Taste, Judgment and AI
Taste is recognizing quality and greatness in the absence of all the evidence. Judgment is committing to it in the presence of some risk. It’s the ultimate foundation of great work.
When you look at a web page or a painting and think "that's beautiful" or "they were uncompromising in quality" you usually can't say why.
You're not measuring anything. But you're sure, and so is everyone else looking at it. That's the thing being recognized. Someone made a series of calls they couldn't prove were right, and the craftsmanship shows.
Taste is built by doing the reps and being able to recognize what good and great mean. It's when you've put in enough of the reps to be able to look at a dozen options and point to which one is the best. Especially when what "best" is ambiguous.
If you expose yourself to the best things that humans have done and try to bring those things into what you're doing, you may just care about taste.
Taste is a lot like a filter.
Through a lot of history, you typically had a smaller pile of options to filter through for great because creating things was expensive. You would probably build just one version of something since that's all you could afford. If it was more than that, great.
You can’t really earn taste just by consuming a narrow feed of the same stuff all the time. When you experience variety, stuff that makes you question or form new opinions, that variety and friction helps sharpen your ability to tell what is quality and what isn’t.
After all, you don't know what you're missing if you never see it. We need to be exposed to the unexpected in order to develop a real sense of taste.
Taste is important, but there's a necessary next step: judgment.
Judgment is that willingness to sign your name to a choice and to be accountable for it. Judgment is making your best quality call on what ships, when the stakes are high enough that you can't just walk away from a bad decision.
Judgment applies when you're selecting the right one amongst many bad options. Taste is what allows you to find the best option to put forward.
Judgment is what allows you to own that choice, to accept the necessary trade-offs, and to be able to explain (and defend) your decision if things go wrong.
In many ways, taste can be borrowed, copied, derived from others, learned from agents, copied from assistants. But judgment cannot be transferred.
Because judgment is insisting on your name attached to the consequences, and agents don't have names yet. Agents get bigger and smarter, but at the same time they're meeting more and more of a ceiling.
They can provide and rank many options well, but once the math is done, when the choice has to be made, we're still reaching for humans, the only beings still left that can be answerable. Agents can continue to grow in their ability to assess options. The next frontier is building capacity for humans to do better than agents at judgment.
This is why the interesting question isn't whether agents can develop taste. You’ve got to assume they can. The question is what's left when they do. Judgment can't be easily copied because it isn't a capability. It's a liability and judgment is insisting on your name attached to the consequences.
And indeed, judgment is not a universal skill. There's no one who's seen all mistakes and been through them all. Skill grows in areas where we've seen and survived many mistakes.
An example in action: Apple
Apple’s design taste is often cited as being exceptional. Think about the MacBook, the iPhone or their other products and it’s often clear they care a lot about how seamlessly things work and feel to their users. But there’s another Apple example that touches on both taste and the accountability of judgment: Maps.
Back in 2012, Apple pulled Google Maps out of iOS and shipped their own version. The taste was there - vector rendering that stayed crisp when you zoomed in, a nice typographic treatment. Unfortunately, there were larger issues at play. Customers complained they saw bridges melt into rivers, searches for Berlin landing them in Antarctica and an entire Australian town misplaced 70km out into the desert. Wowzers.
Rather than blaming this on the iOS Maps Team - Tim Cook himself published an apology on apple.com in the first person taking accountability for decisions made across a large organization over what was probably several years. Reporting at the time suggested he wasn’t the only person asked to sign it, but whether or not thats true, he’s the one whose name was on it. Both the taste and failure were obvious then. The signature was what cost something.
I think we can agree that many people wouldn't sign anything that public and take the accountability. The rough shape of this still applies to how you work. Think of the pull request you approve when you know the tests aren't that great. Or that new UI you rushed out without enough testing or talking to your users. These don't always get the same kind of public apology, but you carry the lessons they teach you. Your taste compounds through exposure and judgment compounds through the consequences.
Closing thoughts
If you're interested in refining your taste, move past phrases like "I like this" or "I hate this" and ask yourself why something succeeds or fails. Study different versions of the same thing and see if you can appreciate the subtle differences. If you want to build judgment I say make the hard calls. Write down what you might have expected would happen and stay close enough to the outcome to learn if you were right and more importantly if you were not.
Over time, trust your inner voice and what you truly believe makes something high quality.
Pangram 4 rated this article as 100% human authored.