There is A Lot Of Great Data Science Articles, But…
Casually reading data science articles in Towards Data Science has been my number 1 go-to site to learn new skills by many writers who share their data science journey. Many of the articles offer great straight-to-the-point content without extensive heavy technical literature reviews that often puts me in a state of confusion when compared to scientific journal articles like arXiv.
More often than not, many of the articles that aim to teach and inspire readers to pursue or do a career switch in data science cover the following topics:
- Which degree/advanced degree gets you further into the field
- Career prospects and typical job descriptions in the data science field
- Recommended programming languages
- Typical machine learning algorithms you should learn
- Recommended online courses to enroll for beginners and beyond
- How to implement algorithm X to solve Y
- New data science framework to spice up your data science process
These topics are absolutely essential when considering a job in the data science field. However in this article, I would like to highlight a certain soft-skill that is absolutely necessary for anyone working in the data science field: communication.
In the real world, data science is not just extracting required data from a database, wrangling a large set of data and dishing out machine learning algorithms to predict or classify given problems.
It's also business.
And when it comes to business, we need not only speak in technical terms with our data team, but to also be able to speak with internal stakeholders and clients in a language understood by all professionals.
In fact, communication has been encapsulated as one of the core data science skills according to KDNuggets in 2019. Based on the infograph alone, I would say that communication is pretty high up in the % want skill.
A Typical Data Scientist Job Requirement
Consider the following job description for a data scientist from Rolls Royce that resembles many other data scientist roles:
Apart from the standard technical data science skills and experience that appear at the top of the list, the highlighted soft-skills pertaining to communication is also an essential trait for data scientists.
So Why Tutoring?
From here on, when I speak about tutoring, I mean technical tutoring that encompasses the STEM(Science, Technology, Engineering, Math) category.
My Personal Experience on Tutoring
When I was in my undergraduate period, I took up an internship as a coding instructor for kids with First Code Academy. I have taught many programming languages including Python, HTML/CSS/JS for basic web programming, C# in conjunction with Unity for basic game development and block-based programming like MIT AppInventor for Android and Scratch.
“What? Coding for kids? Isn't it too difficult for them?"
Yes, it may sound difficult because most of us learned it at least during our late teens or early adulthood. We think that coding is a difficult task that can likely be comprehensible by adults with developed intellect. I picked up programming when I was in my 20s during my undergraduate period.
After getting comfortable with programming basics in my freshmen year, I decided to partake in this internship and boy did I learn a lot in this short period of time!
1. Mastering the concept of ELI5
ELI5(Explain like I'm 5) is Reddit's community solution to many hard-to-explain questions and basically the community tries to “dumb it down” enough for the one asking to understand.
When I was teaching the younger students around the age of 7–10, they do not have a very developed cognitive ability to absorb common programming practices such as variables, loops and conditionals. Teaching these require extensive use of visual aids for them to connect the dots.
Translate this to data science and we speak of data visualization and model explanation that addresses their business problem.
1.2. Data Visualization
When we present to internal stakeholders and clients, we need to assume that their mind is a blank canvas, and we are artists responsible in painting a clear and defined picture in their heads. This is especially true when you are the only one responsible for the project at hand. Nobody knows what and how you did it.
Simply throwing in a couple of key metrics in tables and simple charts won't cut it. Designing interactive executive dashboards and guiding them through it step by step is part an essential trait for many data scientist.
I say many because this may be a job given to business intelligence analysts instead.
In any case, being able to show your beautiful visuals and answer their business requirement is crucial for your role, unless you're a research academic, then maybe it is not as important.
1.3. Model Explainability
Explaining your predictive model based on your algorithm choice is also important apart from hitting the KPI predictive target. Apart from making good predictions, stakeholders may also want to know what variables are responsible for the sensitivity of the prediction.
This is where your choice of ML algorithms may come into play depending on who your end-user is. If accuracy is of utmost importance, then you can consider complex methods such as NN or ensemble methods. Otherwise variable explanation is more suitable by using simpler algorithms such as multiple linear regression or logistic regression.
However, with the current state of ML frameworks, complex models such as random forest(which is also my personal favourite) have the feature importance function that easily visualizes the importance of each feature pertaining to the model which makes it a great source of information for business managers to look into.
2. Strategic Communication
Teaching someone something new and technical is definitely a challenging task, especially when it comes to dealing with numbers and rational concepts. In every lesson I conducted, I noticed that the students are most attentive in the first third of the lesson(each lesson is 90 minutes long). After that they become restless and they start to have difficulties in following the lesson.
This applies to me as well when I attended lectures during my undergraduate period and I'd like to think that many working adults do not have a very long attention span either.
The human brain is not designed to utilize 100% of its memory-forming neurons as shown in this article.
Taking up tutoring has taught me how to create report and presentation content in a strategic manner that maximizes in capturing the attention of the reader or the audience.
Different people will have their own strategies but essentially knowing when to spice up your content when it gets a little dull is key to making a great impression about your work. This means less text and more visuals and knowing how to story-tell your visuals. You can also cook up creative analogies that resonates with the results you are trying to explain.
For me, I use the interactive capabilities of Tableau to my advantage during the communication of results to my stakeholders.
3. Empathy and Patience
“Can you explain this to me again?”
“Oh haha, I completely forgot about it”
They said, with those innocent puppy eyes. These questions come up to me almost every lesson and at first it is a little daunting to have to repeat the explanation again and again.
But then I realized, I am the same, which is why I was not a perfect academic to begin with!
And then I realized, while working, there are people like me who forgets pretty easily. Of course this is not a surprise. Many employees are expected to multi-task with their work and when one switches off their focus from your work for quite some time, they are bound to forget.
So instead of feeling frustrated having to repeat yourself, being exposed to teaching may instill empathy and patience within you that makes you go: “Sure I'll be glad to explain it to you again, here's how it goes…”
I made use of my experience in teaching programming to kids as an example of how you can build your communication capabilities in the data science field. In fact I believe that it has tremendously helped me with my interviews where I have to structure my explanation of my past work experience and projects in a way that the hiring manager can understand without throwing in too much technical jargon unless the hiring manager is technically inclined.
But it doesn't mean you should follow this path, I'm sure there are other ways you can do to supercharge your data science communication skills to the next level.
One way is to consider doing your own passion project and share your findings with your friends. If they enjoyed it, then I'm sure you did great
I hope this article has somehow helped readers improve their communication abilities in the field of data science!