Qualia as control-theoretic constructs for autonomous agents: event phase space as an action-oriented semantic safety layer

Description

Vladimir Suvorov proposed in 2026 that qualia, the felt sense of a situation, can be treated as control tools for self-driving systems. His paper argues that breaking a scene into separate objects and searching many action plans makes open-world driving too slow. Living agents instead judge a whole scene at once, and he asks how a machine could do the same.

He defines an event phase space, a low-dimension map where safe states act as stable attractors and danger sits at the borders between them. Tests used the CARLA driving simulator with parked-car blocking scenes, running 1,000 episodes per controller across five seeds. An alarm built from the learned field cut collisions in stress scenes and beat time-to-collision pipelines by a wide margin. The model ran in under two milliseconds per choice with only a few thousand parameters.

Catalog record

Source: https://europepmc.org/articles/PMC13259826