Representation with Incomplete Votes

Platforms for online civic participation rely heavily on methods condensing thousands of comments into a relevant handful, based whether participants agree or disagree with them. These should guarantee fair representation the participants, as their outcomes may affect health conversation and inform impactful downstream decisions. To that end, we draw literature approval-based committee elections. Our setting is novel in approval votes are incomplete since will typically not vote all comments. We prove this complication renders non-adaptive algorithms impractical terms amount information they must gather. Therefore, develop an adaptive algorithm uses more efficiently by presenting incoming statements appear promising previous participants. method satisfies commonly used notions representation, even when only small fraction Finally, empirical evaluation using real data shows proposed provides representative practice.