Could we control AI behavior not through a stack of prohibitive commands, but through a structure where multiple theories and ...