Tsubamoto & Horii's Collective Predictive Coding model explains parent-infant brain synchrony

A new paper from researchers Yushi Tsubamoto and Takato Horii, posted on arXiv on May 8, 2026, introduces a computational model that may explain how parent-infant dyads achieve inter-brain synchrony (IBS) during real-time interactions. The model combines a POMDP formulation of active interoceptive inference with the Metropolis–Hastings Naming Game (MHNG) derived from the Collective Predictive Coding (CPC) hypothesis. In their setup, the parent observes the infant only through exteroceptive signals, while the infant directly senses its own interoceptive state. The two agents agree on regulatory actions via a shared communicative variable, with acceptance determined by a locally computable Metropolis–Hastings probability.

The agents have asymmetric generative-model knowledge: the parent knows how actions transform visceral states but must learn what the infant's body is communicating, whereas the infant perceives its visceral state directly but must learn how actions affect it. In a 6×6 visceral-state grid world, MHNG-mediated interaction regulated the infant's visceral state more adaptively than one-sided control conditions. Crucially, the two agents' posteriors became rapidly aligned—emerging far earlier than the convergence of their learned generative matrices. This indicates that representational synchrony does not presuppose fully shared world models, offering a minimal constructive account compatible with hyperscanning studies of parent-infant IBS and supporting CPC as a candidate computational basis for inter-brain alignment.