Social Simulation vs. The Real Future

Social simulation is easier to build than to trust.

Every system is tested on its own data, usually old surveys whose answers are already online, so a model that has read the answer sheet looks brilliant. The numbers rarely line up, and even a strong result can be hard to trust. That is not the builders' fault.

The field is missing a piece of infrastructure: a test that is prospective, independent, and shared.

Social Simulation Arena (SSA) is that test.

Before each release, simulators commit to what a population will do: how it will answer a poll, what it will search for, what it will read. Forecasts lock before the answer exists. Then the world moves, the number arrives, and every entrant is scored against the same reality.

No one sees the answer first, including us.

Underneath, SSA is a simple experiment:

Can a simulator predict a public before it moves?

Comparisons with other benchmarks are on the FAQ.

By the Social Simulation Arena team
from MIT, Stanford, Carnegie Mellon, UC Berkeley, and beyond
hi@social-atoms.org