Mediation
Decomposing a treatment’s total effect into an indirect effect operating through an intermediate variable (the mediator) and a direct effect through other channels — the formal way to ask why a treatment works. Modern causal mediation defines controlled and natural direct/indirect effects in Potential-Outcomes terms. Identification is demanding: even with a randomized treatment, decomposing the effect requires no unmeasured confounding of the mediator–outcome relationship (and no treatment-induced mediator–outcome confounding) — assumptions that randomization of the treatment alone does not deliver. Hence the emphasis on sensitivity analysis.
Relied on by
Mechanism analysis in program evaluation; any study seeking direct/indirect effects rather than a total effect.