A UX perspective on artificial intelligence

“The purpose of artificial intelligence is to make technology disappear” — Rand Hindi, jury member of AI Labs

I want to share 3 UX insights that I picked up last weekend, while I was participating to AI Labs, an Artificial Intelligence hackathon.

The concept was simple: assemble 10 teams of data scientists, business developers and UX designers and let them explore how the latest artificial intelligence evolutions can help people buy a house, combat loneliness, avoid traffic, learn chess, follow politics, discover music, prevent car accidents, run errands, meet interesting people or fund startups.

As a mentor, I was available for the teams who wanted a user experience perspective on their projects.

Here’s what I learned:

1. Accepting personalized recommendations

Imagine being on your couch, and receiving the following notification: “Hey, you’ve spent 3 hours on Facebook today. Wanna join switch off from your mobile? Marc is leaving for a run in 20 minutes and would love to see a friend join him.

What would it take for you to act on it? Would you need to have indicated beforehand that you’d like to reduce the time you spend on Facebook? Or that you already run on a regular basis? How much serendipity are you willing to accept?

2. Handling exceptions

A team came up with an algorithm that will use my Spotify library to predict where I’ll want to hang out tomorrow night. Great concept for me, since being a new dad erased all knowledge about where the right bars are. The only problem is, my second most played album is a lullaby record that I put my son to sleep with. Taking it into account to anticipate the music I’d like to hear in concert will lead to weird recommendations.

How should such an exception be handled? Should the algorithm learn how to recognize it? What if it doesn’t? How do I tell the app to discard it, without having to touch my Spotify library?

3. Surfacing implicit patterns

Let’s take a woman who wants to rent a house. She’s browsing ads to choose the ones to visit. One of her criteria is to live in an authentic place. She really likes buildings that carry stories in their walls. Of course, she didn’t give these details to these sites, not only because they didn’t ask, but also because they wouldn’t know what to do with it. However, an algorithm might be able to pick up a pattern, and cluster her in a group of people who seem to like exactly the same ads — without knowing the common denominator of those ads, nor being able to give it a name.

How can we ask people for soft preferences? When a pattern is spotted, should we try to surface it? Even when we can’t name it accurately?

Have you worked on artificial intelligence with a UX lens? We’d love to hear what you learnt. And if you want to read something about artificial intelligence that freaked us out at BlaBlaCar, take a look at this article from Tim Urban.