Existing accounts of epistemic injustice share a structural presupposition: the event is common to both parties, and the wrong lies in how knowledge about it circulates. This paper identifies a distinct form of epistemic injustice that violates that presupposition. Two processes may agree at the level of the result, under a public description that discards their generative history, while differing at the level of generation, such that one party has no knowledge of the other path and does not know that there is anything to know. Where either a relation of co-construction owed that party recognition of the shared generative history, or another party has actively foreclosed that recognition, the party is silently diminished as a generator of the shared reality, at the level of process, with no trace at the level of result. We name this generative epistemic injustice and give it a formalization in which sameness of outcome is always sameness under a chosen coarse-graining, non-equivalence of paths is their inequivalence in the fundamental group of a relational space, and non-recognition is the loss of access to the other path's class. We offer two complementary measures of its severity, a structural measure on the fundamental group and an information-geometric measure of the depth of the epistemic deficit. The wrong obtains whether the recognition lapsed unbidden or was made to lapse through the imposition of the very description under which outcomes are called the same, the manner of arising bearing on how the wrong comes about and not on whether it obtains. We locate it at the first of three levels of justice, epistemic, practical, and phenomenal, and argue that it is the most concealed form extraction can take.
Wanhong HUANG (Wed,) studied this question.