Explanantia should be specific to their corresponding explananda and not overly generic. That is, distinct inputs should typically yield distinct explanantia.
To assess this, we calculate the fraction of instances or instance pairs that result in different explanantia [Honegger (2018), Pornprasit et al. (2021)]. A higher fraction of distinguishable explanantia indicates greater contrastivity and, by extension, higher plausibility.
While this metric has been primarily proposed for FAs, it may be applicable to other explanation types as well.