The separation principle: where do beliefs and desires come from?

On Monday, LessWrong highlighted a new essay that argues the traditional belief‑desire description of agents can be traced back to theorems from optimal control and reinforcement learning. The author invokes the separation principle—a result that lets a complex optimisation be solved by tackling sub‑problems independently—to show that beliefs correspond to an agent’s internal state estimates while desires map onto the optimal policy’s goals. By framing both as emergent properties of an optimal controller, the piece proposes a formal foundation for the belief‑desire model rather than treating it as a folk‑psychology shortcut. This perspective could reshape how disciplines from psychology to economics think about agency.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton