Side Project: Exploring Machine Learning & AI
What’s for dinner? Using Predictive UX to help users decide
Project Goals
In 2018, a client asked me "What does the future of UX look like in five to ten years?" I have been driven to find an answer to this ever since. I embarked on a year-long project with a colleague to explore Predictive User Experiences and Machine Learning, the learnings of which have fundamentally influenced my work ever since. Through this self-driven project, I had three speaking engagements that generated sales leads and career interest for my agency. The work that we did also influenced the agency’s personalization activities.
My role:
Lead and solo designer — academic research, user research, design, project management
Collaboration with back-end developer/engineer Gurwinder Antal
Overcoming a fear of public speaking to present at three conferences in front of groups of 50-120 (DrupalCamp Colorado, Bay Area Drupal Camp, and DrupalCon North America)
Opening a dialogue with the open source Drupal community on machine learning
Gauging user sentiment toward machine learning
I polled 50 people and asked them what their willingness is to have AI make small and large decisions for them. 44% were excited by the idea of a computer making small decisions on their behalf (e.g. allowing AI to set the temperature in your home.) When we asked about large decisions, such as restricting your ability to spend money to help improve your finances, we saw the opposite results. Only 20% were excited by this idea vs. 38% who were opposed.
IBM wrote a great article for The World Economic Forum about why UX is critical to the successful adoption of artificial intelligence and machine learning into society. The author asserts that to successfully implement machine learning, the algorithms and models needs to be there. But it’s not sufficient to drive adoption rates. This is a technology that requires deep change of human behavior and perception and requires careful consideration of the user experience to marry its capabilities with how users engage with digital experiences day to day.
Let AI help you decide what’s for dinner
What’s for dinner? This is a constant debate in my household and many others, I’m sure. How do you decide? Maybe you go on your favorite meal ordering app to find something, which offers you hundreds of options that you have to now browse through and make a selection. And there is so much to choose from! This is where we chose to develop the use case for our demo—a food delivery application powered by a recommendation engine. We hypothesized that AI applied here would help users reduce mental load.
Make it personal
I designed a microquiz that allows us to hone results. One tap indicates cuisines that users love, two taps indicate a favorite and allows us to weight results accordingly. With this feature, we are able to collect explicit data from users without an actual purchase. And from there, can immediately personalize results.
Mix it up
Conscious of helping users avoid a filter bubble, I incorporated a row of recommendations that intentionally do not align with the users’ past behavior. The goal is to show users options for things outside their normal box. I also added a “healthy mode” to allow users to self-select into healthier recommendations.
Key takeaways
Design ethics are critical
This technology has the power to change people’s perception of the world (and it already has). Don’t take it lightly.
AI has human bias built-in
Algorithms are binary and can’t make judgement calls. And they are unavoidably influenced by those who write them.
Hardware is a barrier
It takes A LOT of servers to crunch the data needed for machine learning functions—a major limitation for open-source.
Users are willing to give up data
58% of people polled are uncomfortable with sharing data but willing to do so if they get something in return.
Watch the presentation
Reading list
Introduction to Anticipatory Design, 1st Edition by Joël van Bodegraven. This edition is no longer available–link is to new ebook material.
The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World by Pedro Domingos
Superintelligence: Paths, Dangers, Strategies by Napoleon Ryan
Customization vs. Personalization in the User Experience by Amy Schade, NN/g
Why Scientists Are Upset About The Facebook Filter Bubble Study by David Lumb, FastCo
Instagram has a drug problem. Its algorithms are making it worse by Elizabeth Dwoskin, Washington Post
Why artificial intelligence is learning emotional intelligence by Jesus Mantas, IBM
Measuring the Filter Bubble: How Google is Influencing What You Click by SpreadPrivacy
Building Ethically Aligned AI by Francesca Rossi, IBM
Why Amazon’s Anticipatory Shipping Is Pure Genius by Professor Praveen Kopalle, Darmouth College
How Does Spotify Know You So Well? by Sophia Ciocca, Software Developer at the New York Times
Thanks for making it this far.
Want to check out some more of my stuff? Dig into how I reinforced customer-first values with the Union Bank & Trust site redesign 🏦 See how I built a learning delivery platform for the Human Capital Institute 🧠 Explore how I helped move Firewise USA® from a manual paper process to a low-touch, dynamic web app 🔥 or, learn more about me 👱🏻♀️🐈