The Update: A Practical Guide to Changing Your Mind Like a Scientist

The Stubbornness Problem

Something curious happens when we form an opinion. Watch it in yourself: the moment you decide your new colleague is competent, or that a particular restaurant is overrated, or that your business strategy is the right one—that belief starts to calcify. New information that confirms it feels satisfying. New information that contradicts it feels like an attack.

This isn't a character flaw. It's how human brains evolved to work. In a world where quick decisions meant survival, constantly second-guessing yourself was a liability. But we don't live in that world anymore. Today, the ability to update your thinking gracefully—to change your mind based on evidence rather than ego—is one of the most valuable skills you can develop.

Enter Bayesian thinking. Named after Thomas Bayes, an 18th-century statistician and minister, this approach to reasoning has revolutionized everything from medical diagnosis to spam filtering. But its most powerful application might be the most personal: learning to hold your beliefs lightly and update them systematically.

Beliefs as Probabilities, Not Positions

The core insight of Bayesian thinking is deceptively simple: instead of believing something is true or false, assign it a probability. Not "my business partner is trustworthy" but "I'm about 80% confident my business partner is trustworthy." Not "this marketing strategy will work" but "I'd give this approach a 60% chance of hitting our targets."

This shift changes everything. When beliefs become probabilities, updating them stops feeling like admitting defeat and starts feeling like calibration. You're not abandoning a position—you're refining an estimate.

Here's how it works in practice. You start with a prior probability—your initial estimate before new evidence arrives. Then, when you encounter new information, you adjust that probability up or down based on how likely that evidence would be if your belief were true versus false. The result is your posterior probability—your updated belief.

Let's make this concrete. Say you're 70% confident that your new hire will work out well. After three months, you notice they've missed two important deadlines. How should this update your confidence?

The Bayesian approach asks: How likely would missed deadlines be if this person were going to succeed long-term? How likely if they weren't? If strong performers rarely miss deadlines in your organization, this evidence should shift your confidence downward—maybe to 50%. If deadline-missing is common even among your best people, the update should be smaller.

The beauty is in the proportionality. You're not lurching from "this person is great" to "this person is terrible" based on one data point. You're making a measured adjustment that reflects the actual diagnostic value of the evidence.

The Art of Setting Priors

Before you can update well, you need to start well. This is where many people stumble. Setting your initial probability—your prior—requires a particular kind of honesty.

The most common mistake is overconfidence. Studies consistently show that when people say they're 90% sure of something, they're right about 70% of the time. We systematically overestimate how much we know.

A useful corrective is the outside view. Before asking "How likely is my startup to succeed?" ask "How likely are startups like mine to succeed?" Before asking "Will my marriage last?" ask "What percentage of marriages that look like mine at this stage last?" Base rates—the frequency of outcomes in similar situations—provide a reality check on our optimistic intuitions.

Another technique is to imagine a thoughtful person who disagrees with you. What would they say? What evidence would they point to? This exercise alone often reveals that your initial confidence was inflated.

When I'm setting priors on important decisions, I find it helpful to consult diverse perspectives—people with different backgrounds, expertise, and thinking styles. Tools like thonk can help assemble these varied viewpoints systematically, ensuring you're not just hearing echoes of your own assumptions. The goal isn't to average everyone's opinions but to stress-test your initial estimate against genuinely different ways of seeing the situation.

Evidence Isn't Created Equal

Not all information deserves the same weight. A crucial Bayesian skill is learning to assess the diagnostic value of evidence—how much it should actually shift your beliefs.

Consider two pieces of information about whether a job candidate will succeed:

1. They have impressive credentials from a prestigious university
2. Their former colleague, unsolicited, reached out to warn you about working with them

The first piece of evidence is weak. Many people with impressive credentials underperform, and many without them excel. The base rate of success given prestigious credentials isn't dramatically different from the overall base rate.

The second piece is strong. People rarely go out of their way to warn others unless something significant happened. The likelihood of receiving such a warning if the candidate is problematic is much higher than if they're not. This evidence should move your probability substantially.

