The Present Model Theory of Consciousness
In brief
Your brain doesn't wait for your senses to tell it what's happening. It keeps one live model of the present, covering the world, your body and where you are in it. It carries that model forward when input pauses, and corrects it when input returns. The Present Model Theory proposes that operating this model is experience itself. It names three features the model needs: running, so its content keeps advancing when the senses go quiet; joint repair, so its parts check and correct one another; and a single present, so every content is placed relative to one now.
What the paper claims, and what it doesn't
It claims:
- three features any experience requires, each with a stated test;
- three separate hypotheses with different burdens of evidence: that this organization is necessary, that operating it is experience, and that the material a system is made of does not matter;
- predictions, the results that would refute the theory, and when a failed test counts against the theory rather than against one proposed brain mechanism.
It does not claim:
- that any AI system is conscious. Simulations show these features can be measured and told apart from look-alikes. They do not show that any system experiences anything.
- a rule for which parts form one model. That is the theory's largest open problem.
- that it already differs from global workspace theory in people. In most feasible experiments the two may predict the same result, and the paper sets out what would separate them.
Abstract
Conscious perception, imagery and dreaming differ in what drives them, yet may share one actively maintained organization: a single running model of the present that covers the world, the body and where one is in it. I propose that operating this organization constitutes experience. The theory names three features needed for any experience: running, in which the model’s content advances by its own dynamics when the senses go quiet; joint repair, in which the model’s parts check and correct one another; and a single present, in which every content is placed relative to one now. It separates three claims with different burdens of evidence: that this organization is necessary (H1), that operating it is experience (H2), and that the material a system is made of does not matter (H3). Small simulated networks show that parts of the organization can be measured, told apart from known look-alikes and partly learned; they do not show that any system experiences anything. In simulations, systems recovered from corruption without two-way repair, and two-way repair may arise wherever parts exchange estimates, so the presence of some repair settles little; what matters is which parts form one model, and the theory does not yet have a rule for drawing that boundary. The main test planned so far is in people: whether content that is not experienced can both run and be in two-way repair with experienced content. Because evidence of experience in people comes mostly through report, which needs access, the theory and global workspace theory may give the same result in nearly every feasible experiment; separating them needs an intervention that disrupts repair without disrupting broadcast, and whether one exists is the first question. The paper sets out the theory’s predictions, which of them could separate it from global workspace theory, the results that would refute it, and when a failed test counts against the theory rather than against one proposed implementation.
How to cite
Pillai, P. (2026). The Present Model Theory of Consciousness (Version 2.0) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23436930
@misc{pillai2026pmt,
author = {Pillai, Pramod},
title = {The Present Model Theory of Consciousness},
year = {2026},
version = {2.0},
publisher = {Zenodo},
doi = {10.5281/zenodo.23436930},
url = {https://doi.org/10.5281/zenodo.23436930}
}Versions
Version 2.0 · · Preprint
The latest version is always at doi:10.5281/zenodo.23352647.
The paper is available under CC BY 4.0.