Data Scientists Should Learn Through Play

If there is one thing that really annoys me nowadays, it’s when people look at something I am working on and ask me, often in quite ‘holier than thou’ tones: What is your use case? What is the problem you are trying to solve? As a trained McKinsey consultant, nobody knows better than I do the principle of having to define your problem up front, laying out a use case for the work you are doing.

But if you are learning data science, I think you should throw that principle out the window and take up a new one: Learning Through Play. Underlying this philosophy is the idea that — if your objective is to learn new technical skills — you should think up projects with no particular end objective in mind, other than the fact that you want to do it and you think you might gain something stimulating from participating in the activity. I’m going to argue in this article that the principles of Learning Through Play apply just as much to Data Scientists as they do to pre-schoolers. We will look at what the behavioral psychologists say about Learning Through Play and I hope you will join me in concluding that if you try it any other way you will miss out on important learning that you won’t find anywhere else.

The principles of Learning Through Play

To all pre-school children the world is new and they need to make sense of it in their early years. Most behavioral psychologists and educators agree that the most effective way of doing this is through play. It is through play that children take advantage of a safe environment to explore and test out the skills they will need as they grow — their social and cognitive skills, their emotional maturity and their self-confidence.

The book Einstein Never Used Flash Cards lists five characteristics of play which differentiates it from work:

  1. The participant needs to enjoy it
  2. There are no goals or prescribed learning
  3. It is spontaneous and the participant engages voluntarily
  4. The participant engages actively, not passively
  5. There is an element of ‘pretend’ or ‘make-believe’

Importantly, there are two critical differentiators between work and play. The first is who it is initiated by. Play is mostly initiated by the participant, never by a leader, co-ordinator or authority figure. And play is completely process-oriented with no defined up-front objective.

Outside of my work, I frequently take on personal learning projects to continue to develop my technical Data Science skills and to keep up to date with the latest technology. One of the many examples of this is when I recently completed a personal project where I worked on creating a network of characters from the TV series Friends so that I could analyze and visualize that network. 

I didn’t realize this so explicitly until I am writing it down now, but all my personal learning projects follow the principles of Learning Through Play. To use my Friends project as an example:

  1.  I knew I would enjoy doing it
  2. I had no specific end product in mind. I knew there would be an end product and I knew that I would learn new skills in doing it, but I had not defined what those would be up front. In other words, it was process-driven, not objective-driven.
  3. I thought of it out of thin air and proceeded spontaneously to tackle it
  4. I actively pursued it to a natural end point that I was satisfied with.
  5. It was not a real work situation and so had a ‘make-believe’ element that eliminated any risk or personal pressure outside of my own joy in making progress.

Why should Data Scientists Learn Through Play?

Like the world of a pre-schooler, the world of a Data Scientist is massive and at times overwhelming. Where do I start? What do I do next? How do I know I can do this? What might go wrong? But the more you play, the less massive those questions seem and the more confident you become that you can take on any challenge that comes your way.

Don’t get me wrong, I am not arguing against formal education and learning objectives in Data Science — these are absolutely still necessary especially for people starting out in the field. But I am arguing that great data scientists can only become great data scientists by Learning Through Play. Here are three reasons why:

  1.  Formal education can only cover less that 1% of all the tools, methods and possibilities that exist in Data Science today. At some point, you are on your own when it comes to discovering the skills that you need. That discovery is enabled by play.
  2. Knowledge and competency is only effectively gained and retained in someone’s mind if they have found a use for it. Given that our work and jobs are by their nature narrow, there is usually a limit to the knowledge and skills that are needed to perform day to day duties. So if we want to learn new and different skills, we need other outlets, other projects. Play projects.
  3. When we play with data, we are by definition enjoying ourselves. This creates a better environment for learning and retaining knowledge than when we are simply doing what we are told to do. We are more likely to learn when things work and don’t work, meaning we can more competently apply them in the future when the stakes are higher. 

Do you Learn Through Play?

I have absolutely no doubt that I Learn Through Play. All the new things I have used or applied at work in my recent memory have been things that I was exposed to through fun personal projects. If I had not taken on those projects, I would never have even known about the methods that I then applied in my work. 

But how about you? Here are some questions to ask yourself to determine if you are effectively Learning Through Play:

  1. Do I take on projects outside of work for my own personal enjoyment?
  2. Am I taking them on with a spirit of exploration and personal interest?
  3. Am I driving the thinking on how to attack these projects and make progress on them (and not merely blind copying something someone else has done)?

Learn Through Play has been my mantra for several years, though I’m only just getting round to articulating it now. I’m convinced it should be the mantra of all Data Scientists.

Leave a Reply

%d bloggers like this: