AI Research and UX Design
Not long ago I had an epiphany. What a brilliant mystery of the brain, epiphanies are.
I was showering (the holiest of thinking places) and kicked the motors of spatial thinking into a slow rumble. Prior to showering I had just wrapped up a prototyping session and it hit me…
Why not model a thought hierarchy of the user with cumulative research and testing data?
Ok so that’s a pretty ambiguous thought out of context… sorry. Let me explain. One of the most important ways to challenge yourself as a UX Designer is in finding new ways to observe your user. This stems from the indelible quality you must have as a UX Designer, empathy. I’m deeply interested in AI and through my studies have found some parallels between AI and my work as a UX Designer.
The holy grail of AI research is replicating how the human mind works. What this means is that researchers have to break components of the mind down to such a fundamental level that it can be replicated programmatically. When we start to understand how to model a component of the brain, we can begin to predict how that component impacts human behavior. Neuroscience leads the way here, but AI research has taken a great interest in discovering how we think at a deeper level.
How AI might influence design
Currently we have a pretty standard process when designing for people. We conduct user research, create user personas, prototype ideas and relentlessly iterate to refine the ideas. When executed correctly, this process generates great results — but doesn’t always give you the full picture. Even after an exhaustive amount of iterations or enhancements.
Among all the research in AI taking place, one of the most recognizable figures in the space is Ray Kurzweil. His theory on the Neocortex (among the vast range of other things he’s done) is being used as a foundation for how Google tackles AI’s ability to understand and communicate in natural language. His theory suggests that the human brain is made up of 300 million modules each of which can learn, recognize and implement a pattern. Furthermore, all of these modules are organized into a hierarchy created by our own thinking. It’s organized in such a way that the things we think and care about the most sit at the top of the hierarchy and it descends down from there.
Let’s look above at this very basic hierarchy model to give some visual context. Visualize each dot as a data point that makes up a user’s thought hierarchy. This hierarchy indicates that Jenny is a Tax Accountant, with Accounting sitting at the top of her thoughts. It’s what she does professionally, it’s what she thinks about for 40–60 (hello tax season) hours a week and thus — it sits at the top. But within that data point or ‘symbol’ of Tax Accounting is an even deeper series of nested thought hierarchies.
As you can see from left to right things can become nested at a pretty deep level. Additionally, the above model doesn’t even factor in the patterns that develop between these symbols. Which, ultimately forms the complexity of our memories or what we all know as knowledge.
What are the benefits of this?
What if you could see a hierarchy of things that matter the most to your users? A prioritized list? This would help tremendously to support design decisions.
Ok so you’re thinking…”I already do that in my processes!”.
I’m not talking about what we learn from individual exercises. I’m talking about a hierarchy that’s the product of data aggregation — from ALL of your processes and tools in a culminate manner.
That’s great, how do we do that?
We aggregate feedback and results from our processes and tools throughout the design process and beyond. It needs to be an ongoing and living process. Here’s a brief list of what I use to help generate this data:
- User Interviews
- Card Sorting
- Whiteboarding
- User Personas
- Google Anlaytics
- Inspectlet
- etc
What we do is analyze the aggregate output of these processes and tools from all our users and search for consistencies. Then we use those consistencies to build a hierarchy of subjects that are important to the user. In it’s entirety you end up aggregating things people say they want, need, what goals they have, things they click on, tap on and interact with at an over-arching level.
What’s important to realize is that most UX Designers use these tools and processes as a way to develop/refine their designs already. However, there tends to be a disconnect between what what you see in each of these processes individually and what the data is saying in culmination. You have the same data coming from different areas and in a different format.
Google analytics may be telling you that someone is going to a page in greater frequency. Your heatmaps may be lighting up the user’s interactions leading to that same page. During prototyping the user may have realized the need for that same thing in retrospect to a flow. But the user may not have indicated a need for it during initial user interviews. They’re all different pieces of data — but in summary the need becomes more apparent when summarized. Individually they might be considered an anomaly or less of a priority.
How do I aggregate this data?
Simply put, a spreadsheet. There’s no need to complicate it with a convoluted process. It’s an easy transition for a lot of UX Designers too because they already use spreadsheets to capture User Persona data. I personally have a tab for User Personas, Google Analytics, User Interviews, etc etc. I then have a master tab that weighs the consistencies and prevalence of specific subjects across all the tabs.
From there I build a thought hierarchy and I constantly refine it as things change. This hierarchy gives me direction on what I might be missing. You’d be surprised what the data says when you take a step back and look at it all from a different angle.
What else?
One question that’s been posed to me is how I handle conflicting data. What if you have half your data saying one thing, but the other saying something completely different? For the sake of time and your attention span, I will go over that in a future post. :)
But before you go… please know that this process is really just the tip of the iceberg! There is additional insight as a UX Designer that you can garner from aspects of hierarchical thinking that I didn’t even delve into yet. It revolves around how memories are formed and how identification of their sequential makeup can help you modify user behavior. Coming soon as well!