Charles Duhigg wrote the book Smarter Faster Better, in which he discusses the eight secrets of being productive in life and business. One of the eight principles is making effective decisions by forecasting the future.
In this post, we’ll discuss using probabilistic thinking and Bayes’ theorem to make better forecasts. And better forecasts are the key to better decisions.
Decision Theory
We can use modern psychology to understand how people make decisions. Psychology aims to simplify human behaviour with understandable rules. When we boil it down, many of our most important decisions are attempts to forecast the future.
For example, deciding to settle down and get married is forecasting that no one better will come along. While the idea may seem completely unromantic, getting married is actually a result of calculated costs and benefits.
Good decision-making depends on the basic ability to envision what might come next. People who are better at envisioning various futures are able to choose the best ones for themselves.
What is forecasting?
Forecasting is an imprecise, often terrifying science that forces us to confront how much we don’t know. The practice can be scary as it forces us to come to terms with the uncertainty of our future.
In 2011, the federal Office of the Director of National Intelligence started a study to understand how to produce the most accurate predictions of the future. Several universities were asked to compete in a forecasting competition. Some key insights stood out as enhancers of forecasting accuracy.
Firstly, research and statistical techniques training helped people make more accurate predictions about the future. Even simple training that taught people to think about the future helped improve their prediction accuracy.
The most valuable information was learning to think probabilistically. These courses involved breaking down the future into a series of possible scenarios.
For example, we can predict how long we’ll stay together with a romantic partner. Our future will depend on many factors. Firstly, having goals that coincide makes it more likely that you will grow in the same direction. We can also pull statistics that show how having children might impact the relationship.
With this data, we can then adjust the likelihood based on our own previous experiences and what we think will happen.
The key takeaway here is that we can make the vague future much more predictable by calculating what we do and don’t know. It all comes down to probabilistic thinking.
How to Think Probabilistically
Probabilistic thinking teaches us to not focus on the future as something that is going to happen. Instead, we can learn to see the future as a series of possibilities that might happen.
Probabilistic thinking forces us to hold multiple, conflicting outcomes in our mind to estimate their relative likelihoods. It involves looking at the full range of possibilities and assigning the likelihood of each outcome.
Probabilistic predictions also allow us to combine contradictory futures into a single prediction. By averaging out a broad spectrum, we’re left with a better forecast than if we’d only had one data point to start with.
Aggregating contradictory information allows us to use probability to improve our predictions of the future.
What Is Bayesian Thinking?
Bayesian thinking involves forecasting based on our existing data points. Bayes’ theorem is useful because it allows us to start with any existing data we may already have. Then we can update our predictions as more data becomes available to us.
The truth is everyone thinks in this way everyday, whether we realise it or not. When we notice new information in our environment, we connect new dots. This improved understanding allows us to make better predictions.
For example, most people can predict that a healthy 30 year old will probably live for another 50 years. We can also predict that a film that’s already made $60 million will make at least another $30 million. Both of these examples are accurate predictions following Bayesian thinking.
For a computer to make Bayesian predictions requires running thousands of models simultaneously and comparing millions of results.
People, however, intuitively understand why different types of forecast require different kinds of reasoning. Humans make these kinds of calculations without even thinking much about it.
Bayes’ rule states that even with very little data, we can make surprisingly accurate predictions. Bayesian thinking also allows us to skew existing data to our observations about the world.
The Risks of Bayesian Inference
Sometimes though, we make mistakes using the Bayes model. The risk comes down to not having enough information.
For example, students in a study believed that Pharaohs would had already reigned for 11 years would reign for another 23 years. The problem is that 35 would be considered an elderly age in ancient Egypt. Unlike European royalty, most Egyptian pharaohs would not reign for more than 20 years.
We need to fully understand the context to make better predictions. Having more data will improve the accuracy of our forecasts.
How to Succeed with Bayesian Predictions
When something bad happens to you or someone else, try and figure out why. Digging for this type of insight will be invaluable for future forecasting. We need to update our assumptions as we go along.
To succeed with Bayes’ theorem, try to envision various futures. Assign a probability to each of these futures based on your existing data points.
