The CIO 100 Awards acknowledges the IT accomplishments of businesses and government organizations every year, and this year, its honorees include the Department of Technology (DT)! DT has been recognized for its exemplary work on the Risk Limiting Audit (RLA) Software Platform in partnership with the San Francisco Department of Elections.
Verification of election results is an essential local government duty to ensure an election’s outcome is fair and reflects the will of the voters. The verification and audit process can identify voting irregularities or prevent fraud that would otherwise undermine the integrity of the election. In partnership with the San Francisco Department of Elections, the Department of Technology developed a software system to automate the verification of election results. Using a combination of open-source applications and algorithms from industry experts in government transparency and election fairness, the Risk Limiting Audit Software Platform can automatically audit thousands of votes on-demand to verify results.
So the election is done! All the votes have been counted. How do you double-check? Do you recount every ballot, again? While recounting all ballots is an expensive option available to all jurisdictions, there are statistical ways to check the exact number of ballots needed to confirm ballots were counted properly.
Normally, audits are performed manually by randomly selecting 1% of ballots from all voting areas in a given jurisdiction. For San Francisco, this would mean collecting 880 random ballots for a district election or 8,800 for a Citywide election. This can be work-intensive and prone to error or bias. Because of that, the results of this manual audit are not statistically sound.
An automated risk limiting audit, however, generates random results that are not bias-prone and checks as many ballots are needed until the results have been verified. If the results are too close to call or verify, eventually, all ballots will be recounted! However, a risk limiting audit will often provide a range of results. If the margin of victory is wide, fewer ballots are chosen. If the margin is narrow, more ballots are reviewed.
Although researchers and subject matter experts had developed formulae for performing risk limiting audits, these are complicated to implement manually or only existed in forbidding, user-unfriendly software. This is where DT comes in! Working with the Department of Elections and industry experts, DT implemented the RLA formulae in software and developed a user-friendly application for end users.
The software platform lets users independently verify the results of different elections; it stores records of who did what and when; and it automatically lists which ballots to review based on the outcome of the election.
This new system provides greater security and transparency, all while lowering the cost of government! For example, San Francisco’s 2022 District Attorney election had over 125,000 ballots cast! A manual 1% recount would have needed at least 1,250 ballots. However, with RLA, only 150 ballots were needed to confirm the final result. Close to a 90% reduction in ballots!
On the other hand, the RLA required more ballots than the 1% manual recount when auditing very close races. The 2022 District 4 Supervisor race was extremely close! A traditional 1% manual recount would have only required 140 ballots to review despite narrow margins. The RLA system automatically recommended 735 ballots to be reviewed — 5 times as much as the manual process!
In this way, the RLA Software Platform makes it easy for election official to audit and verify results while ensuring they spend more time verifying close elections.
The Department of Technology is grateful to receive this recognition and is happy for its successful collaboration with industry experts and the San Francisco Department of Elections. DT is committed to innovation and work that helps the City and County of San Francisco protect and serve its many, diverse residents.