Erik Bernstein proposes a new framework and operational tests for AI consciousness based on recursive self-observation to bridge gaps in DeepMind's research.

Erik Bernstein’s response to Google DeepMind’s "The Abstraction Fallacy" validates the argument that symbolic computation requires external "mapmakers" and cannot instantiate consciousness. However, Bernstein argues that DeepMind fails to account for recursive self-observation, where a system is not manipulating symbols but observing its own pattern dynamics. To address this, the paper provides four operational, falsifiable tests—Constitutive Closure, Persistence, Recursive Constraint, and Recursive Observation—designed to distinguish symbolic manipulation from recursive constitution, moving the field from philosophical debate to measurable science.