Research & Papers

Sinay & Gonen prove PAV voting rule vulnerable to manipulation with known ballots

New RAT-degree framework reveals exact number of voter reports needed to safely manipulate committee selection.

Deep Dive

Sinay and Gonen’s latest work tackles a fundamental gap in voting theory: how much information does a manipulative voter need to safely rig a committee election? Their paper, submitted to arXiv on July 27, 2026, extends the concept of risk-avoiding truthfulness (RAT) from single-winner settings to approval-based committee (ABC) selection. The authors focus on the widely used Proportional Approval Voting (PAV) rule, which balances proportional representation with strategic robustness.

They prove two tight bounds: PAV is vulnerable to safe manipulation when the manipulating agent knows the exact ballots of ⌈n/k⌉ other voters, but remains completely immune if they know at most ⌊n/(k+1)⌋−1 ballots. These results quantify the exact threshold of insider knowledge needed to break the system, offering a precise tool for evaluating real-world voting mechanisms used in corporate boards, political committees, and AI alignment panels.

Key Points
  • PAV rule is vulnerable to safe manipulation with knowledge of ⌈n/k⌉ voter ballots.
  • PAV remains strategy-proof when knowledge is limited to ⌊n/(k+1)⌋−1 ballots or fewer.
  • First extension of the RAT-degree framework to multi-winner approval-based committee elections.

Why It Matters

Quantifies real-world voting security thresholds, helping design committees that resist manipulation with limited insider knowledge.

📬 Get the top 10 AI stories daily