incentivized to seek praise and avoid the taunts of those who matter to us.

How someone wants to be seen and who they want to be with will push them towards holding some beliefs and not others. Believing that everyone has a right to healthcare communicates affiliation with political leftists. Believing that evolution is true expresses admiration for scientists and science generally.

Intuitive and Reflective Beliefs

Substantive beliefs are usually what cognitive scientist Dan Sperber calls intuitive beliefs - beliefs we automatically draw inferences from. We form these beliefs through simple perceptual and inferential processes. We often let them roam freely in our minds and guide action. If you notice there's a table in my room, that belief is intuitive. You'll freely make inferences about what other furniture would fit in my room, how many people could eat at the table, and so on. Similarly, if you notice a car coming before you cross the street, you'll infer that you should wait to cross.

Contrast intuitive with reflective beliefs. Reflective beliefs don't have immediate behavioral or psychological consequences. They tend to have abstract and complex content.

Reflective beliefs are compartmentalized. If I have false beliefs about God's existence, that will have no direct impact on my life. I will not ever know, at least not on this earth, whether or not I am correct. Similarly, if I am wrong about reflective beliefs concerning the justice of the Iraq war, the feasibility of single-payer healthcare, or the theory of evolution, that will make no immediate difference. Neither I nor anyone else regularly uses such beliefs to navigate through the world. It takes serious mental effort to draw out the inferences of each one. And there's little incentive to do the required intellectual labor.

Complex and abstract political, scientific, and religious beliefs are reflective and social. They are not treated literally.

Instead, they are principally about meaning, community, and entertainment.

Believing in Fake News

Let's look at the phenomenon of fake news. In a strange turn of events, many people came to believe that a Washington DC pizza parlor was a host for an elite child-trafficking cabal in 2016.

The conspiracy returned earlier this summer, now merged with other theories:

In the first week of June, comments, likes, and shares of PizzaGate also spiked to more than 800,000 on Facebook and nearly 600,000 on Instagram, according to data from CrowdTangle, a Facebook-owned tool for analyzing social interactions. That compares with 512,000 interactions on Facebook and 93,000 on Instagram during the first week of December 2016.

How many people believed this story in the substantive sense? Only a minority.

A few believers rated the store poorly online or left threatening messages on alleged participants' voicemails.

But in general, conspiracy theorists don't act in a way that their beliefs would predict. In their view, the government, the rich, and other influential members are enmeshed in complicated and powerful networks. In response to this, most will share their suspicions with friends, post on the internet, and attend public lectures. That's the extent of their conviction.

Only a handful of 9/11 truthers would share their doubts with a first responder. Conspiracy theorists hound parents of the Sandy Hook massacre, but it's a minority of the actual audience of "believers."

In 2016, one man entered the pizza parlor armed, demanding to know what was going on. He opened fire in a closet.

His actions would make sense if the conspiracy were substantive and action-guiding. But instead, he is a moron who took the game too seriously. Belief in conspiracy theories is reflective, not to be taken literally. Most believers implicitly understand this.

Believing in God

There are significant differences between religion and conspiracy theories, but the typical conspiracy theorist is like the typical religious believer in this respect.

When I was a child, I thought it was odd that other kids in Sunday school didn't read the required passages and struggled to pay attention. We were studying the word of God. If Christianity is true, nothing could be more important! With few close exceptions, many Christians I knew were not seriously devoted to their beliefs. If it were true, how could you afford to be anything but a zealot?

The best explanation for most believers' behavior is that they don't believe that Christianity is literally true.

Consider belief in an afterlife. The afterlife is not something one sees or feels concretely. Instead, it is something one believes in through interpreting testimony, abstract argument, or extraordinary experiences. Even the most confident religious fundamentalists do not believe in an afterlife, in the same way, that we believe in cars or tables. With few exceptions, people who believe in an afterlife do not engage in riskier behaviors or grieve loved ones' death any less. These are indicators of a reflective and social belief.

Religion is about practice, not a substantive belief. The primary role for beliefs like "the Bible is the word of God" and "there is an afterlife" are social ones. Their content does not promote action. Instead, religion is primarily about attending church, participating in Sunday school, tithing, and generally being a good community member. Religious believers are prosocial within their community, but only a minority of them are zealots.

In fact, the zealots are often pushed out. In Sunday school, I was the weirdo who took things too seriously. Later, when I converted to atheism, I made the opposite mistake. I thought religion was primarily about accepting or rejecting substantive beliefs.

Trader and philosopher Nassim Taleb captures this in the aphorism:

Atheists are just modern versions of religious fundamentalists: they both take religion too literally.

Anti-Fake News Powers Fake News

One response to misinformation, whether political or religious, is to insist on stronger epistemic norms. Punish fake news! But this strategy inadvertently elevates the status of fake news.

What these responses miss is that sharing fake news is a social act. When you do so, you're alienating yourself from mainstream orthodoxy and binding yourself to a niche.

Think of it like burning bridges. If someone has nowhere else to go, they're bound to be loyal.

This dynamic explains why some political leaders, like Donald Trump, require subordinates to tell blatant lies. The economist Tyler Cowen explained:

By requiring subordinates to speak untruths, a leader can undercut their independent standing, including their standing with the public, with the media and with other members of the administration. That makes those individuals grow more dependent on the leader and less likely to mount independent rebellions against the structure of command. Promoting such chains of lies is a classic tactic when a leader distrusts his subordinates and expects to continue to distrust them in the future.

Recent disputes over the election demonstrate how political talk extracts its value by polarizing.

Skepticism over the election results expressed loyalty to a particular group because it's bound to get you expelled from another one. If it didn't do that, then it wouldn't have any political use.

The value of spreading and consuming fake news value derives from the perceived value of its expression. Or more precisely, the perceived difference between sharing and not sharing fake news. The act of sharing something politically insensitive is, in many situations, relatively risk-free. But if it's perceived as a risky, brave act, there's more value in the ideas' expression. If a messenger pays a high cost to deliver a message, it must be important.

Correcting tweets is one way to increase the cost of a message and clarify where political players stand. So we should expect corrections by social media companies to strengthen loyalties, not to promote the truth.

A recent Cornell study shows exactly this:

Corrections... had dramatically different effects on the beliefs of Democrats and Republicans. Among Democrats, exposure to all three corrections of varying strength decreased belief in mail fraud; in the flagged and Twitter correction the decrease was 10%.

Among Republicans, however, we observe the exact opposite. Belief in voter fraud was actually higher in all three corrections treatments than in the control group baseline – evidence of a backfire effect.

If we promote higher epistemic norms, that raises the cost of sharing misinformation. If there's already a social reason to share the fake news, this cost makes the signal more genuine.

Anti-fake news messaging is fuel for fake news.

If the correction were for a substantive, action-guiding belief, then it would have more force. But by attempting to promote truth through fact-checking, Twitter made the same mistake I did in Sunday school: mistaking a social belief for a substantive one.

Take Me Seriously, Not Literally

Once one realizes that most political and religious beliefs aren’t substantive, believers look less irrational.

By and large, people are not gullible drones or lemmings. Instead, they are ordinary, mostly selfish creatures responding to whatever situation they happen to be in.

For most, the reward for truth and rigor isn’t high for political beliefs.

Instead, we use these beliefs to entertain and bind us to one another.

Thanks to Sachin Maini, Étienne Fortier-Dubois, Dan Stern, Elaine Lin, and Paul Orlando for reading earlier versions of this essay.