Skip to content
Home » Why you need Mental Models in Data Science

Why you need Mental Models in Data Science

I am back to the usual routine: morning coffee with Python, lunch with Machine Learning and dinner with Spark. If you are expecting me to talk about either one of these topics, you should stop reading because this week I am talking about Charlie Munger.

Charlie Munger, for those of you who have (still) not been introduced to his genius, is the long-time business partner of Warren Buffet. If by reading this, you are thinking this post is going to be about investing (again), let me tell you why you are wrong. Us humans think in associations. Warren Buffet is famous for investing so when I mentioned Charlie Munger is Buffet’s partner, our immediate thinking is Munger = Buffet = Investing. This is first conclusion bias, whereby we consider the most obvious associations to the facts presented in front of us and latch on to our “first conclusion”. The genius of Munger, and the topic of this post, is not investing, it’s Munger’s views on Mental Models.

What are Mental Models, you ask? Think of them as universal laws and principles from one part of our world that are applicable to another. Let me give you an example: did you know that Brownian Motion, a concept from Physics, is considered by some to be the foundational basis behind quantitative investing? I didn’t, until about a week ago.

So why am I talking about Mental Models. Why is this even relevant to Data Science? Let’s look at the following line: “you’ve got to have multiple models — because if you just have one or two that you’re using, …you’ll torture reality so that it fits your models”. Does that sound like the definition of overfitting to you? Do you know who I stole this quote off? One of the greatest Decision Scientists in the world today…tricked you again, Charlie Munger. Obviously, he said it in a completely different context, but if you’re a Data Scientist, you probably did not think about that. Instead you immediately applied it to your domain, and that’s a dangerous thing!

Still not convinced? Let’s pick up the concept of Social Proof from Psychology now. It says that we are constantly being influenced, both consciously and subconsciously, by what we see others do an approve. If everybody is buying into something, we think its better. Its proven by social science. Now stop here and google Machine Learning. You will probably come up with a lot of articles and company adverts (can I call it propaganda?) extolling its virtues. If you’ve worked as a professional in the industry, you know that the actual work is nothing like that (I haven’t but I’ve spoken to people who do). It seems that the companies are beings too (legally speaking, they are), they want to buy into something that everyone seems to be doing. As Munger puts it, “We don’t like to be the one guy who’s out of step”.

I’ll conclude by trying to convince you with a final model that Munger calls ‘surfing’. It is what inspired me to leave my world of human-driven financial analysis and start studying Data Science from scratch. If a surfer catches a wave, he stays there, without putting much effort of his own. If he gets off however, he is swallowed by the very same waves. The surfing experience is about catching the next, biggest wave possible because the biggest ones lead the furthest. Think of how Python, a relatively new language, has taken over R as the go-to-language for Machine Learning. Someone who thought they could surf the waves of R forever is probably rethinking the choice. For those interested in Data Science, it’s all about learning and at times even creating better technologies to keep up with the crowd.

If I have not convinced you on why Mental Models might be useful to make better decisions as a Data Scientist, let me tell you that 6 out of every 10 Data Scientist considers it an essential skill for the future. Happy Learning!

Read Munger’s 1994 Speech about Models here. Farnam Street also has a list of Models here.