Joanna McKenzie is our Principal Data Scientist in the Data Science team. Here she gives her thoughts on Design Thinking in data science projects and the need for a Creative Data Science approach.
Design Thinking for Data
I’m a data scientist, and at the Data Lab I get to scope data science projects in a lot of different contexts and for a lot of organisations at different points in their data journey. Last year, I was heavily involved in running our “Innovation Week”, which, at the time at least, was a week-long immersion in Design Thinking (this year’s was a little different). As the data scientist in the team, my job was to be thinking always about how design thinking applied to a data science project.
Throughout the run-up to Innovation Week 2020, I was alternately inspired by the whole design thinking process and irritated by its apparent limitations for a data project. It just didn’t seem equipped to get a data project actually working. As I put it when I was feeding back my thoughts to the team at The Data Lab, it feels like the first stage is applying data to design thinking projects, and we captured that, but what we didn’t get was applying design thinking to data projects to really develop a Creative Data Science approach.
Bringing Life To Work
The third bit of context I’m going to provide to you here is that in my spare time I like to draw. I’ve spent most of the last ten years trying to build competence with shape and form, pen and pencil, colour and texture. But lately, I’ve been thinking a lot more about the process of producing a final art work. It’s an iceberg, you see — most of the lines or marks the artist makes never make it to the finished work. They’re in preparatory sketches, or they’re erased or painted over; perhaps they’re even in the skill building work the artist did to prepare them to make this artwork. As an amateur artist, you start with an idea of what you want, maybe try out a few things in your sketchbook and see what works, experiment with feeling, expression and atmosphere. Once you’re fully comfortable with all of that, you can start on your final work.
Is Data Science Creative?
Data Science is a science — it’s there in the name. Like science, we have tools and techniques that we apply to find something out — and the same approach applied to the same data will result in the same effect. Mostly.
But like art, data science is iterative, it relies on skills built up prior to the project itself, it benefits from an experimental approach, and applying the same techniques can sometimes surprise you. And data science is also an iceberg. What you get at the end is only the result of the process of building it, most of which is hidden.
How is Data Science like Art?
What has struck me recently, though, is that the parallel goes beyond that. See, a professional artist might often have to work with a client to develop their work. It’s not just their own idea they have to represent. Just as a professional data scientist is delivering projects that provide value for some business client, so too does a professional artist deliver artworks that reflect their clients’ perspectives, desires and expectations. The clients, for both professionals, will come with some ideas of what they want that may not work in practice, and the professional will have to manage that. It’s very much the same thing on both sides.
Now, producing artwork is not just a creative industry, it’s almost the creative industry — so emblematic of creative work that it’s the go-to example. If there are so many parallels between producing artwork and producing data science, then it stands to reason that exploring data science through a lens of a creative approach to showcase creative data science should be enlightening; there are things that we, as data scientists, should be able to learn from centuries of experience producing bespoke work for clients. And so we circle back to innovation week. Surely the design thinking process, which has grown from the creative industries, must have more direct applicability to the data science process than I originally thought?
Feasibility in Design Thinking
In looking at what I found dissatisfying with the design thinking process, my first instinct was to focus on the lack of any support or materials in the process in which I could take into account what the data would actually support. I looked at this as a technical feasibility angle — there’s no point in having ten people round a table suggesting that we should be predicting lottery numbers from darts scoring sheets if such a tool isn’t going to work in practice.
Certainly there’s a truth to this; but where it falls down is that technical feasibility is also a challenge in, for instance, product design, which thrives on the design thinking process. So it didn’t quite make sense that this should be a barrier for data projects when it isn’t for similarly technical challenges in other areas. That said, there is a clear difference that we already know from software engineering: people have an intuitive sense of what’s possible in the physical world that they often lack in the computing, virtual or mathematical worlds. The constraints of the physical world are part and parcel of how we live every day: if you tell people to think about what a new chair should look like they’re unlikely to ask for it to be made out of water or to support people upside down.
The constraints of a software or data product, on the other hand, are only really well known to those of us who work with software or data, and other people can and often do ask for things that are contradictory or totally infeasible. So I do think that a creative data science approach does require some new tools to enable people to get a handle on what’s actually achievable.
Design Thinking For Data
I also think, though, that there’s another limit on how well we can presently align design thinking with data projects, and that is simply that the data, particularly the data science, part of a project is actually only a subset of a broader effort. In a data science project I take an existing dataset and apply some exploration and algorithms and produce an output, and that’s basically the data science part.
But in order to get to that point, the data had to initially be collected, quality controlled, monitored and stored. And to actually get any value from my output, it has to be put in the position where a decision can be enabled from that output, either as part of a broader system or by putting it in front of a person. Data science is a part of that process and an enabler of that process, but if you try to apply Design Thinking tools just to that part, you’re likely to fail. Not because Design Thinking doesn’t work for data, but because you’re already part of the way into the project, and therefore you have already introduced constraints. The unconstrained ideation that’s so core to design thinking is a mindset that’s incompatible with a project that’s already half-done.
Working with what we’re given
The thing about all of this is that data science has emerged as a discipline essentially because organisations started finding themselves with lots of data or the ability to access lots of data, and wondered if all of this could be valuable. My background is in physics, and there, as in science more generally, datasets are very carefully designed to empower the process of answering the questions the researchers are considering.
But data science often isn’t like that. In data science we very often take erratic, undesigned datasets whose only real distinguishing feature is their sheer volume, and we try to use computing power and cleverness to build something valuable from that. It’s a useful process and can sometimes feel like spinning straw into gold, but it does remove a powerful tool from our arsenal — that of carefully cultivating a dataset that will provide the maximum insight into whatever it is we really need to know.
Making Creative Data Science Work
I’m pretty convinced that while data science has aspects that it shares with science, it also diverges from the scientific approach. While it’s got a lot in common with creative industries too, there are spaces where approaches that work there won’t work for us without some careful thought and some new approaches. And I’m fully convinced that data science as it is today isn’t yet what it will look like when it settles into a professional career — something we’re only at the beginning of, because only the younger data scientists have actually got any qualifications in data science.
What a fully developed creative data science approach has to look like is still in flux. Perhaps in the future we’ll have a design thinking department in large organisations who will scope out internal or external pain points and then commission a team of software and data experts to deal with the well-documented and identified problem. Whatever the solution ends up being, we’re still settling into our space in the world, and that’s one of the reasons I find it so exciting.
Read next: One of our MSc students gives his thoughts on soft skills: A Data Lab MSc Student’s take on data science soft skills