Materialism › Computational
O'Reilly-Shah's State Space Theory and Computational Dynamic Monism
Consciousness as Self-Referential Dynamical Embedding in Plastic Networks
O'Reilly-Shah proposes that phenomenal consciousness is hierarchical delay coordinate embedding in plastic recurrent neural networks: recurrent feedback lets the brain reconstruct the dynamical history of its environment and, at the top of the hierarchy, of its own embedding activity. His metaphysics, Computational Dynamic Monism, identifies consciousness with this temporally extended process itself. He argues the identity dissolves the hard problem, since the remaining gap is only the difference between describing a process and being it.
Vikas N. O'Reilly-Shah
What consciousness is
Phenomenal experience is identical to hierarchical, self-referential delay coordinate embedding carried out by plastic recurrent networks, a temporally extended computational process, not a static neural state, functional role, or intrinsic property of matter.
Mind and body
Identity between experience and a physical process, on the model of lightning and atmospheric electrical discharge. One ontological category, physical process, underwrites both the scientific description and the subjective character available to the system.
Mechanism
Delay coordinate embedding, grounded in Takens’ embedding theorem, in which time-delayed copies of a signal reconstruct the full state space of the system that produced it. Recurrent feedback supplies the delays.
Core claims
- Hart and colleagues proved that contracting recurrent networks driven by structured input embed the input dynamics through generalized synchronization.
- Uribarri and Mindlin, and Ostrow, Eisen, and Fiete, found the geometric signature of time-delay embeddings in trained recurrent networks and neural sequence models.
- Bats and rodents represent three-dimensional space on a topological torus, an example of perception mapping regular environmental dynamics onto low-dimensional attractors.
- O'Reilly-Shah, Selvitella, and Schurger argue that no static feedforward network can stay equivalent to a plastic recurrent network over unbounded time, which weakens the unfolding argument.
- The theory predicts that metamers, categorical perception, and discrimination thresholds reflect a topology-preserving embedding working with finite predictive resources.
Criticisms
The unfolding argument of Doerig and colleagues, and its Turing-completeness generalization by Herzog and colleagues, holds that any computable function can be reimplemented in a system with different causal structure. O'Reilly-Shah accepts this for input-output functions but argues it loses force once plasticity and environmental coupling must be matched indefinitely.
Implications
- AI consciousness: Conditional
- A system with the same plastic, processual trajectory would be conscious, but O'Reilly-Shah considers exact equivalence extremely unlikely given unpredictable input and chaotic sensitivity, and static feedforward systems do not qualify.
- Survival beyond death: No
- Since consciousness is identical to an ongoing physical process in recurrent networks, it follows from the view that it ends when that process ends.
- Virtual immortality: Unlikely
- A copy would need to match the original’s plasticity, environmental coupling, and trajectory indefinitely, which O'Reilly-Shah regards as practically unattainable.
Related theories
- Markwell’s Process Identity Theory (Overlap)
- Integrated Information Theory (Contrast)
- Qualia Space (Compatible)
- Lamme's Recurrent Processing Theory (Compatible)
Sources
- Delay Coordinate Embedding as Neuronally Implemented Information Processing: The State Space Theory of Consciousness, Vikas N. O’Reilly-Shah (2025)
- A Caveat Regarding the Unfolding Argument: Implications of Plasticity for Computational Theories of Consciousness, V. N. O’Reilly-Shah, A. Selvitella, and A. Schurger (2025)
- Computational Dynamic Monism: Process Metaphysics for the State Space Theory of Consciousness, Vikas N. O’Reilly-Shah (2026)
- State Space Theory as a Unifying Framework for Consciousness, Vikas N. O’Reilly-Shah (2026)
- Smooth Embeddings in Contracting Recurrent Networks Driven by Regular Dynamics: A Synthesis for Neural Representation, V. N. O’Reilly-Shah and A. M. Selvitella (2026)
- Detecting Strange Attractors in Turbulence, F. Takens (1981)
- Embedding and Approximation Theorems for Echo State Networks, A. Hart, J. Hook, and J. Dawes (2020)
- Dynamical Time Series Embeddings in Recurrent Neural Networks, G. Uribarri and G. B. Mindlin (2022)
- Delay Embedding Theory of Neural Sequence Models, M. Ostrow, A. Eisen, and I. Fiete (2024)
- The Unfolding Argument: Why IIT and Other Causal Structure Theories Cannot Explain Consciousness, A. Doerig, A. Schurger, A. Hess, and M. H. Herzog (2019)
- The Multiple Realization Book, T. W. Polger and L. A. Shapiro (2016)