Build a Bayesian instinct by looking at past choices. Ask yourself why you were so sure it would turn out one way, and why you were wrong. You’ll likely gain new insight for your next important decision.
You’ll make better decisions by committing to this Bayesian outlook in life. After all, the ability to balance contradictory scenarios is itself an accurate predictor of success.
Balance Success & Failure Data Points
The Social Media Age has made it abundantly clear that people prefer to share the good things that happen to them. As a society, we’d much rather understand why great companies succeed. We don’t talk as much about the companies that fail.
Accurate forecasting, however, relies on exposing ourselves to both successes and disappointments. Too often we focus on the successes in the world: Apple, Google, Facebook. And we forget to consider what happened to the losers: Atari, Alta Vista, Friendster.
We harness the true power of Bayesian predictions with new and varied insights. That’s why focusing exclusively on successes won’t help us. If it did, dropping out of university might seem like the main predictor of success.
We need both accurate and balanced data as a starting point.
That’s why we should expose ourselves equally to successes and disappointments. Successful people tend to seek out more information about failures. For example, successful entrepreneurs tend to read more about companies that went out of business. This helps them understand the complexities of the modern economy.
People tend to avoid asking friends who were just fired questions that might seem rude. But this conversation would be full of insight. Next time a deal doesn’t go through, call up the other party and ask what happened. Maybe it’s something you could consider next time.
Use your experiences and those of people around you as learning opportunities to get more information about how the world works.
Probability Case Studies
In the 2011 national intelligence study, teams were asked to make several predictions. The first prediction involved president Sarkozy’s chance of being re-elected in 2012.
Forecasting Sarkozy’s 2012 Re-election
One group of participants attempted to predict the French election outcome by combining three factors: incumbency, approval ratings and economic growth rates.
Incumbency means that current presidents tend to have a 67% chance of winning re-election, because people like stability. Approval ratings, on the other hand, were low at 25%. And a forecast at the time focused on economic stagnation predicted his re-election at 45%.
The team took the average of these three probabilities to determine Sarkozy had a 46% chance of being re-elected. They predicted that Sarkozy would lose the election by 4%.
In reality, Sarkozy received 48.6% of the votes in 2012. Less than 2% from victory and surprisingly close to the team’s prediction.
This example shows how we can take three different perspectives and take an average to get the best prediction.
Probabilistic training increased prediction accuracy by up to 50%
A Poker Probability Example
In this example, let’s say you’re playing Poker. That’s Texas Hold’em to be exact.
There are four communal cards on the table, two of which are hearts. You’re also holding two hearts. One more card will be added to the table, and there’s a 9 in 37 chance that the last card is a heart, giving you a flush. That’s about a 20% chance you’ll win this hand.
In this round your opponent raises by $10 for a total pot of $100.
A novice player might fold after calculating the 20% odds. A professional, however, would calculate it in a different way. Playing this hand 100 times would cost a total of $1000 and return $2000. That’s $1000 profit after 100 rounds.
A 20% chance of winning 10x return means you will, on average, double your investment.
A Bayesian Poker Example
We can also consider poker from a Bayes’ rule perspective, starting each new game with assumptions about people. We’ll use the obvious signals at first, which includes their appearance and personality.
Someone who seems like a 40 year-old businessman might only care about telling his friends he played poker at the casino. A 22 year old with a poker t-shirt may play more online and thus have a limited strategy.
Of course throughout the game, these people will make decisions that tell you more about them. The 40 year-old could be a good bluffer, and the 22 year-old may start acting like a rich kid.
You can use these data points to predict their likelihood of having a good or bad hand in each round.
Effective Decision-making in a Nutshell
Effective decisions come down to making accurate predictions of the future. And we can learn to make more accurate predictions of the future with forecasting, probabilistic training and Bayesian thinking.
Forecasting helps us look to the future to predict possible outcomes. We can make much better decisions by improving our understanding of what the future holds for us.
Probabilistic thinking helps us take various predictions and aggregate them to provide a more accurate prediction of the future. This type of thinking helps us produce better predictions.
Bayesian thinking allows us to start with a small set of information. Then we can update our assumptions as we gather more data.
Make better decisions with probabilistic thinking and Mind & Practice today.