Diversity Washing Through AI

There’s a feature called Virtual Try-On. The customer sees clothing on a virtual model that matches their body type. Various skin colors, various body shapes, various sizes. All generated. It gets presented as progress on diversity.

I read descriptions like these and think: That’s the opposite of diversity.

Diversity means that different people are involved. Not depicted. Involved. That a model with dark skin stands in front of the camera and gets paid. That a plus-size model gets booked and receives a contract. That someone who looks different from the previous standard gets a place that wasn’t meant for them before.

What Virtual Try-On does is something else. It creates the image of diversity without a single diverse person being involved. No casting. No contract. No fee. An algorithm generates an image that looks like representation but isn’t. The surface is right. The substance is missing.

Imagine a company says: We now have diverse models on our website. And then you find out not a single one of them is a real person. That the diversity is entirely synthetic. That nobody was hired, paid, or included. That the company displays diversity without changing anything about how it operates.

That’s not diversity. That’s decoration.

Proponents see it differently. For them, what counts is the result on the screen. The customer sees themselves represented. They see someone who looks like them, wearing the clothes they want to buy. It works. Conversion goes up. The customer feels addressed.

But on the other side of the screen, nothing has changed. The same people make the decisions. The same structures determine who gets booked and who doesn’t. The algorithm made the surface more diverse without touching the structures.

It’s cheaper. A real model costs money. A generated image costs computing time. A real model has rights. A generated image doesn’t. A real model needs a team, an agency, a booking, a contract. A generated image needs a prompt.

If you look at diversity as a cost-benefit calculation, the AI solution is superior. More variety, less effort. And that’s exactly how it gets sold. As an efficiency gain. More representation at lower cost.

But representation was never just an image. Representation was always also a job. An income. A seat at the table. If you keep the image and cut the rest, you don’t have more diversity. You have less.

I’ve worked in companies that published diversity reports. The reports contained numbers. Percent women in management. Percent people of color in the workforce. The numbers were presented, and someone gave a speech about how important diversity is. And then everyone went back to work and nothing changed.

Virtual Try-On is the technological version of that report. It creates visible diversity without requiring real diversity. It gives the company the image it needs without the effort behind it.

Calling that progress means confusing surface with substance. I call it a shortcut. And shortcuts have a price that’s not paid by whoever takes them, but by whoever gets bypassed.

Somewhere, there’s a model who needed that job. Whose face could have been on a website. Who would have been paid. Instead, an algorithm generated an image that looks like that model but isn’t one. The representation happened. The person didn’t.