December 2025 Update

This post is part of our monthly supporter updates and reflects product development and thinking at the time it was written. Here is our latest update.

Hi all,

Here's what's new this month from Artificial Societies:

We made major research breakthroughs

Our models now reach 93% individual consistency (on par with human participants), 96% collective diversity (matching human opinion spread), 87% distribution accuracy, and 80% perfect ranking prediction against real-world survey data. We've deployed these improvements to our enterprise platform, making our personas measurably more human.

Building on this, we developed emotion analysis: simulated populations can now be surveyed across eight emotional dimensions, showing whether your messaging is eliciting the emotions you intend.

We developed Radiant — our enterprise platform

We built a standalone simulation platform for enterprise customers, named after the Prime Radiant in Asimov's Foundation. Radiant now supports team-level sharing, 3-way comparison across societies or narratives, and 3D network visualisation. We also improved our advertising concept testing, taking high-level concepts or storyboards and generating tailored outputs that reflect a brand's priorities.

We began new enterprise pilots

We're calling this our Forward Deployed strategy: deliberate exploration of all the different ways to provide human insights for businesses, from high-stakes narrative decisions to high-frequency tasks like choosing which ad or thumbnail to run.

One pilot involves building a tailored Artificial Society to help a client test narratives for the US AI policymaking community. Using our deep research infrastructure, we built 1,000+ AI personas of relevant members and mapped out how they influence each other.