Does Integrated Information Theory (IIT) make experimental predictions about consciousness?
David Reichert, Rif Saurous
PsyArXiv December 5, 2025 preprint DOI: 10.31234/osf.io/gauqm_v1 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper |
|---|---|
| Key points | Argues that the experimental predictions used to test Integrated Information Theory do not appear to follow from its mathematical framework and likely cannot do so even in principle, meaning experimental outcomes have no bearing on the theory's hypotheses about phenomenal experience. Concludes that refuting this argument would require showing that IIT's mathematical formalism connects its consciousness hypotheses to concrete observable variables. |
Abstract
Integrated Information Theory (IIT; Albantakis et al., 2023) is a well-known theory of consciousness. In recent years, it has drawn vocal criticism (e.g., IIT-Concerned et al., 2023, 2025). Much of that criticism has focused on a subset of IIT’s predictions possibly being counterintuitive, untestable, or both: for example, IIT predicts that simple circuits of logic gates are conscious. At the same time, proponents (Tononi et al., 2025) as well as “neutral” parties (Gomez-Marin & Seth, 2025) have defended IIT on the grounds that it does make experimental predictions, which could falsify or corroborate the theory. Here, we consider several of these predictions and accompanying experimental work. We find that the predictions being tested experimentally do not appear to follow from IIT’s own mathematical framework, and likely cannot follow even in principle. If so, the experimental outcomes have no bearing on the validity of IIT’s hypotheses about phenomenal experience itself. We conclude that, to disprove this argument, IIT’s mathematical formalism would need to be shown to connect the relevant hypotheses about consciousness with concrete experimentally observable variables.