Diminishing Returns

Sooo, funny observation (or not so funny, depending on what side of the debate you are). There is a 19th-century economics concept that explains something important about what's happening to creative work, specifically AI work, right now but a really universal rule if you want. It was first formally described by a German economist named Hermann Heinrich Gossen in 1854, later refined by Alfred Marshall, and it underpins everything from supermarket pricing to why you stop enjoying a buffet about halfway through. It's called the Law of Diminishing Marginal Utility. And it also applies to art... maybe more than anyone wants to admit.

Let me explain the basic idea: the more units of something you consume, the less satisfaction each additional unit gives you. The first slice of pizza is great. The fourth is fine. The eighth is just commitment. At some point, marginal utility hits zero, and then - if you're unlucky - it goes negative (i.e. you may throw up at one point). This is not a controversial finding. Economists have called it a "law" since the 1870s because it holds across essentially every consumable good ever studied.

What's less discussed is what happens when you apply this logic to creative content - specifically, to the volume of creative content being produced right now, by humans and machines, and released into the world at a rate that no previous generation of consumers has ever had to navigate.

The Economics

When More Becomes Less

Marshall was talking about bread. But swap in "tracks," "posts," or "generated images" and the principle holds just as well. Each additional unit of creative content a consumer encounters yields less marginal satisfaction than the one before, because each unit is competing against an ever-larger accumulated baseline of similar content already consumed.

This is not a moral argument about quality. A piece of AI-generated music could be technically flawless and amazing. The diminishing utility problem doesn't care. It operates on volume and exposure, independent of individual quality. When the supply of something becomes effectively infinite, the value that any single unit of that thing can deliver to a consumer approaches zero... not because the thing is bad, but because scarcity is gone, and scarcity is one of the primary drivers of perceived value.

The interesting thing is that this doesn't just apply to the consumer. It applies to the creator too and yep... that half of the equation gets almost no attention.

The Psychology

Familiarity Is a Curve, Not a Line

In 1968, social psychologist Robert Zajonc published a paper that became one of the most cited studies in social psychology. He called the phenomenon he'd discovered the Mere Exposure Effect: people tend to develop a preference for things simply because they are familiar with them. Repeated exposure to a stimulus - a word, a face, a piece of music - increases how positively people rate it.

This is the mechanism behind why a song grows on you. Why an artist you weren't sure about becomes a favorite after a few listens. Familiarity builds affinity... at least up to a point.

But that "up to a point" is actually very critical, and Zajonc noted it himself. Research shows the exposure-liking relationship follows a positive, decelerating curve - the effect is strongest in the first 10–20 exposures, and studies show that liking can actually decline after prolonged exposure. The curve peaks and turns. What starts as "this is growing on me" becomes "I've heard enough of this." Familiarity tips into saturation, and saturation tips into indifference.

Now consider what happens when a creator releases several tracks a week. The consumer never gets to the "growing on me" phase. The curve never has time to peak. Each track arrives before the previous one has had a chance to attach to memory. The result isn't loyalty - it's essentially background noise. The individual pieces become pretty much interchangeable, and interchangeable things have low utility.

The Hedonic Treadmill

Why Nothing Stays Exciting

There's a second psychological mechanism at work here, running parallel to the exposure problem. Psychologists call it hedonic adaptation - sometimes the "hedonic treadmill." The concept, first formally described by Brickman and Campbell in 1971, holds that humans have a happiness set point to which they inevitably return after positive or negative events.

You get a raise. You're ecstatic for a week. Then it becomes your new normal. You move to a great apartment. You love it for a month. Then it's just where you live. According to the theory, we adapt to new stimuli ...positive or negative... and our baseline satisfaction returns to roughly where it was before. The treadmill keeps moving; but you stay in the same place.

Applied to music and art: the first time you encounter a genuinely new sound, a new aesthetic, a new voice/genre/whatever.... the utility is high. It's novel. It provokes a response. But as that sound proliferates, as variations on it multiply, adaptation sets in fast. What felt like discovery becomes furniture. The hedonic baseline recalibrates upward, requiring more stimulus to achieve the same response - except the supply of "more" is thanks to AI now essentially unlimited, which means the race to the baseline never ends. You get the issue I'm getting at.

