Aesthetic Value Judgment

The Missing Piece in the Creative AI Debate

AI creative work has taken a beating in the media and social media recently. Some of the criticism is fair. A lot of AI generated art does sound or look off, even bad. But the tool isn’t the reason. The person using it is.

Aesthetic value judgment is the ability to look at a piece of creative work and, using experience, knowledge, and acumen, gauge its actual worth against what the creator was trying to achieve.

Someone who has spent decades steeped in an art form develops a far sharper sense of this than someone who hasn’t. This holds true whether the work is entirely human made or built with AI tools from the ground up. Aesthetics, the branch of philosophy concerned with beauty and the value of creative work, has always been about this same question: does the work connect to others, and why.

This concept is exactly what’s missing from the mainstream conversation about AI. Media coverage and social media chatter miss it because they misunderstand what a creative tool actually does. The public imagines AI as a generation engine that replaces a creator. For a skilled professional, it functions more like a high-speed execution engine, one that still needs a hand on the wheel.

Decades of creative and technical experience give a seasoned professional a real advantage here, and it shows up in a few specific ways.

Curation is the first. AI can spit out a thousand variations of an image, a sentence, or a melody in minutes, but it has no taste of its own. It cannot tell a generic, statistically average result from something genuinely alive. A trained ear or eye can sort through the noise very quickly, spotting which version carries real emotional weight, which composition holds together, which one will actually reach the audience it’s meant for. They do this by comparing what they see or hear, against the vault of knowledge in their mind from evaluating creative works for years, decades often.

Someone without that baseline often can’t see past a surface impression. They notice something looks clean or sounds pleasant and stop looking any further, because they don’t yet have the reference points to look deeper.

Catching the micro-flaws is the second. Generative tools tend to fail at the edges. The main subject looks right, but the way light wraps around it is subtly wrong. A shadow carries an unnatural color shift. A music track loses an edge in EQ, is over compressed, “shimmers” when it shouldn’t. A trained eye, or ear, catches these things in a fraction of a second, because years of building things by hand teach you exactly where the seams usually are. That same experience makes it possible to step in and fix the problem, whether by hand or by re-prompting the AI with the right correction in mind. Or in many cases, both, as a form of hybrid production.

Prompting with real vocabulary is the third. Getting a precise result out of any advanced tool depends entirely on the quality of what you feed it. Vague prompts like “epic lighting”, or “catchy melody” just push the AI toward an average of its training data. Someone who understands the actual mechanics of a craft can reference a specific lighting setup, a particular historical art style or fusion of styles, a historical music movement, a compositional choice. In music that might mean a specific key, meter, instrumentation, or arrangement style. In writing it might mean pacing, voice, or the texture of a scene. In art it may mean knowing the subtle difference between impressionism and expressionism, and when and why you’d want one over the other. That level of specificity guides the machine with intention instead of leaving the outcome to chance.

Here’s where it gets more interesting, though, and where I have to be honest about the limits of expertise and caution an experienced artist to think openly and dynamically. A reflection of Shunryū Suzuki’s famous work Zen Mind, Beginner’s Mind. One where “the expert knows of many limitations, while the beginner knows none.”

I once had a teacher, some decades ago, who could hear a chord progression, one he must have encountered a thousand times over his career, and wave it off without a second thought. I watched him do it more than once. What struck me even then was that his dismissal wasn’t just quick. It was fixated. The familiarity of that progression had become a kind of static in his ear, something he couldn’t hear past. A beginner sitting in the same room, someone without thirty years of pattern recognition weighing on them, could hear the same progression and catch something the teacher missed entirely: an overarching feeling in the piece that was fresh, real, and worth paying attention to. Not because the beginner had better ears. Because the beginner’s ears weren’t blocked by the exact thing the expert’s were.

This is the trap hiding inside expertise itself. The deeper your knowledge runs, the more finely tuned your radar becomes for certain kinds of flaws and certain kinds of familiarity. That same tuning can also wall off entire categories of experience from reaching you at all. An expert can become so certain of what they already know that they stop listening, looking, or reading with any real openness. That’s a much harder trap to notice than it sounds, especially with the sheer volume of content AI is able to generate, because it doesn’t feel like closed-mindedness from the inside. It feels like knowing with certainty what you’re talking about.

Aesthetic value judgment without intellectual humility becomes aesthetic rigidity.

Philosopher Bertrand Russell, in a push of individualism, once quipped:

“Pay no attention to the authority of others, for equal and contrary authorities can always be found elsewhere!”

Russell was needling people a bit when he said it, and you could read it as an excuse to dismiss expertise altogether, which I don’t think is the point. But there’s a real grain of truth in it. No amount of experience makes a person’s taste infallible, and the moment you start treating your own judgment as beyond question is the moment you stop growing. Staying open, genuinely open, to what a beginner hears, or what a new tool makes possible, is often much harder than it sounds. It’s a discipline, not a personality trait, and even the most seasoned, respected voices in any field can lose their grip on it without noticing.

This same tension, between refined taste and the humility to keep refining it, is also what pulls people across the spectrum of hybrid creation. Dissatisfaction with something mediocre is itself a form of taste. Dissatisfaction with aspects of a work is itself a form of taste. And it’s usually what drives someone from passively accepting whatever a tool hands them toward actively shaping it into something better.

None of this means expertise doesn’t matter. It clearly does. A seasoned professional using AI isn’t just an old dog doing tricks. They’re closer to a director working with a faster, more responsive crew. A composer working with an esteemed ensemble of musicians. They are someone who can spot the flaw, name the fix, and push the work somewhere a beginner may not know to look. But that same director, or composer, owes it to their work to keep an eye or ear open for what the crew, or the beginner standing next to them, might hear that they can’t. Taste sharpens with experience. It can also calcify with it. The goal isn’t choosing one truth over the other. It’s holding both at once, and using that to build one’s knowledge to even greater depths.

Inspiration for this essay comes from Aristotle’s Poetics, GF Hegel’s Lectures on Fine Art, Edmund Burke’s On The Sublime and Beautiful, Arthur Schopenhauer’s The World as Will and Representation, and Douglas Hostatadster’s GEB, and obviously Suzuki’s Zen Mind, Beginner’s Mind.

Edit: As an amusing anecdote to this, I wrote the vast majority of this myself, putting what I felt was a significant amount of mental energy into carefully coining the phrase aesthetic value judgment itself, but also the importance of it’s meaning in today’s rapidly changing world. It took a few drafts over the course of months, getting feedback from Anthropic Claude and GPT on tone (not wanting to sound too snobby, mostly), and editorial concerns (grammar, flow, syntax, etc.), or when something I wrote didn’t read well, making various changes along the way using what I thought was the best wording in the end. I even included a real-life story from my youth as an example. When I had Substack check this for AI, it said it was 100% AI written, zero human! This is the world we live in my friends!