The general principle: evidence is diagnostic to the extent that it would be much more likely in one scenario than another. A test that comes back positive regardless of whether you have a condition tells you nothing. A test that's positive only when the condition is present tells you everything.

Apply this to everyday decisions:

• Your friend recommends a restaurant. How diagnostic is this? It depends on whether they recommend everything or are highly selective.
• A business metric improved after you made a change. How diagnostic? It depends on how often that metric fluctuates anyway.
• Someone promises they've changed. How diagnostic? It depends on how often people make such promises without following through.

The Update Ritual

Knowing the theory is one thing. Actually updating your beliefs in the moment is another. Here's a practical ritual for important decisions:

Step 1: State your current probability. Write it down. "I'm 65% confident this product launch will hit our targets."

Step 2: When new evidence arrives, pause. Don't immediately react. Instead, ask: "What would I expect to see if my belief were true? What if it were false?"

Step 3: Assess the evidence's diagnostic value. Is this the kind of thing that would happen either way, or does it strongly favor one hypothesis?

Step 4: Make a proportional update. Strong evidence for your belief? Move up. Strong evidence against? Move down. Weak evidence? Small adjustment or none.

Step 5: Record your new probability. "Based on the early customer feedback, I'm now 55% confident."

The act of writing forces precision. It also creates a record you can learn from. Over time, you'll discover your own patterns—where you tend to be overconfident, which types of evidence you overweight or ignore.

When to Stop Updating

Bayesian thinking can become its own trap. Some people update so readily that they never commit to anything. Every new data point sends them spinning in a new direction. This isn't wisdom—it's paralysis.

The solution is to distinguish between decisions that need to be made and beliefs that can remain provisional. You can hold a 60% confidence that a strategy will work while still committing fully to executing it. The belief stays probabilistic; the action becomes definite.

There's also a point where further updating yields diminishing returns. If you're 95% confident in something, it would take extraordinary evidence to move you to 98%. At some point, you've gathered enough information to act. Continuing to seek more is just procrastination wearing the mask of rigor.

The wisdom lies in knowing which mode you're in: still gathering evidence, or ready to decide. As we often explore on thonk, the best decisions usually involve both—gathering diverse input, then committing with conviction even amid remaining uncertainty.

The Humility Dividend

There's a deeper reward to Bayesian thinking beyond better decisions. It cultivates a particular kind of humility—not the self-deprecating kind, but the intellectually honest kind.

When you hold beliefs as probabilities, you implicitly acknowledge that you might be wrong. This makes you more curious, more open to perspectives that differ from your own, more willing to seek out information that might challenge you. You become a better listener because you're genuinely interested in evidence that could update your view.

This posture is rare and valuable. Most conversations are people defending positions. Bayesian thinkers are doing something different—they're trying to get closer to the truth, and they recognize that other people might have pieces of that truth they're missing.

It also brings a strange peace. When you're not defending beliefs as part of your identity, contradictory evidence loses its sting. It's just information. It's an opportunity to be less wrong than you were before.

Starting Small

You don't need to become a probability-calculating machine overnight. Start with one domain where you make frequent decisions—hiring, investments, health choices, relationship judgments.

For the next month, try this:

• Before forming a strong opinion, assign it a probability
• Write down what evidence would make you update up or down
• When that evidence arrives, actually update
• Track your predictions and see how well-calibrated you are

You'll likely discover you're overconfident in some areas and underconfident in others. You'll notice which evidence you tend to dismiss and which you overweight. This self-knowledge is the real prize.

Bayesian thinking isn't about becoming a robot. It's about becoming more honest with yourself about what you know and don't know. It's about treating your beliefs as tools for navigating reality rather than possessions to protect. And in a world that rewards adaptability over stubbornness, that's a skill worth developing.

The next time you're certain about something, pause. Ask yourself: what probability would I actually assign to this? What would it take to change my mind? You might find that the question itself is the beginning of clearer thinking.