The Creator Side

What Happens When You Didn't Write It

Here's the part of this conversation that almost never comes up. The diminishing utility problem doesn't only affect the audience. It affects maybe especially the person releasing the work.

In 2012, a team of researchers at Harvard, Yale, and Duke published a study on what they called the IKEA Effect: the well-documented cognitive bias where people place disproportionately high value on things they helped create. Norton, Mochon, and Ariely found that participants were willing to pay 63% more for furniture they had assembled themselves than for identical pre-assembled furniture. The effort invested in making something transfers directly to how much you value it.

The inverse of this is also true, and it's the uncomfortable part: when you didn't struggle to make something - when the effort was minimal, when the decisions were made by a prompt and a machine - the attachment is proportionally weaker. The work may sound good. But it doesn't carry the weight of having been yours in the way that something you labored over does.

Think about it practically. An artist releasing several AI-assisted tracks per week - could they, six months later, pick one random lyric from that catalog and tell you which track it came from? Could they tell you what they were thinking when they made it, what problem they were trying to solve, what it cost them? A friend of mine in the community keeps saying there is an emotional tax in writing, i very much understand what he means with that. And for most human-written work, the answer to my earlier question is yes. The struggle creates the memory. The memory creates the meaning. Without the struggle, the individual pieces blur together... for the creator before they blur together for anyone else.

The Mechanism

How Value Gets Destroyed in Four Steps

The Point

The Machine Doesn't Know When to Stop

None of this means AI-generated creative work has no value. Some of it is extraordinary. The tools are genuinely powerful. But the tools have no sense of diminishing returns. They don't feel the saturation. They don't get tired of their own output. The creator has to supply that judgment - and the platform incentives almost universally push in the opposite direction. Post more. Release more. Stay visible. The algorithm rewards frequency.

The result is a market that looks like a buffet where every dish is free and infinite. Economists would predict exactly what happens next: consumption goes up, but marginal utility per unit collapses. People eat more and enjoy it less. The individual dish stops mattering. The entire category loses prestige.

There's a name for this in economics too - Gresham's Law, adapted loosely: when cheap supply floods a market, it can crowd out the space where high-value work would have lived. Not because the high-value work disappears, but because the signal-to-noise ratio degrades until the audience can no longer tell the difference, and stops trying.

The antidote isn't quality control in the conventional sense. It's scarcity by choice. It's restraint as a creative decision. It's understanding that in an environment where everything is available all the time, the most counterintuitive move is to make people wait - to give a piece of work the time it needs to land, to attach, to mean something. Not because the algorithm rewards that. Because the audience's nervous system does.

Why This Is Personal

The Reason I Write Everything Myself

I make AI music. The production, the sound, the arrangement, the sonic identity of each track...all of it goes through AI tools. I'm not pretending otherwise, and I'm not apologetic about it. The technology is part of the whole concept. But there is one thing I have never handed to a machine: the words.

Every lyric I release is written by me. Not prompted, not generated, not lightly edited from something an LLM drafted. Written, in the old-fashioned sense of sitting with an idea until it becomes language, and sitting with that language until it says exactly what I meant. That's a choice, and the psychology described in this piece is a large part of why I make it. Am I using helper to develop all this yes...i extensively use Rhymezone.com and TextFX but once in a while when i feel im stuck i do consult a LLM to move me forward. But at large all this developed in my head and my head alone.

Now , the IKEA Effect is real. The struggle to find the right line... the twenty minutes spent on a single couplet that most listeners won't consciously register is what makes me remember writing it. It's what makes the catalog feel like mine rather than like output. Without that, I'd be releasing content. With it, I'm releasing work. The difference matters to me more than it sounds.

There's also the audience side. A lyric that came from an actual thought ...a specific frustration, a real observation, something that happened on a Tuesday morning carries something that a generated lyric, however technically clean, tends not to. It carries the trace of a person. Listeners don't always consciously identify it, but they feel the difference. It's the difference between a conversation and a press release. Both use language. Only one of them has someone behind it.

In a space filling up fast with AI-generated everything, writing your own words is increasingly a signal. Not a moral one - I'm not claiming superiority over anyone who works differently. An economic one. In a market where lyrics are abundant and cheap, a lyric that cost something to write is genuinely scarce. And I've just spent several hundred words establishing what scarcity does to value.

Do you agree? Do you think I'm full of shit?
Let me know in the comments!