Adaptively Weighted Audits of Instant-Runoff Voting Elections: AWAIRE

Abstract An election audit is risk-limiting if the limits (to a pre-specified threshold) chance that an erroneous electoral outcome will be certified. Extant methods for auditing instant-runoff voting (IRV) elections are either not or require cast vote records (CVRs), system’s electronic record of votes on each ballot. CVRs always available, instance, in jurisdictions tabulate IRV contests manually. We develop RLA method (AWAIRE) uses adaptively weighted averages test supermartingales to efficiently when available. The adaptive weighting ‘learns’ efficient set hypotheses confirm outcome. When accurate AWAIRE can use them increase efficiency match performance existing CVRs. provide open-source prototype implementation handle with up six candidates. Simulations using data from real show likely practice. discuss how extend computational approach more Adaptively general tool, useful beyond audits collections sequentially while rigorously controlling familywise error rate.