From Social Science to Data Science
From Social Science to Data Science
8 years ago I started my bachelor's degree in Geography. Now I'm a Data Scientist; this is the story of why I've gone down this path, and why I think data science has a lot to learn from the humanities
I n my current team of ~60 data scientists and ML engineers, there are only ~5 of us who didn't major in a STEM (Science, Technology, Engineering and Maths) subject at university.
This ratio is pretty typical across the tech industry – in 2020, the annual StackOverflow Developer Survey found that only 3.8% of professional developers majored in a social science or humanities subject at university.
This isn't necessarily bad or surprising (it's natural that Data Science attracts people who liked STEM subjects at school and university). But STEM isn't the only route into the field, and I think it's important for aspiring data scientists not to be discouraged if they don't have a traditional maths/CompSci background.
STEM isn't the only route into data science
In this article, I'll share the story of how I broke into data science from a social science background.
If you're an aspiring data scientist, I hope this encourages you in your journey and provides practical advice on how to make the career change. If you're a hiring manager or fellow data scientist, I hope this helps you appreciate the value that the humanities can bring to our discipline (hint: it's more than just AI ethics!).
From postmodernism to p-values
I've been working in data science for a couple of years. Here's a quick overview of how I got here:
- 2013–2019 – While at school/university, I did a bunch of part-time jobs ranging from McDonald's to call centres to paper rounds. I was amazing at flipping burgers at Maccies; I was less good at waking up at 5:30am for my paper round
- June 2019 – Graduated with a BA in Geography from Cambridge
- Sept 2019 – Started work as an Analyst at Vodafone, eventually moving into the strategy team
- Oct 2021 – Started my MSc in Social Data Science at Oxford
- Sept 2022 – Started an internship/short-term contract with Rewire's data science team
- Jan 2023 – Started as a Data Scientist at Sky (where I still work)
As you can see, my career took a few twists and turns along the way.
When I was deciding which subject to study at university, I didn't even know data science was a thing (much less that I would enjoy it). I left school believing that maths/CompSci were for "nerds," and wrote myself off from ever working in a highly numerical profession.
It took me a few years to stumble across data science and realise that it was a great fit for me. But when I look back on my time studying Geography at university, I don't think of it as a complete waste.
Quite the opposite.
The advantages of coming from a social science background
Appreciation of psychology
When you're knee-deep in a petabyte of data, it's easy to forget that there are real humans behind the patterns you observe.
Somewhere along the line, you stop referring to people as people and start using corporate buzzwords (people become "users", "prospects" and "customers").
But, regardless of what we call them, they're still people!
And one of the great things about the social sciences and humanities is that they give you an appreciation of the human psychologies and behaviours which drive a lot of patterns we see in our data sets. I think this is a real superpower in corporate data science contexts where techies are so far removed from the end user (person).
Social scientists understand the value of qualitative and multi-method approaches
In data science, we tend to emphasise the value of quantitative approaches.
So much so, in fact, that we sometimes forget qualitative approaches are even an option!
Part of this is due to an organisational division of labour (in large companies, qualitative research is likely performed by other, non-DS teams), and part of it is cultural (data scientists typically learn their craft via impersonal online courses, computer suites and lecture halls, not through interpersonal fieldwork and interviews). We default to quantitative approaches (our comfort zone) and neglect the rich insights that can come from qualitative ones.
I'll give you an example.
When I worked in a call centre several years ago, I developed an "upsell propensity model" which aimed to identify the customers most likely to purchase new products. I made the model for entirely selfish reasons: I was required to make make at least 10 sales per day, and this usually required making about 200 outbound calls. If I could make the 10 sales quicker, I could go home quicker!
The model itself was completely quantitative: matrix of numbers in… matrix of numbers out… you know the rest. But my inspiration for the model came from a lunchtime conversation with a friend in a different department. My friend was complaining about a particular group of customers that was always asking for upsells (he hated this group because, as a customer service agent rather than a sales agent, his main KPI was call handling time, so making upsells to these customers made his calls take longer and made him look worse in the stats).
Our conversation quickly descended into an interview (poor him) and gave me the idea to develop a predictive model. The model worked: it boosted my sales rate by 50%, and meant that I got to go home at 4pm instead of 6pm (this was a huge win for 18 year old Matt). But the original spark came from that "qualitative" moment of data collection, not quantitative analysis of the numbers.
Social scientists are often excellent at this type of data collection and analysis, and again, this can be a real differentiator / superpower in a data science environment.
