Mental Model Mondays #1: Overfitting

Hi friends, hope we’re all keeping well.

Ah yes, another new series, but this time I swear I’ll be consistent with it, since I think I’ve finally found out a way-just have shorter posts🙃.

Anyway, this series will sort of be a successor to a previous one I had, called ‘Mental Models That You Should Know About’.

For those of you who aren’t aware of what a ‘mental model’ is, Wikipedia gives a top notch definition which I shall now CTRL C, CTRL V for you now:

“A mental model is an explanation of someone’s thought process about how something works in the real world. It is a representation of the surrounding world, the relationships between its various parts and a person’s intuitive perception about his or her own acts and their consequences.”

A really talented author at Wikipedia (Biogeographist)

In more layman’s terms, they’re different ways to think about different things, in an advantageous way that gives you a clearer sense of what’s going on.

Some mental models I covered in my previous series included anchoring, survivorship bias, conjunction fallacy, etc. You may have heard of some of them.

Anyway, my plan is to post a new mental model every Monday (hence the long ass name), and see where it goes. I find these things to be quite fascinating to think about, since psychology is by far the most interesting science.

Alright, I assure you all future Mental Model Monday (you know what, I’ll just abbreviate it to MMM) posts won’t be this long. With introductions and summaries out of the way, let’s have a brief look at a model that even I’ve had trouble wrapping my head around-overfitting.

---

💭Concept 💭

Overfitting is a concept really only used in statistics and machine learning, but it can also be used to understand many real life situations as well.

In statistics and machine , overfitting occurs when a model tries to predict something, based on noisy data (basically, meaningless/corrupt data that can’t be interpreted by computers or machines). If a model is overfitted, any results it produces are inaccurate because they don’t reflect the real data.

To say it another way, the main problem with overfitting is that the machine has basically memorised the data we have inputted, instead of actually doing its job and trying to predict how unseen data would look like.

I understand that still may sound confusing, so let’s look at some real life examples of overfitting.

---

🧐Examples🧐

-In an example that applies to myself, let’s think of overfitting in an exam context. Say I’m studying for an exam by ONLY doing practise questions from previous years and memorising them, instead of learning the general concept of HOW to tackle these questions.

Obviously, this would not bode well in the actual exam, since the questions would all be brand new and I would then fail miserably.

This is exam overfitting, as I’ve essentially memorised meaningless data (the answers to previous years’ questions), and tried to apply them in a real life scenario (in the actual exam), and surprise surprise I’m now screwed because I didn’t learn the GENERAL method of HOW to do the questions.

-In a much simpler and probably relatable example, let’s say you have a fever and a cough at the same time, but don’t want to spend money by going to the doctor.

Upon googling these two VERY COMMON symptoms, you’ve now concluded that you have 10 different types of cancers and have 3 weeks left to live. In this case, you are over fitting your symptoms, when you should instead be thinking more generally and what’s most probable.

-One final example of overfitting is in relationships. After having one bad partner and then concluding that that entire gender is evil and trash is dating overfitting. Can you see how if you fail to generalise things, you can fall victim to potentially absurd conclusions?

---

Basically, overfitting occurs when you use an overly complicated explanation when a simpler one will do, and also when you use bad data to try predict a trend.

It’s by far one of the more complicated mental models that may take a few examples to fully grasp, but it’s also a really interesting one, in my opinion🙃.

---

Cool, that’ll probably wrap up this MMM post then. If you’ve any questions or even feedback for this type of series, I’d love to hear it and share a discussion!

Great, I’ll hopefully see you guys in another post then, stay safe 👋🏼