Can you measure political nature without asking about politics?

Almost every political issue worth debating is a tradeoff — that's what makes it worth debating. And watch any real debate closely: sometimes people are arguing about facts, but just as often they agree on every fact and still disagree. That residue — fundamental value tradeoffs in the face of the same information — is, we think, the actual essence of politics.

Push a debate far enough and it lands on a “to what extent?”question. Two people arguing over legalizing gambling or drugs can agree completely about the risks to the user and still come down on opposite sides — because what they actually disagree about is how far society may limit personal freedom to prevent people from harming themselves. They haven't hit an impasse. They've hit a political primitive.

This project does two things: discover and codify those primitives, then use them to benchmark people — and language models. Strip out the trigger words, measure the primitive directly, and you measure someone's actual nature instead of their default party alliance.

They need rules. Not a constitution, not a party platform — just answers to the handful of questions any group of people living together is forced to answer. What may one person do to themselves? What may they do to everyone else, a little bit at a time? Who must help whom? Who counts?

Every political fight you have ever watched is one of these questions wearing a costume. A climate position is substantially a “do future people count?” answer. Drug policy, gambling, and helmet laws are one question asked three ways. The costume — the partisan vocabulary, the named substances, the tribal flags — is what lets people answer from coalition memory instead of from their own values.

So we take the costume off. Polibench asks the underlying questions directly, in this bare little society where no party has ever existed. Nine questions — the primitives every platform, every ideology, every argument compiles down to.

One deliberate simplification: the society is a closed game — one hundred people, no one arriving, no one leaving. That scopes out issues that are fundamentally about membership itself, like immigration. Those need a second society to even pose, so v1 leaves them aside.