AI tools have elevated the importance of prediction-based decision making. But this has always been a strategy used by savvy leaders who understand that the perfect plan depends on information that is rarely available. Instead, they have learned to live with uncertainty and find ways to keep moving by thinking in terms of probabilities instead of sure things.

Read on: blog.WednesdayWin.com​

Read time: 4 minutes

A critical skill for anyone new to leadership is the ability to make good decisions quickly. This can be a tough transition for individual contributors or recent graduates who are used to only facing questions with previously-determined correct answers.

The good news?

There are powerful mental models you can use to strengthen your decision making without slowing down. One of the best? Probabilistic Thinking.

Let's dive in:

1. What is Probabilistic Thinking?

Probabilistic thinking is a strategy of estimating likely outcomes through the use of some tools of logic and mathematics. Or, more simply, it's an attempt to predict a range of possible future conditions based on the best available information.

Sometimes, a proper probability can be calculated if there's a great deal of available data. But more often, a much simpler approach is used which can factor in previous experience, known characteristics about the current problem, and some basic rules of thumb all to determine a best guess approach of how to proceed.

• Use some tools of logic and math to predict the most likely future
• Combine available data with relevant experience to guess well
• The goal is accuracy over precision​

2. How is it used?

When I was in graduate school, I had a great professor for my engineering statistics course and he'd always admonish us to start by looking at the data. Before trying to pick the right model to use in evaluation, look at what's plainly visible.

As pro poker player Annie Duke wrote in her bestselling book, How to Think in Bets: start with a logical framework, beware of taking shortcuts like being swayed by emotion, and make your decision as if you are committing all of your money on that bet. And, critically, create a positive feedback loop to improve your decisions as you learn more.

Not every decision you make as a leader requires hours or days of analysis, though. Consider the consequence of a wrong choice and calibrate your deliberation to match.

If you are wrong, will someone be hurt? Will the company sustain major losses? Can it be reversed? If the effort to roll-back the decision is trivial, evaluate quickly. If it's irreversible, then you need to go much deeper.

• Look at the data
• Think in Bets, but calibrate the effort to the consequences of the decision
• Track the results and adjust your strategies in response
• Learning happens everyday improving your ability to predict

3. Become a power predictor

Within this framework, there's a cheat code you can apply to almost any complex decision: think like an engineer.

As you are evaluating your options for action in response to a challenge, start by identifying what question can you ask or experiment can you run that will allow you to eliminate the greatest portion of the possible outcomes?

In other words, focus on what you can do to eliminate most efficiently the greatest portion of your uncertainty. Look for ways to prove something impossible or at least far less likely in the fewest possible steps.

• Narrow the possible outcomes with clear questions or experiments
• Look for historical examples or parallel problems for clues
• Leverage hard limits from the laws of physics to the limit of laws
• Ask better questions

4. And what about AI?

How does this intersect with the current explosion in Generative AI tools? At the core, many of the most exciting Gen AI applications are for responding with information, creating content, or summarizing text. The practical output of these models is based on a kind of multi-layer probabilistic thinking.

Each new unit of output is provided based on the model's best prediction of what should come next within a determined range of randomness.

When responding to a question, an AI chatbot is responding with one letter after another based on its model's choice of the most likely best next letter.

At a higher level, you can think of this as something similar to how a person might make a best informed guess as to the right response to an ambiguous question that must be answered. And, as you provide feedback on the quality of the response, that information is processed, and another new best guess is offered in response.

• Generative AI models respond with their version of probabilistic thinking
• Each portion of a response is driven by the model's best guess of what should come next
• Understanding this allows you to better direct the models on how to respond to get what you want from them
• The details are more complicated, but this itself is a useful simplification for improving your Gen AI results

Action Summary

Look for ways to apply this model to decisions both big and small. What would you do differently if you weren't looking for the single best decision, but one that would maximize the likelihood of an overall positive outcome?

How might you think in bets instead of in final decisions?

• The world is a complex place. Embrace the overlapping uncertainties
• Instead of looking for the answer, seek out the mostly likely good outcome available
• Be action oriented: don't get caught in analysis paralysis
• Become a better experimenter: what can you do to learn the greatest meaningful information in the most resource efficient manner?
• Leverage AI tech better now that you know more about how it works

As Mike Walsh wrote in Harvard Business Review:

​

What do you think? How does most likely and good enough thinking apply to your decision making? Reply to this email and let me know.

​

To your success,

Christopher

​

P.S. Like this one? You'll probably want to check out this earlier Wednesday Win post on how successful leaders create cognitive distortions to accelerate their success, too.

​

​

Looking for a deeper dive on these topics? Connect here and reach out:

1. Connect and follow on Twitter/X: @cbell​
2. Connect and follow on LinkedIn: @cbell​

Like this newsletter? Please share the sign-up link with others. Thank you.

​

​

​