One of the great things about machine learning is that it can support and empower our human experience, an effort called human-centered machine learning at Google. But what does machine learning have to do with user experience (UX) design? Let’s start by defining each term and then we’ll take a look at how they affect one another.
What is Machine Learning?
Machine learning (ML) is, according to wiki, “the scientific study of algorithms and statistical models that computer systems use to perform a specific task without using explicit instructions, relying on patterns.” And in layman’s terms, machine learning makes it possible for machines to learn from experiences, adjust to new inputs and perform human-like tasks.
What is User Experience Design?
In the words of Don Norman, user experience design “encompasses all aspects of the end-user’s interaction with the company, its services, and its products.” Simply stated, UX design is a human-first approach to solving problems.
Want to know more about UX design? Read my article about it:
In order to better understand the user experience of machine learning, let’s discuss some UX design processes and tools that could come in handy.
The User Experience Design Process
A UX design process focuses on enhancing a product’s value and accessibility to the end user and that might look like conducting user research, creating personas, wireframing and prototyping, user testing or a UX audit, or even going through a design sprint process. As a UX designer, every project brings its own challenges, and knowing you have your toolkit essentials ready to go could save you a ton of time and energy.
Here is a list of my 6 UX must-haves for any project.
Download the UX Checklist
Why UX is So Important to Talk About in Machine Learning
Output of ML solution tends to look like magic. People don’t understand it, so they push back against ML solutions. ”People climb mountains and expose themselves voluntarily to all kinds of risks, but they don’t like risk inflicted upon them that they don’t understand or have control over …” said Dr. Roger E. Kasperson, director of a risk study center at Clark University in Worcester, Mass, and I’d argue that the same concept of fear could be applied to ML outputs. ML solutions do not properly interpret the user’s context. ML algorithms still need a human to know what human problems to solve.
How are UX Design and Machine Learning Similar?
Both UX design and ML are centered around learning the end user’s behavior and predicting user actions. Both UX design and ML are trying to make products valuable, desirable, usable, useful, accessible, findable, and credible to the end users.
The UX of Machine Learning
Now that we understand both UX design and ML, let’s use a user research approach/process to solve machine learning problems. Here is a list of questions we should be asking during the process to get the most out of ML.
Define a Problem Statement
We should strive to find meaningful problems that are aligned with a human need. We can do so through user research to find pain points, expectations and mental models about the problem we are trying to solve. Here is a really good quote by one of the senior researchers at Google that explains this.
“If you aren’t aligned with a human need, you’re just going to build a very powerful system to address a very small — or perhaps nonexistent — problem.”– Jess Holbrook; Human-Centered Machine Learning
Does Our Problem Require a Machine Learning Solution?
Once you’ve found your customer pain points, you need to ask if this problem needs a machine learning solution. Not all problems and pain points require machine learning solutions. A good example of this is Typeform. As of writing this blog post, Typeform does not require a ML enhancement to feel personal and smart. For anyone unfamiliar with Typeform, here is an explanation:
Typeform is “bridging the gap between data collection and customer interaction.”– Robert Muñoz, Co-CEO and Co-Founder of Typeform
The big challenge here is knowing which pain points could use a ML enhancement to greatly increase the UX of your product. After conducting research to better understand your target audience, product teams can go through an ideation process to come up with feature/product suggestions and then then vote on which feature provides the most user impact and the impact/experience can be enhanced by a machine learning solution.
Example of Product Experience that is Enhanced by a ML Solution
GBoard: “A virtual keyboard you can download onto your Android/IOS device. Gboard has everything you love about Google Keyboard—speed and reliability, Glide Typing, voice typing, handwriting, and more—plus Google Search built in. No more app switching; just search and share, right from your keyboard.” – Google.
Focusing on the handwriting feature alone, Gboard predicts with precision and transcribes our handwriting to text. For those of us who write faster than we type, this tool is very convenient.
Weigh the Cost of a ML Solution Mistake
You can decide whether to use a ML solution by weighing the cost of a ML solution mistake. When errors happen, how is the user’s experience affected? Most times we are super excited when our predictions are right. What happens when they are wrong?
Just as with any software development processes, we have to make tradeoffs. In this case it’s between Precision and Recall—core concepts in machine learning classification problems.
Precision means we are trying to reduce the number of wrong predictions even if we lose some right ones along the way. Recall means we are trying to get as many right predictions as possible even if we allow some wrong ones in.
Establish a Feedback Loop
Our machine learning model needs data to keep learning and make predictions with more accuracy to create a personalized experience for the end users. Our end goal here is to encourage users to give feedback that benefits them and our model. Here is an example:
Focusing on user research and feedback can help enhance the user experience of your ML solutions. ML is here to support our human experience, so let’s remember the importance of keeping the user at the center of current and future ML solutions.