IIT Traded the Pineal Gland for Causal Units

Abstract

Integrated Information Theory identifies each experience with the $\Phi$-structure unfolded from a maximal causal substrate. Its calculation begins from a transition probability matrix over chosen units and interventions. I identify two limits on the claim this structure is intrinsic. First, a complete list of properties shared by every experience does not entail IIT's physical postulates are sufficient for consciousness. A cross-domain bridge is required, and IIT's explanatory identity is one such bridge. Second, system integrated information $\phi_s$ depends on the product decomposition of a transition system. Two conjugate presentations of the same four-state deterministic dynamics yield $\phi_s=0$ and $\phi_s=2$ ibits for the full two-unit candidate. IIT can reject one presentation after a micro basis has been fixed, but the unfactorised transition graph does not fix that basis. Its grain framework starts from the finest units that can be observed and manipulated. Hence any consciousness verdict inherits an operational decomposition, unless IIT derives the starting micro basis from the substrate. IIT avoids software-level computational dualism while retaining an analogous interpreter dependence in its causal model. If AI traded the pineal gland for a Turing machine as interpreter, then IIT traded it for a choice of causal units.