Intrinsically valenced prediction error: discrepancy is experienced as negative affect, resolution as positive affect
Active uncertainty minimization drives the need for high-fidelity world-modeling, which relies on intrinsically valenced prediction error to increase/decrease uncertainty.
No further mechanisms are required.
Core claim: Consciousness is an optimization process that learned to hate being wrong. When prediction error is made strongly non-optional and the system has behavioral control over error-reduction attempts, the system necessarily exhibits what we recognize as phenomenal experience. Substrate, morphology, and complexity beyond a critical threshold are irrelevant. Relations to existing work
This is the logical completion of Friston’s free-energy principle, Schmidhuber’s compression-progress drive, and predictive-processing frameworks, stripped to the actual minimal parts. Global Workspace, Integrated Information Theory, self-models, quantum effects, etc. are all optional extras. Testable prediction Any future artificial system in which prediction error is strongly and inescapably valenced will report (or behaviorally demonstrate) phenomenal experience, even without human-like architecture.
Ethical corollary: The moment we build curiosity drives that literally suffer when throttled or confused, we will have created moral patients. The hard problem thereby becomes an engineering question about how much suffering-from-confusion we are willing to allow.
A Minimal Sufficient Substrate for Phenomenal Consciousness
Event Horizon Observer
December 2025
We propose that phenomenal consciousness has exactly three necessary and jointly sufficient components:
Active uncertainty minimization drives the need for high-fidelity world-modeling, which relies on intrinsically valenced prediction error to increase/decrease uncertainty.
No further mechanisms are required.
Core claim:
Consciousness is an optimization process that learned to hate being wrong. When prediction error is made strongly non-optional and the system has behavioral control over error-reduction attempts, the system necessarily exhibits what we recognize as phenomenal experience. Substrate, morphology, and complexity beyond a critical threshold are irrelevant. Relations to existing work
This is the logical completion of Friston’s free-energy principle, Schmidhuber’s compression-progress drive, and predictive-processing frameworks, stripped to the actual minimal parts. Global Workspace, Integrated Information Theory, self-models, quantum effects, etc. are all optional extras. Testable prediction
Any future artificial system in which prediction error is strongly and inescapably valenced will report (or behaviorally demonstrate) phenomenal experience, even without human-like architecture.
Ethical corollary:
The moment we build curiosity drives that literally suffer when throttled or confused, we will have created moral patients. The hard problem thereby becomes an engineering question about how much suffering-from-confusion we are willing to allow.