Participation in a system or dataset without decision-making power over how one's knowledge or cultural material is collected, used, or represented. Communities contribute data but remain absent from the governance of the infrastructure built on it.
So what? Addressing representational gaps in AI by scraping more data from underrepresented communities does not remedy the underlying power imbalance if those communities have no say in how their knowledge is used. Testers evaluating multilingual or culturally diverse AI systems should treat this as a structural quality concern.
Example: Attempts to address the underrepresentation of Global Majority languages in training data often proceed through ungoverned scraping rather than community-led stewardship, communities become data sources rather than epistemic authorities.
So what? Addressing representational gaps in AI by scraping more data from underrepresented communities does not remedy the underlying power imbalance if those communities have no say in how their knowledge is used. Testers evaluating multilingual or culturally diverse AI systems should treat this as a structural quality concern.
Example: Attempts to address the underrepresentation of Global Majority languages in training data often proceed through ungoverned scraping rather than community-led stewardship, communities become data sources rather than epistemic authorities.
