The potential-outcomes (Neyman-Rubin) framework gives every unit two values, Y(1) with treatment and Y(0) without, and defines the individual causal effect as their difference. Since each unit reveals only the outcome matching the treatment it actually received, individual effects are unobservable by construction, a fact known as the fundamental problem of causal inference.
The framework turns causal estimation into a missing-data problem: averages like the ATE are recoverable, but only under assumptions about how treatment was assigned, because the assignment process decides which half of the table is missing.
