Bayesian Updating: How We Predict the Future
I hope you’re enjoying the course so far!
Today’s mental model is going to show us the right way to use a crystal ball when thinking about the future. With Bayesian Updating, you’ll see that there’s nothing mysterious about it at all.
What Is Bayesian Updating and Why Is It True?
This mental model is named after English minister Thomas Bayes, who wrote about the concept in an essay published in 1763, later inspiring economists and probability theorists, including the renowned Pierre-Simon Laplace.
What Bayes suggested for future forecasters was this: Factor in everything known prior to the prediction. Then, for every guess made, collect additional information and modify the prediction again. Continue to update in this fashion.
This is Bayesian Updating.
Machine learning is a good example of Bayesian Updating in action. Everything you do to your devices is collected in a cache of memory, and your devices try to learn how to better predict what you will do. Take the example of a Google search autofill or autocorrect on your phone.
This is a great way to explain why Bayesian Updating works:
When we make decisions about the world, we base them on what we think we know.
When we make intelligent decisions about the world, we factor in all prior known information.
When we make the most intelligent decisions possible, we update our understanding of what that means each time, based on previous decisions.
We can’t ever know the future with certainty, but we can eliminate numerous wrong assumptions about the future and continually update our expectations with them.
Applying Bayesian Updating
You make decisions continually. Many times, you might be reacting and not sure of the answer. Do you say yes to the interview opportunity with an obscure blogger that’s going to take two hours of your time to prepare? Do you rush to get the leaky toilet fixed and interrupt your work day (even though it still works just fine)?
I can certainly tell you that there are decisions I now make differently.
Once upon a time, when an email would pop up in my inbox, I’d be quick to check it and reply.
Today, I ignore it unless it’s an emergency (I don’t even keep my email tab open).
Why do I make this decision? Because I’ve learned that through choosing the first action, my productivity goes down, I get stressed out unnecessarily, and most importantly, I waste time writing longer emails, with unnecessary verbiage, that I really should either not write at all or take one to two sentences to write instead.
You likely have many situations in your life where you can learn from outcomes from the same kinds of choices. Going back to Lesson 1, a morning routine is a good example. What’s the best way to start your day? There isn’t a perfect way, but you can learn from continually trying different things and keeping track of how you function, which is the best way to start.
The best situation to apply Bayesian Updating is any outcome that matters to you.
For example, if you want to spend 20 hours a week reading, what choices must you make for that to happen? You might start by trying to read for ten hours on Saturday and ten hours on Sunday, since these are your days off.
Monday comes, and you check your data: three hours total! Ouch!
But you try again next week, and you apply Bayesian Updating to learn from the last time. Of the three hours, you got them in near the end of the day. So, you try three hours at the end of every day. This will land you at 21 hours, so there’s grace room.
Next Monday comes and you check your data: 13 hours. Much better!
Most of those were 45- to 60-minute sessions on weeknights. Reassessing your weekends, you realize you just have too many family commitments. You plan to go off to a quiet coffee shop the next weekend. Instead of shooting for 20 hours, you want to beat 13 hours, and your goal will be to keep beating your previous week until you eventually hit 20 hours.
You can see in this example how Bayesian Updating can help you arrive at an outcome, but you’ll have to continually update your understanding of how you fail to achieve it, so as to try other pathways.
Bayesian Updating is a mental model that allows you to continually improve your decisions based on using everything you know beforehand and everything you learn from previous decisions.
Do the above Bayesian Updating example on reading for 20 hours, but with any important outcome you want to see each week.
Tomorrow, we’ll explore a mental model that’s about the power of silence. Stay tuned!
Share with friends