Don’t You Just Love being in Control?!

Earlier this year, Alex Tim and I delivered a workshop at the UX STRAT conference in Amsterdam. We developed our “AI Design for Enterprise Products” workshop for UX designers across industries, and centered around our experience working on CLAIRE, Informatica’s cutting edge AI technology featured in many of the company’s products. We spent the majority of our time discussing our thoughts on Informatica’s five UX design principles for delivering a greater user experience for Artificial Intelligence (AI).

During that workshop and others that followed, I learned that the principle attendees identified with most was Control. And I believe that this is the case because as independent thinkers, being in control is a preferred state of mind in every aspect of our lives, and even more so when we interact with AI.

Image is a photo of our workshop’s illustrated notes created by attendee Corey Roth, during the UX STRAT workshop in the US

Humans are typically most comfortable when in control. “Black box” systems steer users away from their comfort zone and into unplanned interactions, confusing pathways and unpredictable outcomes.

Very often, people don’t “get” AI. Which makes sense when you remember that the technology of artificial intelligence is based on complicated calculations and machine learning of multiple data sources. There are algorithms that crunch an inexhaustible amount of data that we humans simply cannot comprehend. Moreover, many of the rules that AI is based on are created by complete strangers. Inevitably, users only experience the end result of this brilliant process, and gain no insight into the fine workings under the hood of the AI software which, in turn, hinders their ability to waive control.

The question we need to ask ourselves as user experience designers is: how can we help users feel in control and, at the same time, secure in their decision? During the workshop, I found it easier to demonstrate this principle, and the following related UX tips, by sharing my experiences with the navigation system installed in my car:

  • Provide insights into what the user should expect. When a navigation system offers me a route, it shows me a map with the suggested route highlighted. This way, I can see which roads I will take until I reach my destination, and I can decide what to do next based on that insight. Some systems will offer me alternative routes based on my preference. For example, it might ask me whether I wish to include highways or stick to small country roads. Asking for my preferences will assist me later in my understanding of what to expect from the system suggestions.

Informatica’s CLAIRE Lookup recommendations inside Cloud Data Integration Mapping Designer. Using CLAIRE Recommendations to present the user with a lookup condition, the user can inspect the condition and decide to accept or reject it. (Image by Jim Reed.)
  • Focus on failure and don’t assume success. It is easy to dismiss or ignore a product’s recommendation, but not so easy to recover from the actions you choose (or don’t choose) to take based on those recommendations. For instance, if the suggested route has taken me into an unexpected traffic jam, the system should be able to recalculate and suggest an alternative route. This is why we should always consider supporting “Undo and Redo” actions when designing product interaction models.
  • Ask for user feedback and have the system react to that feedback. The navigation software in my car warns me about incidents (such as a car stopped on the side of the road) that might interfere with my driving. The system also asks for my own feedback if I happen to come across such incidents.
  • Allow editing. User preferences may change over time, so it’s important to consider how you’ll give users control over the preferences they communicate to the AI system—that is, always allow for adjustments. Often, when I take a familiar route, I’ll decide to deviate from the system’s suggestions (as I still know best). So the system should be able to recalculate the route according to my preference.
  • Allow users to turn the AI interface on and off.  The navigation system I use has three operating modes. Mode 1 shows the route on the map and offers audible driving directions. This is the mode I use when I visit new places. Mode 2 shows the selected route only, without audible direction. This is the mode I use when I’m driving home from work in peak hours, just so I’ll know what to expect. Mode 3 shows the map without any selected route, only providing advisories about heavy traffic or route cautions. That’s the map I use on weekends.

    I l-o-v-e this example, as I believe it underscores the idea of just how versatile AI is, and the many types of use cases and personas it can support.

Want to learn more?

If you want to learn more about our workshop, “AI Design for Enterprise Products,” you can listen to our podcast.

The Control design principle is one technique we use to implement CLAIRE. You can learn more about other principles for AI design: Trust, Clarify, Simplify, and Humanize.