A caveat regarding the unfolding argument: implications of plasticity.

The unfolding argument claims that causal structure cannot matter for consciousness because any recurrent neural network can be replaced by a feedforward network with the same input-output behavior. This paper shows a boundary condition: when a network has rapid plasticity—weights that change quickly based on history—the equivalence fails. Mathematical proofs demonstrate that such systems encode information in ways a static feedforward network cannot capture, including history-dependent dynamics, complex temporal encoding, and perturbational instability. The results do not prove that recurrence or plasticity is necessary for consciousness, but they show that the unfolding argument does not block empirical tests of whether these properties matter.

Study at a glance

Abstract

The unfolding argument in the neuroscience of consciousness posits that causal structure cannot account for consciousness because any recurrent neural network (RNN) can be "unfolded" into a functionally equivalent feedforward neural network (FNN) with identical input-output behavior. Subsequent debate has focused on dynamical properties and philosophy of science critiques. We examine a boundary condition on the unfolding argument for RNN systems with rapid plasticity in their connection weights. We demonstrate through rigorous mathematical proofs that rapid plasticity negates the functional equivalence between the RNN and the FNN. Our proofs address history-dependent plasticity, dynamical systems analysis, information-theoretic considerations, perturbational stability, complexity growth, and resource limitations. We demonstrate that neuronal systems that possess properties such as plasticity, history dependence, and complex temporal information encoding have features that cannot be captured by a static FNN. We show that plasticity is a concrete instance of lenient dependency between behavioral and internal observables, restoring empirical testability to theories that incorporate plasticity on perception-relevant timescales. Our results do not establish that recurrence, plasticity, or process are necessary for consciousness; they establish that the unfolding argument does not preclude empirical investigation of whether these properties matter.