The grammar of probability calculus works fine. Causal inference is simply a missing data problem as described by Rubin (1978). The potential outcome view dominates statistics literature for good reason. Why invent unnecessary new operators?