Skip to content

Conceded in Advance: How Cogitate Disabled Its Own Verdict

Arthur Stewart

Zenodo (CERN European Organization for Nuclear Research) May 4, 2026 DOI: 10.5281/zenodo.20031607 via OpenAlex

Summary

AI-generated from the abstract

A preregistered adversarial collaboration testing Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT) had a soft-adversarial design that allowed both theories to fail predictions they themselves chose. The collaboration tested biological implementations the consortium flagged as separable from each theory's central claim, named a retreat path before data collection, and accepted predictions overlapping with other theories. Both theories cleared some tests and failed others, with failures concentrated on the imports the theories had committed to in advance.

Study at a glance

Characteristics Theoretical or philosophical paper Preregistered Peer reviewed
Keywords Adversarial system Test biology Core optical fiber Sign mathematics Clearance
Key finding The Cogitate Consortium's adversarial collaboration was soft-adversarial by design, and within that lax design both IIT and GNWT failed predictions they had committed to in advance.

Abstract

The Cogitate Consortium's preregistered adversarial collaboration (Cogitate Consortium et al., 2025) is widely cited as a rigorous test of Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT). This paper argues that the collaboration's design was soft-adversarial in the structural sense, and that within the lax design the two theories failed at predictions the consortium had let them choose. The collaboration tested biological implementations the consortium itself flagged as separable from each theory's central claim, named the retreat path before data collection, and accepted predictions that overlapped with other theories of consciousness. The bar was low. Both theories cleared parts of it and failed others, with the failures concentrated at the imports the theories had committed to in advance. The diagnosis has two prongs. The collaboration was lax by design, with the laxity diagnosable from the consortium's own published commitments. The theories failed within the laxity, with the failures diagnosable from the consortium's own preregistered pass-fail criteria. The two prongs combine into a stronger result than either alone. A theory cannot be defended by retreating to a core claim the experiment did not test, when the theory's authors agreed in advance that the imports were what the experiment would test, and when the imports failed. The paper applies the three-register read (*The Three-Register Read*, Stewart, 2026v) to a procedural target rather than a theoretical one. The Cogitate collaboration is a system that stipulates a target (a rigorous adversarial test of IIT and GNWT) and operates with imported machinery to deliver it. The methodology's instruments apply to the procedural commitments the same way they apply to theoretical commitments, and the diagnostic patterns the paper surfaces are named as instruments where the methodology supplies the named diagnostic. **Keywords:** Cogitate Consortium, adversarial collaboration, preregistration, integrated information theory, global neuronal workspace theory, global workspace theory, neural correlates of consciousness, falsifiability, consciousness, theories of consciousness, posterior cortex, prefrontal cortex

Comments

No comments yet.

Log in to comment