The Hardest Problem I’ve Ever Seen

Matt Blaze opens by describing how US elections are run: mostly by counties and townships, under state law, with budgets competing against everyday local services. He stresses two uncomfortable facts. Real technical vulnerabilities exist in parts of election infrastructure, yet no credible evidence has ever shown a technical attack changing a US election outcome.

He traces the path from the 2000 Florida recount, where punch-card ballots and hanging chads became a national embarrassment, to the Help America Vote Act. That law pushed jurisdictions toward paper-based optical scan systems and touchscreen machines. Software-driven direct recording machines brought new trust problems, because voters cannot read their own votes and software failures can quietly alter results.

The central argument is that software is unreliable, so election systems should be designed so that a software flaw cannot cause an unrecoverable wrong outcome. Ron Rivest's idea of software independence, and Philip Stark's risk-limiting audits, show how hand-checking a small random sample of paper ballots can bound the risk of a wrong result. Blaze closes by urging technologists to help local officials, consider becoming election workers, and help retire paperless touchscreen machines.