Estimating the probability of assigning treatments under a given context (i.e. propensity scores) would enable two useful features
Contextualized Causality: correcting for context-specific propensity allows us to directly estimate context-specific causal treatment effects
Assessing the Meaningfulness of Propensity: propensity can be understood as bias, where one group receives treatment more often than another. This bias is not necessarily bad if treatments affect groups differently, e.g. one group benefits from a treatment while the same treatment is harmful or neutral for another group. Estimating context-specific treatment effects along with context-specific propensity enables detecting treatment biases that are meaningful versus unsupported.
Estimating the probability of assigning treatments under a given context (i.e. propensity scores) would enable two useful features