Interesting. I am wondering what should be the observable in agents scenario (and therefore we could test whether a theory is correct), no matter EUM or active inference.