Skip to content
Shadow Theory

Shadow Theory / the consciousness programme

A perspective.
A boundary.
A physical realization.

Consciousness demands more than a description of behaviour. It demands an account of why a physical organization constitutes a perspective—and a way to determine which organization is actually there.

One programme / four publications

One foundation.
Three new investigations.

SPC-2 gives the framework its constitutive laws. Three connected investigations put its boundaries, learning procedures and realization criteria under precise mathematical pressure. Follow the argument all the way through.

The laws and the evidence belong together.

The founding monograph develops the philosophical distinction between awareness and its organized manifestation. SPC-2 then states explicit laws for qualifying perspectives, phenomenal structure and continuation. Its mathematics gives that position a form that can be inspected: physical realization, native predictive organization, joint realization and process identity each do a particular job.

The next question is unavoidable. How do we know that the boundary we have drawn belongs to the organization itself? How do we recover an interface when the mechanism is opaque? What evidence identifies a realization rather than merely a convenient description of it?

Papers 2–4 pursue those questions as one connected investigation. A constitutive law, an operational reconstruction and an identification certificate address different steps of the argument. Putting them together makes the framework more exact at each step.

Read the four constitutive laws ↗

A boundary must survive a change in how we look.

Paper 2 begins with a vulnerability in the original boundary construction. A target that is minimal before mixing can lose its apparent dependence after marginalization. The candidate graph can therefore change with the level at which a system is described.

The response is a deeper operational treatment: reconstruct complete joint experiments, declare the resources available to a severed simulator, and test whether a causal interface preserves the distinctions a surrounding context can expose. Exact masking examples, resource comparisons, quotient and composition results, approximation bounds and identification limits make the proposal concrete.

This development is visible alongside the preserved monograph. The original boundary law remains the foundation’s original statement; the new work explains where its diagnostics need to be stronger. A robust diagnostic constrains a successor boundary law without selecting a unique phenomenal assignment.

Read: What makes a boundary real? ↗

Learning an interface is a sequence of achievements.

A representation may have enough capacity to express an adequate interface while the learner never produces it. A learner may produce it while calibration chooses another candidate. A selected candidate may perform well on the declared held-out contexts while a different context exposes its failure.

Paper 3 separates those events. Its matched study evaluates complete interaction laws across a declared suite of opaque stochastic fixtures. The strongest learner passes 1,063 of 1,152 contexts and 37 of 48 fixtures. The fixture-level path through representational capacity, finite-budget production and selection is 48, then 39, then 37.

Those distinctions give the numerical results their meaning. The article follows the primary matched comparison, stage-specific failures, uncertainty and composition checks, and the separately labelled retrospective fitting study. It explains both what the learning procedure achieved and exactly which stronger conclusions its protocol cannot carry.

Read: Can an interface be learned? ↗

A realization is identified by what interventions reveal.

Paper 4 moves from a useful predictive representation to binary coordinate identification under a specified response model. The response family supplies geometric structure: when the prescribed conditions hold, it fixes the binary chart up to coordinate permutation and bit complementation. The paper supplies the identification argument and error certificates.

It then uses an explicit five-register construction to show why changing a binary encoding is not a harmless relabelling of the declared intervention mechanism. The stochastic continuation results make that obstruction precise. Finally, a bounded computation compares inherited SPC-2 assignments with an official IIT intrinsic-grain search, including its selected grains and numerical outcomes.

Microchart identification and macrograin selection remain separate tasks. A certificate for coordinates does not automatically choose the scale at which a constitutive law should operate. Keeping that distinction explicit is what lets the comparison teach us something about the two frameworks.

Read: What identifies a realization? ↗

Read for the argument. Stay for every detail.

Start with the three lead articles to see how the programme develops. Each opens directly into the full technical investigation: every source section, proof, equation, table, figure, appendix and reference is readable on the site. The section readers retain source numbering so a result can be followed consistently across the website and publication.

For the philosophical position, begin with the monograph’s source-to-experience problem and the argument for an additional manifestation law. For the new operational work, begin with Paper 2 and continue in order. The glossary, FAQ and programme search connect the vocabularies.

Three investigations / 49 complete reading sections

The publications, open in full.