← Home

Subjectivity & Design

“On the Web, design has become much more of a science than an art. Because you can iterate so quickly, because you can measure so precisely, you can actually find small differences and mathematically learn which one is right.” - Marissa Mayer.

That quote makes my skin crawl. Design as objectively right or wrong is a bit ridiculous. But based on click-through rates? That's plain horrifying. That said, whenever someone says, “Design is subjective,” I also cringe. Buttons too small to press, unclear copy; how could those be anything but bad?

When we call something subjective, it's often really a difference in value judgements. In other words, we're really looking at a failure to state our criteria of evaluation. Google, for example, might define ‘good design’ as whatever makes people click on ads. I define it through something resembling Rawlsian social contract. If Google and I disagree about whether something constitutes good design, it’s probably because we have different conceptions of goodness, not because we disagree about the characteristics of the design itself. We're looking at imprecision, not real subjectivity.

Other times, when things are subjective because they're hard to measure, it’s a bit more tricky. If Google and I both agree that all other things being equal, a design that produces more joy in a typical user is better, we can still disagree about two designs based on which we believe will produce more joy. Here though, one of us is wrong. It’s just hard to prove which one of us. (Theoretically, we could hook users up to fMRI machines, and scan their brains while they used the product, but their preferences may change in an fMRI.) Furthermore, a designer is often aiming for a more complex emotive response than just ‘happy.’ Excitement, security, comfort, curiosity; we want to design for what it means to be a human. This is why designers need to be human; a designer with a canny sense of intuition and the ability to conduct research can come to an accurate understanding fairly easily, allowing them to make changes much more reliably and precisely than using A/B testing to throw spaghetti at the wall until the metrics look right.

That’s not to say, however, that A/B testing is some kind of inhumane and bizarre imposition on the design process. Great artists, it turns out, often A/B tested their own work. They’d create a new version, see how they liked the new composition, and then make further changes, until they came to a final version that they found good. (For more on this, see Why Mondrian Was A Great Optimizer.) A/B testing is another way to look at whether people respond to designs in the ways we hope they will. While Mondrian can evaluate his own pieces by examining the composition, a designer needs to see how people come to understand their work through time, and so being able to see the way users respond to the artifact can be hugely useful tool.

A/B testing is useful to designers not because it can pick the ‘right’ design; it's useful because it offers much more precise information about how choices affect actions than can be obtained by doing conventional user tests. This gives designers the knowledge they need to craft artifacts that get closer to the conception of good design that they (or the corporation they work for) choose.