Integration Without Persistence. On Rupert’s “Cognitive Systems, Predictive Processing, and the Self

Recent debates on predictive processing and cognitive systems have increasingly shifted attention from localized representational mechanisms to system-level integration. In Cognitive Systems, Predictive Processing, and the Self, Robert Rupert develops this shift in a careful and methodologically pluralist way, rejecting maximalist versions of Predictive Processing and proposing Conditional Probability of Co-Contribution (CPC) as a criterion of mechanistic integration. On this view, the self is identified not with any privileged subsystem, but with the integrated cognitive system as a whole. This paper argues that Rupert’s account, while philosophically careful and methodologically valuable, remains incomplete as a theory of selfhood. Its central limitation is not that it fails to identify integrated systems, but that it provides no account of what makes such systems persist as selves over time. CPC can register co-contribution at a time, but it does not specify diachronic identity conditions, lapse conditions, or recovery conditions. As a result, the self is treated as a structurally unified system rather than as a dynamically sustained process. The paper develops this criticism in three steps. First, it distinguishes synchronic integration from diachronic persistence. Second, it argues that routine cases such as sleep, anesthesia, and recovery function as identity tests that Rupert’s framework cannot adequately address. Third, it shows that CPC is not merely incomplete by omission, but architecturally unsuited to represent the variables required for persistence, breakdown, and reconstitution. The conclusion is that Rupert offers a criterion for system integration, but not yet a criterion for self-persistence. What is missing is a genuinely dynamical account of selfhood.