AI and Design

How product designers and data scientists can (and should) work together

Rosarie O'Regan
Oct 22, 2018 · 7 min read
Image for post
Image for post
Illustration by Not Flipper

Zalando uses artificial intelligence technology to help “reimagine fashion for the good of all” by analyzing shopping and search trends to create personalised shopping experiences. For instance, we might recommend specific items based on past purchases, a new brand based on recent browsing history, or a size based on previous purchases.

While using AI for data analysis is an excellent tool, it can sometimes neglect the more human side of personalisation. Fashion is fickle and user preferences frequently change as life circumstances evolve: college graduate to professional, festival goer to new parent or a person who normally only shops for corporate clothing on Zalando and is occasionally interested in more casual looks, to name just a few examples. Basing recommendations on past behaviours does not always guarantee success…what a user bought last month may not accurately reflect what they want or how they feel about fashion today.

Basing recommendations on past behaviours does not always guarantee success…what a user bought last month may not accurately reflect what they want or how they feel about fashion today.

The potential implications of AI and machine learning are enormous and quite frankly largely unknown. As designers, we must consistently ask ourselves how we can design for the visceral in our AI enabled products. We can make smart machines, but in the end they are still just acting like machines. Given that fashion is very personal and therefore evocative, how can we expect machines to make good decisions without factoring in a more emotional side? Or as Danielle Krettek put it “It’s basically time for technology’s EQ to equal its IQ”. Until more emotional AI is a reality, it is critical that designers work side-by-side with data scientists to help build machine decision making processes that considers this emotional layer.

Until more emotional AI is a reality, it is critical that designers work side-by-side with data scientists to help build machine decision making processes…

I’ve often found myself wondering if our tried and trusted design processes will continue to hold up in the face of AI as we collaborate more closely with data scientists. And if not, what should we change?

The 4Ds Process

Here at Zalando we use a design process referred to as the 4Ds: discover, define, design and deliver (loosely based on the Design Councils’ “Double Diamond” seen below). Our model is comprised of an iterative process that uses both divergent and convergent thinking to first discover what problem it is we’re trying to solve, and only then going on to deliver solutions that ultimately improve the customer experience. Below I discuss each step of the design process and outline some practical ways data scientists and engineers can (and should) work alongside designers.

Image for post
Image for post

Discover

Over the last two years, I’ve learned to adapt our design process to better integrate with machine learning teams. One of the most important lessons I’ve learned is the need to take a step back from our “designer bubble” for a moment, make note of our own biases and ensure that we have empathy for our colleagues — their backgrounds, assumptions, and ways of working. For a team to work well together, every member must make the effort to empathize with each other, before collectively attempting to understand our customers.

Meet each other halfway

Data scientists and designers often have different backgrounds, terminologies and processes, making it both intimidating and exciting to work with each other. Designers must be proactive in learning about and understanding AI and ML concepts.

To get started with the basics of ML training see these AI resources for creatives, this summary of Machine Learning for Designers or Sam Drozdov’s nice introduction. Want to learn even more? Have a listen to the Machine Learning Applied podcast series or sign up for the free online course on elements of artificial intelligence.

Of course the same holds true for data scientists, as they may not be familiar with the concepts of UX and design. We should take the opportunity to offer training on the design process, explain user research methods, discuss the advantages of prototyping and of course talk about why it’s so important to understand how each other work. Taking the time to do this will make the entire project experience more pleasant for everyone, while simultaneously increasing the design maturity of the entire organization!

Involve data scientists in research activities

Including the data scientists and data engineers in the discovery process is one of the easiest and most enjoyable ways to expose team members to the design process. Share out research insights and make them tangible and visible to the entire team (use photos and video clips, persona posters). Also, the engineers’ mathematical mindsets can come in quite handy when using quantitative research methods!

Image for post
Image for post

Define

Help the team stay focused on the user problem

In engineering-heavy teams, it’s often easy to neglect consideration of the users’ needs. Designers can help ensure the team doesn’t fall into the trap of putting tech before the user. As amazing as AI is, it can’t tell us what problem we should be focusing on, or what user needs we should be tackling. Not all problems can be solved with AI, therefore we should not start with the assumption that it is the only solution. First, we must identify the user need, and only then assess whether AI can help solve the problem. I’ve found this to be difficult for AI teams, but design maturity can help ensure the customer is considered first.

Design

Prototype & user test with real data

It can be difficult to properly evaluate the experience of a personalised product using dummy data, and this is where the benefits of a multidisciplinary team really kicks in. Data scientists/engineers and designers should work together to create realistic prototypes using real customer data. With this method, we are evaluating the prototype not only in terms of usability but also accuracy. Are our recommendations actually relevant to this customer? Are our suggestions of real interest to users?

Practically, this may mean create more personalised prototypes by using real data. This can be time-consuming, but yields extremely valuable insights. We’ve sometimes used a Wizard of Oz method that “fakes” the product’s response to a users’ action (e.g. test a chatbox interaction by having someone answer each message while pretending to be a bot).

Strive for transparency

Given the number of data breaches and examples where AI has really gotten it all wrong (check out this visualisation from AI Now to see the extent of the problem), people are understandably worried about the transformative impact AI will have on both the individual level and society at large. At Zalando we always strive to be transparent about how we use data and try our best to explain why we recommend the items we do (shout out to my colleague Clementine Jinhee Declercq who worked on this initiative 🙌)

Image for post
Image for post
Brand recommendations on Zalando

Deliver

How good is good enough?

Machines are not always right. It’s important for us to understand just how “right” the algorithm needs to be, and what the consequences of any errors might be on users. For example, we’ve learned the impact of recommending the wrong size is far greater (disappointment, returning the item) than recommending a color or material they don’t like.

Build in a feedback loop

It is also good to design a way for users to easily give feedback on recommendations (e.g I don’t like this item, I don’t like this brand, I don’t like the colour). Such feedback can be used to retrain the model and provide customers with more relevant recommendations over time.

Empathy is key

So, does our design process still hold up? Yes, I think it does. But to be honest, I’ve learned that process is secondary to mindset. If we, as designers, can instill a mindset of customer centricity and empathy in our colleagues, then it is of little consequence what process we use. The customer must be foremost in the teams’ minds in order to ensure AI enabled products are built responsibly, inclusively and in the best interest of users.

Big thanks to my colleague, Philip Kelly, Senior Data Scientist, for helping me with this post ✌

Rosarie is a senior product designer (UX) with Zalando’s Dublin office.

Zalando Design

A collections of stories from Zalando Design

Rosarie O'Regan

Written by

Lead Product Designer

Zalando Design

We are a diverse, fast-growing community of more than 130 designers, researchers and writers from over 30 countries. Read about how we combine empathy, science and creativity to make engaging and joyful user experiences to ultimately reimagine fashion for the good of all.

Rosarie O'Regan

Written by

Lead Product Designer

Zalando Design

We are a diverse, fast-growing community of more than 130 designers, researchers and writers from over 30 countries. Read about how we combine empathy, science and creativity to make engaging and joyful user experiences to ultimately reimagine fashion for the good of all.

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store