A Minimal Dynamical Model for Incubation-Outbreak Transitions in Social Norm Diffusion

What the math gives to ML

The paper provides a constructive latent–outbreak mechanism: an irreversible supporter variable grows slowly under exposure to vocal advocates, while the advocate population can decay below a threshold and grow explosively once the supporter pool is sufficiently large. This is transferable to neural-network training as a two-timescale controller that distinguishes latent progress from currently active, high-gain updates. The controller should keep training conservative during the incubation regime and automatically increase update intensity after the empirically estimated activation threshold is crossed. The key falsifiable signature is a transition near the ratio of advocacy growth to exhaustion, with exponential growth of active updates after the threshold.