The Viable System Model & Multi-Scale Agency

A LessWrong researcher is applying vintage cybernetics to AI safety. Jonas Hallgren is reinterpreting Stafford Beer's Viable Systems Model—a fifty-year-old organizational theory on managing complexity through hierarchy—as a potential blueprint for safely scaling AI agents. The model explains how organizations divide work among semi-autonomous teams while maintaining alignment through coordination layers. Hallgren argues similar structures could govern hierarchical AI systems, where sub-agents operate with local independence yet stay coordinated at higher levels. He's reframing the theory through Active Inference, a mathematical framework from neuroscience, and acknowledges his interpretation is provisional—but positions it as a bridge between management science and AI alignment research.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton