Preventing Confirmation Bias In UX

Leah Cunningham: All it takes is one bad design decision to have a negative impact on your digital products. One reason that bad decisions happen is a human behavior called confirmation bias.

Confirmation bias is when a person or team uses pre-existing opinions or assumptions while ignoring other options.

Here’s an example. Google+ was a social network that pushed the concept of circles, a complicated interface that users weren’t actually interested in learning. Google had made assumptions about what users really wanted and failed, leading to its eventual shutdown. The cost has been estimated at half a billion dollars, so very costly assumptions indeed.

Hi. My name is Leah Cunningham. In this tutorial, you’ll learn different strategies to prevent confirmation bias in design. In the early days of UX, company leaders were obsessed with counting clicks. There was even a saying that no page should take more than three clicks to access.

In fact, this is just another form of bias, and the results are often complex navigation patterns more focused on click reduction rather than intuitive design. Steve Krug sums up this concept nicely in a quote from his classic book, Don’t Make Me Think.

It doesn’t matter how many times I have to click, as long as each click is a mindless, unambiguous choice.

— Steve Krug

No matter how well intentioned we are, we all have biases.

Big Idea: Confirmation bias spreads because everyone makes assumptions based on experience.

So where does it all begin? Confirmation bias happens when a biased research result is incorporated into designs. This happens because it’s really easy to fall into a habit of tweaking even the most balanced research output to fit with your personal assumptions.

Best Practice: One way to prevent this is to avoid selective reporting of research findings that focus only on the positive aspects of a design.

Additionally, it can be easy to accidentally skew the integrity of the data you collect just through how the questions are asked. This leads us to another best practice.

Best Practice: Always structure research questions in a way that is neutral and makes no assumptions about what the correct answer is.

It’s also critical to ensure you’re casting a wide net to understand all viewpoints. To help with this, always consider data from multiple sources. There’s a method I could recommend called triangulation.

Big Idea: Triangulation is a model for reviewing research results and asks you to consider all of the following:

• What was said?
• What was observed?
• And what was done?

Following these steps ensures the broadest range of feedback is being considered. Using the triangulation method helps reduce the risk of biased designs, which is especially important in the modern pattern library production environment. Here’s an example. Imagine you’re designing an investment dashboard for independent financial analysts. You might make the assumption that they want as much data visible in as little space as possible.

Why? Because that’s how popular software such as Bloomberg Terminal does it. However, you can’t assume all financial analysts consume and act on content in the same way. Do independent financial analysts really have the same needs as the corporate financial analysts who use Bloomberg? Maybe. Maybe not. But you can’t assume.

So it turns out this biased pattern is actually too complex and confusing for your audience of independent financial analysts. But guess what, it ends up at a pattern library anyway. Because of their nature, pattern libraries are shared among large teams. And now your small problem of one confusing design element has potentially spread, much like a virus, to other digital products within your company, impacting additional users and expanding the risk.

Best Practice: To prevent confirmation bias from spreading through pattern libraries, it’s important to have a rigorous, cross-functional review team for any new patterns being added to a central library.

Beyond that, it’s important to ensure there are proper approval processes, usage guidelines, and documented best practices for items accepted into the pattern library. This will help maintain the integrity of the system and prevent the replication of any issues in future designs.

Recap: As we said at the beginning, challenging your own assumptions is critical to reducing bias and minimizing product risk. It’s worth the hard work.

For more guidance on how to do this, check out the Resources section of this tutorial. Thanks for watching, and be sure to check out the rest of our Take 5 tutorials at thegymnasium.com.