Reflections from Thinking Fast and Slow — On Intuition and Expertise

Is Intuition Actually a Good Thing?

Mukundh Murthy
6 min readSep 2, 2020

Have you ever thought about how much you’re truly in ‘control’ of? Humans think of ourselves as rational creatures, with full or at least majority control over the decisions and judgements that we make.

Daniel Kahneman’s Thinking Fast and Slow sheds a new light over this common assumption. Kahneman has puts forth the idea that we are made up of two minds, one ‘fast’ and one ‘slow.’

Our fast mind falls prey to certain heuristics and substitutions — that is, ways of thinking that fail to represent reality as it actually is. One simple example is the availability heuristic. We are constantly exposed to what’s portrayed by the media; therefore, we tend to propose higher than accurate probabilities for infrequent occurrences like car and plane accidents. We substitute the magnitude of emotion that we feel at qualitative level for more quantitative and factual objective-based answers.

On the other hand, our slow mind is the one that often experience cognitive dissonance with our fast mind. It tends to question the fast theories, corollaries, and assumptions cooked up by the fast mind. It’s the voice in your head that challenges your first answer when you look at a riddle or a puzzle, telling you that the most obvious answer isn’t going to be the right one, as it’s likely a trick question.

I’d encourage you to take a look at Kahneman’s book directly for a more thorough description of the fast versus the slow mind (aka System 1 versus System 2).

Here, however, I’d like to specifically meditate over the ideas of intuition and expertise, especially as they relate to the current global situation and thoughts being put out by other authors and thinkers.

Difference between System 1 (fast) and System 2 (slow)

Why I’m writing about this and why it matters.

During this global pandemic, innovation and expertise seem paramount to moving ourselves out of this dire situation. I’ve been amazed to see bioinformatics and cellular biology combine with machine learning, distributed computing, and quantum computing in this global fight.

Here’s the main idea I’m currently grappling with:

  • Does intuition increase or decrease the frequency of breakthrough advancements over gradual improvements?

What is Intuition?

As introduced by Kahneman, intuition can be thought of as a cognitive response or state, triggered by environmental or knowledge-based cues that trigger activation of your associative memory.

The associative memory is the part of your memory built through repetitive experiences with similar cues and similar responses. Essentially, it’s a pattern based activation system.

Therefore, intuition requires a couple main components:

  • A set of repetitive components in the system
  • A set of responses (right/wrong answers) to be paired with the cues
  • The ability to perceive the association between cues and right/wrong answers.

This is essentially what intuition is at its core. One plausible reason for why intuition my sometimes seem magical is that over decades, someone’s associative memory may subconsciously detect certain surrogate and accompanying variables that he/she doesn’t constantly recognize. In other words, a doctor who seems to magically do a hard diagnosis on the spot accurately may be unconsciously pulling on other variables through his/her associate memory that help in that specific moment. Again, the reason this intuition seems so magical is because the doctor doesn’t consciously know what those variables are.

The real question to ask here is how many times is enough to perceive the patterns accurately in a system? Do I need to see examples five times, ten times, one thousand times?

The answer here may be that we don’t really know. The world displays stochasticity — in essence, there’s a certain amount of randomness occurring in our world due to probablistic sampling of certain outcomes.

Therefore, there’s a great chance that a pattern or trend that we deduce, even over a seemingly long period of time isn’t really a pattern or trend. Kahneman calls this What You See is All There is aka WYSIATI.

WYSIATI is inevitable in our world, as we will never see all the examples for particular concept. Therefore, there’s a risk of building intuitions over a large period of time that have a small applicability domain–In other words, these intuitions would be accurate for a small amount of examples but not for the examples that we haven’t seen yet. I would allude this to the problem of generalization in machine learning. Models are very good at fitting their training data, but often unable to extrapolate to datasets outside the domain used for training.

Following the machine learning analogy, the solution invented by ML pioneers has been dropout — extending back to our use case, preventing the formation of specialized intuitions. Phrased another way, polymathy is what dropout would look like when extended back to human learning and pattern recognition.


In a recent The Daily Experiment podcast episode with Kevin Doxzen, a PhD student out of the Doudna and science communications specialist at the Innovative Genomics Institute, we discussed a recent blog post of his on The Role of Experts in Society.

We came to the idea that Expertise scales somewhat exponentially with knowledge and that knowledge=information + experience.

The equation above implicitly states a very important idea. Knowledge is not the same as information. Information can be accrued by reading multiple articles on the web, whereas the transformation of that information into knowledge is achieved through experience, which means applying that information in a practical setting.

By connecting back to the section on intuition, we can see that intuition derived from information alone can be more misleading, as experience can give the surrogate and associating variables that take decision making into a much higher dimensional space.

However, here we experience an inherent contradiction with two arguments going against each other.

  • Expert intuition can may be more accurate in certain cases due to larger amounts of experience for spotting applicable patterns, though they may be misleading in cases which require extrapolation.
  • Polymathy may be better for innovation. It allows our system one to develop associative coherences that span a large variety of domains. However the conversion of information into knowledge through specialization will be much harder for polymaths (it’s objectively harder to gain an equal level of depth for a larger variety of fields).

The answer to this contradiction, in my opinion, is that we need both experts and polymaths. Experts for the situations that are most like the situations we’ve seen before, and polymaths for the situations that we’ve never seen before to explore newer paradigms.

The Downside to Polymathy — an analogue of Underfitting

In both Kevin’s blog post mentioned earlier and Mark Manson’s recent mention of ‘The Death of Expertise,’ authors mention that larger groups of people are starting to distrust experts due to the larger amounts of information and articles online, and an ‘illusion of confidence’ given what they’ve read and heard.

The benefits of polymathy include a larger probability for groundbreaking insights due to a lack of associative constraints placed by too much experience. However, polymaths must understand that being a polymath means being wrong more often.

I know I through a lot of ideas in this article, but here are some key points and conclusions that I hope you’ve been able to derive.

  • Intuition can be misleading sometimes due to WYSIATI
  • Experts and Polymaths are relevant in different situations.
  • Being an expert means having more repetition to rely off of but being more prone to WYSIATI
  • Polymathy involves a higher probability of being wrong in a specific situation, but better ability to extrapolate.

Hey! I’m Mukundh Murthy, a 16 year old passionate about the intersection between machine learning and drug discovery. Thanks for reading this article! I hope you found it helpful :)

Feel free to check out my other articles on Medium and connect with me on LinkedIn!

If you’d like to discuss any of the topics above, I’d love to get in touch with you!(Send me an email at or message me on LinkedIn) Also, feel free to check out my website at

Please sign up for my monthly newsletter here if you’re interested in following my progress :)




Mukundh Murthy

Innovator passionate about the intersection between structural biology, machine learning, and chemiinformatics. Currently @ 99andbeyond.