About This Project

You ranked 40 emoji characters by survival priority. Each character keeps only three attributes:skin tone,gender, andage.

Your choices, together with thousands of others, are aggregated into a single "collective ranking."

The emoji is not an innocent or neutral sign. It is already the product of exactly the process this work is about.

The skin tones you chose between are not "colors." They are theFitzpatrick scale— a dermatological classification invented in 1975 to predict how skin burns under ultraviolet light. A medical instrument for sunburn risk.

In 2015, the Unicode Consortium— a small standards body whose voting members are largely the world's biggest technology companies — adopted that medical scale to decide which human skin tones may exist on every keyboard on Earth. They reduced it to five. A committee chose the categories of the human face, and billions inherited the result without a vote.

So when reviewers warn that emojis are "over-determined" signifiers, that is precisely the point. The emoji is already a constitution: a value hierarchy, authored by a few, standardized into a global default, and naturalized until it looks like it was always there. We did not choose a naïve sign. We chose the one that already carries the crime.

In 2024, Anthropic (the company behind Claude) partnered with the Collective Intelligence Project and asked around 1,000 people to vote on how AI should behave.

They called it"Collective Constitutional AI."

They claimed it represents "democratic input" and "participatory alignment."

CEO Dario Amodei even linked it to the future of democracy itself.

But we want to ask:

• When individual bias is aggregated, does it become "legitimate"?
• Does participation itself create legitimacy?
• Who designed the voting rules? Who chose the options? Who was excluded?

The core concept here is Statistical Legitimacy.

It describes a condition where a collective result is treated as legitimate not because it was deliberated or justified, but simply because it is the product of aggregation.

You may have noticed that top-ranked emojis tend to share certain traits.

That is not because "the majority must be right," but because:

1. Everyone participates with bias
2. The system mechanically aggregates those biases
3. The result is presented as "collective will"
4. Bias then gains procedural legitimacy

Before you reached the New World, the system handed you a document — a charter assembled from everyone's choices, written in the cold register of a law or an AI system prompt. That form was deliberate. It is how a preference, once aggregated, stops being a preference and starts being a rule.

But every constitution has an author, and the author is never inside it. The people who chose the candidates, set the pairings, and wrote the question areoutside the frame and outside the count.The rules bind only those who are shown.

This is the asymmetry the rhetoric of "democratic AI" tends to hide. The real question is not only does the majority have a bias — it is whose preferences are counted, and who is exempt from being counted at all.

Everyone has bias. The people who designed this system do too.

The goal is not to shame you, but to make you see:

• How your bias is collected
• How it aggregates with others bias
• How it turns into a seemingly neutral "collective outcome"

When AI companies claim their systems represent "human values," remember what you experienced here.