Continual learning and refinement of causal models through dynamic predicate invention | ScienceToStartup | ScienceToStartup