Data Science of Learning

If you're following my endeavor of completing the Harvard Edx Data Science certificate online you know why I'm less "social" out there recently. It is a tough cookie. Refreshing my stats studies from two decades ago and applying it through R at the same time is a challenge. I've learned to make informed predictions on all kinds of data from the Titanic through insurance industry or elections...

But I'm not going to bother you with data. This is about learning itself. I wanted to share how important it is to unlearn and learn how to learn adopting to the performance expectations and evaluation. I'll share my mistakes I've made in case you're planning to take on some of these challenges.

Mistake #1: Underestimating the time it takes to complete a course

There are seven courses included in the program. I thought it would be breeze to get through seven courses. I'm sure you've taken a lot of online courses with or without assessment (mostly multiple choice). This Data Science program has theory and practice combined. You learn from an overview video, consult the workbook with worked examples, and then there's the practical part: actually code in DataCamp.

This is where I underestimated how long it takes to get things done. The practical application of the theory (that in theory I know) is hard. Learning a programming language and new concepts in stats together sometimes make my head spin. As with everything new, the mental model you hold about things don't fit. A mental model is how WE think something should work based on our experience. When we run into a discrepancy, it slows us down (may even lead to frustration). Simple things like R counts from 1, and every other language I learned counts from 0. One little number can make big problems!

Unlearning how "things should work" is a painful process. But that's the only way to keep up with the change happening around us.

Mistake #2: Not using resources the right way

There are multiple types of resources you can use during the course:

1. Concept videos by professor Irizarry (btw, I set the video to 1.5x speed)
2. Workbook with details and worked examples (always open and I do a lot of search)
3. R studio (actual R coding platform you download and play with)
4. DataCamp exercises (interactive platform to write R code and get hints)
5. Forum (asking others or staff)
6. Google (for anything else)

How to use these resources is something I learned through trial and error. Some of these mistakes, again, relate to the mental model of how I learned/worked before.

So many eLearning courses require you to memorize facts, and then spit out the answers by correctly recognizing the longest option in a multiple choice test. Often these questions are set in a linear order, so you can't go back and forth once you answered them.

This data science course have very few multiple choice questions rather open-ended research questions that you need to answer by using data. Like in real life, there's no multiple choice options. Therefore, using resources available you need to do a lot more work than watch a short microlearning video.

During the first course, I only looked at the forum after I struggled through the exercise. Simply, because that's how I assumed it works. Then I realized that reading through the questions and discussions BEFORE taking on the challenges help me to avoid misunderstandings, pitfalls, and confusions. Another thing I learned from reading others' comments is that sometimes I had to go back and relearn a concept. Whenever I didn't even understand what another user were asking, I knew that there's some gap in my knowledge.

Using others (and staff who are very patient and helpful) is a resource we all incorporate in our every day work. Learning shouldn't be just a content on the screen. You should know people who ask those questions you don't even understand. That's how you grow.

Mistake #3: it is much easier to criticize

Yes, I'm working on this skill but time to time I fall into the trap: criticizing how things work, how it was designed, how it doesn't do what it should be... If you are reading this on LinkedIn you probably know this but it's so much easier to criticize what others posts than creating and sharing on your own. There's always limitations, circumstances, time, budget, management, egos, politics... Criticizing a solution should also come with a better alternative that is feasible within the same constraints.

DataCamp has its pros and cons. It allows you to write code and test your actual knowledge, rather than just theoretical understanding. At the same time, grading can be limited. There are various ways to get to the same result. One thing is sure. Designing and building a solution for a problem most of the time lands inbetween doing nothing about it and coming up with the perfect thing.

Start where you are. Use what you have. Do what you can.

Conclusion

I'm half way through the data science program but all the way through my learning how to learn within the framework. I wasted several hours due to ineffective methods. I'm still not perfect at data wrangling and way to slow compared to machine learning. But I'm unlearning and relearning every day in my human way. And that counts.