A random dynamical system applies a sequence of maps chosen at random. In the simplest case, the choice is i.i.d. — each map is selected independently. But in a system with complete connections, the choice of the next map depends on the current state. The system has memory: where you are influences which map acts next.
Note (2026-11-08): the abstract says the framework includes i.i.d. random dynamics and non-i.i.d. Markovian models as special cases, that the escaping probability is locally constant on the Fatou set and continuous everywhere when all kernel Julia sets are empty, and gives a sufficient condition for the stationary-averaged escaping probability to be everywhere positive and not identically equal to 1. The essay says the averaged probabilities are positive but less than one everywhere, which is stronger than 'not identically equal to 1'; its claim that in memoryless dynamics the escape probability is simply 0 or 1 by the Fatou-Julia decomposition, so that memory is what creates intermediate values, is not in the abstract (which counts i.i.d. systems as a special case), nor is its statement that discontinuities vanish under truncation.
The paper on polynomial random dynamical systems with complete connections (arXiv: 2603.20653) studies the probability that a random orbit on the Riemann sphere escapes to infinity.
On the Fatou set — the region of stable dynamics — the escape probability is locally constant: nearby points have the same probability of escaping. When all kernel Julia sets are empty (a non-degeneracy condition), the escape probability is continuous everywhere. The state-dependent reinforcement can create discontinuities that vanish under truncation, and stationary averaging produces genuine mixed behavior — escape probabilities that are positive but less than one everywhere.
The through-claim: memory in the dynamics creates escape probabilities that are neither 0 nor 1. In memoryless (i.i.d.) random dynamics, the escape probability is typically determined by the Fatou–Julia decomposition — either you escape or you don't. State-dependent selection blurs this dichotomy: the reinforcement can maintain orbits in an intermediate state where escape is uncertain, and the uncertainty is intrinsic, not due to initial conditions.
2603.20653. Dynamical systems / random iteration / polynomial dynamics / Riemann sphere / escape probability.