In the book "Thinking, Fast and Slow" by Daniel Kahneman, I'm exploring specific chapters that shed light on how our brains work when making decisions. In this article, I'll focus on the key ideas from these selected chapters. We'll dive into quick thinking, slow thinking, and when to trust your gut. If you're curious about how your mind works and want practical insights into better decision-making based on these specific chapters, stick around. Let's explore it together.
In this book, Daniel describes two modes of thinking.
System 1: This is the fast mode. It provides output very quickly using heuristics. It is designed to provide quick response in dangerous situations. For example, we run when we see a lion running towards us. It does not engage in detailed calculations but quickly processes environmental signals to produce outputs. For this system, whatever is quickly available is all that matters. Our mood information is readily available. So, any decision made by system 1 is likely to be affected by our mood.
System 2: This is the slow mode of thinking. It engages in detailed calculations to process the inputs. It also does complex searches in the memory to find relevant data. It is, however, lazy. If system 1 can answer the question, system 2 will rest in the default mode.
Clearly, the intuitions are the results of system 1. The two systems interact with each other. System 1 can provide starting points for calculations to system 2. Many repeated calculations from system 2 can be effectively done by system 1. So, now the question is: Can we trust the intuitions made by an expert? If so, when?
Intuition vs. formula
We start with the fields where the experts don't do so well. The critical property of these fields is that they are low-validity environments with a high degree of uncertainty and unpredictability. Here are some examples:
Evaluating the future performance of a student
Evaluating the future performance of a large business over the long term
Predicting the winner of an upcoming football game
Assessing the suitability of foster parents.
The list goes on. Experts in these respective domains can provide intuitive answers. But it turns out that these answers are outperformed by simple formulas. This holds true even when the formula considers fewer metrics to evaluate the solution than a human expert. More surprisingly, even when the formula outputs are available to human experts, the humans still fail to match the accuracies. They override the formula answers too often, which reduces their accuracy.
One key reason humans do worse than the formulas is that intuition is inconsistent. Recall that system 1 outputs depend on the environment in which we make the decision. As per an example given in the book, expert radiologists contradict themselves 20% of the time when classifying an X-ray image as "normal" vs. "abnormal". Unreliable judgments can't be valid predictors of anything.
In such low-validity environments, we should instead rely on formulas. A simple process to create a formula that will outperform human experts is as follows. Take 5-6 key features for the question you want to answer. Find a way to measure those features. Give them a scaled score of 1-5. Simply add the scores or perform a simple linear regression to get the final outcome.
When experts do well
Expert intuitions are sometimes good. For example, a trained firefighter will have an excellent intuition about what to do in certain emergencies. An expert chess player in classical games can come up with high-quality moves for a given complex position in a few seconds. These intuitions are far better than intuitions made by average humans. These scenarios are complex enough for expert intuition to outperform a simple regression formula with 5-6 features.
So, what is so special about these domains? These are high-validity environments with the following two properties:
- They are regular enough to be predictable.
- There is an opportunity to learn these regularities through prolonged practice.
An expert chess player has undergone rigorous training of over 10,000 hours or more. During this training, they have seen many patterns. They receive quick feedback on their action. The faster and unambiguous the feedback is, the easier it is to master the skill. Expertise in a domain requires mastery of many mini-skills.
An intuition is nothing more than pattern recognition. This is why it works well in environments where there are high regularities.
Unfortunately, the confidence of an expert is not a good metric to evaluate how good their intuition is. The confidence is an output of system 1. When we are at cognitive ease, for example, when a story comes easily to mind, with no contradictions, we feel confident. This has nothing to do with the accuracy of the output.
A true expert knows the limit of their knowledge. They know that certain areas in their domain of expertise are low-validity environments. We can use these guidelines about the validity of the environment to figure out when to trust our own intuitions as well.
Video: The Universe is Hostile to Computers (Veritasium) (This is how to justify hardware bugs)
Quote: “Everything that is really great and inspiring is created by the individual who can labor in freedom.”— Albert Einstein