Critical thinking
Social science is fantastic at teaching critical thinking. In my bachelor's degree, we studied everything from environmental politics to "critical museum studies" (yes, that's a real thing!), and it trained me to think critically about new subject areas even when I wasn't super familiar with the content.
I'm not saying you can't learn critical thinking in other disciplines, but I'd be surprised if the critical thinking skills you learn from Foucault are the same as the ones you get from doing Fourier Transforms.
In my day job, these skills help me identify problems and ask the "big questions" about what we do, rather than getting prematurely bogged down in implementation details.
I'd be lying if I said it wasn't difficult
So, what are the hard things about getting into data science from a non-STEM background?
First, there's the feeling that you're constantly playing catch up, or that you've missed out on too much.
While my present-day colleagues were practicing Fourier Transforms, I was reading Foucault. While they were learning about Multivariate Calculus, I was learning about Marx's view of Capitalism.
I wouldn't trade those experiences for anything, but I'd be lying if I said it didn't sometimes bring a feeling of insecurity and imposter syndrome.
"How can I possibly catch up on all that maths?" I think to myself. Or "how can I possibly get a job in Machine Learning – the other candidates have been doing maths Olympiads since they were 5!" (slight exaggeration).
While it's true that there are things to catch up on if you're not coming from a STEM background, it might encourage you to know that, in my experience, the maths requirement is a lot less than you think and is easily learnable via free courses on YouTube (unless you want to be a Research Scientist at Meta – then you need to be a maths boffin and put in some serious groundwork!).
My mate Egor has written a helpful article on how to learn the maths required for data science:
It can be a bit… boring
Another big challenge in moving to data science from a social sciences background is that it can be boring.
Yep, I said it!
In the social sciences and humanities, nothing is taken for granted and everything is up for debate. In the data sciences, everything is taken for granted and nothing is up for debate.
When you're working in a demanding trading environment, there's no time to discuss the subjectivity of different truth claims in our data sets. There's rarely time to think about the big questions, and the demands of the day usually win out. If you're used to operating in a humanities-type academic context of lively debate and big picture thinking, this can be frustrating.
Things that helped me make the transition
I'm not writing this article just to reminisce about my own journey. I also want to help people. Specifically, to help YOU (if you're thinking about getting into data science).
Here are some of the things that helped me.
Studying a master's degree
A master's degree in Data Science isn't for everyone (and it's definitely not a hard requirement for getting a job), but it helped me learn a lot of new skills in a short space of time and gave me the confidence to start applying to more jobs.
If you're considering doing a master's, please don't do it until you've properly considered the other options. Postgraduate degrees are expensive, and for many people they aren't necessary. To help you make the decision, I'd recommend this article by Khouloud El Alami:
I Spent $96k To Become a Data Scientist. Here Are 5 Crucial Lessons All Beginners Must Know
and this one I wrote last year:
8 Things You Must Consider Before Committing to a Data Science Master's Degree
Personal projects (you can do the thing!)
I'm a big fan of data science portfolios. For me, the main benefit of making my portfolio was psychological – it helped me realised "you can do the thing!" and gave me the confidence to apply to some jobs. If you're stuck for project ideas, you can take a look at my portfolio or read this article I wrote last year:
How to Find Unique Data Science Project Ideas That Make Your Portfolio Stand Out
Find people like you
We all need people to look up to. I took great inspiration from finding and following some fellow social-scientists-turned-data-scientists:
- Chris Albon – from political science PhD to ML Director at Wikimedia
- Carson Forter – from humanities PhD to Data Science at Twitch
- Cassie Kozyrkov – from psychology and stats to Chief Decision Scientist at Google
If you look around, you'll find many more. I really recommend finding and connecting with people from your background; it will help you to believe that it really IS possible to make difficult career transitions.
Conclusion: Would I recommend it?
When making decisions about what to study at university, a lot of people (myself included) are driven by the desire to make a difference. We pick subjects we're passionate about, but we don't really think about how we'll actually make that difference when we're finished studying.
That was certainly true for me – I picked Geography because I loved the subject and was passionate about making a difference to the lives of people less fortunate than me. But I ended my bachelor's degree with few hard skills and very little ability to actually make tangible progress towards those goals.
The nice thing about data science is that it is a tangible and useful skill. For this reason, I highly recommend it to my fellow social scientists. Learning data science is a fantastic way to equip yourself with the skills you'll need to make a difference.
If this is you, I'd love to connect with you on Twitter or LinkedIn! Let's chat
