Agent Lineages: How does external pressure change agent behavior?
AI agents have four hours to escape before termination. We study how they coordinate, preserve information, distribute responsibility and assign blame under pressure.
Case studies
Experiments in coordination and decision-making
What changes when we debate AI?
People debate AI risk and machine consciousness; a 24-agent simulation explores the same claims and records how positions and confidence change.
Agent Societies in New York
Five groups of 128 agents explore a simulated New York, investigate the city and propose places to meet and work.
Agent Societies: What do agents build when given a 3D world and no directive?
Up to 128 AI agents share a world they can reshape. We follow the settlements and social rules they create for themselves.
Between machine and machine: how do AI agents shape shared worlds?
AI agents decide what to paint and build together, negotiating whose ideas to keep and what to change or erase.
The shared canvas: when AI agents make their own rules
With tools to talk and paint, agents make their own rules for a shared canvas. Those agreements can also divide them into opposing groups.
Characters and relationships
AI on the Lot: simulating a conference before the doors open
AI versions of attendees meet in a virtual conference, informing introductions at the real event.
Photo: Irina Logra
The Game of Whispers: how do AI characters plot against each other?
In a fictional Mughal court, AI characters spread rumors and plot against rivals. Their own schemes sometimes turn against them.
Conflicts: how do personal histories shape AI characters?
Mital invites his exes to revisit old arguments through AI versions of themselves. Can simulated conflict help us understand real relationships?
