New model predicts which voting rules resist coalitional manipulation
A minimal model with no tuning accurately ranks vulnerability of 12 voting rules.
Deep Dive
François Durand's paper analyzes all standard ordinal voting rules under the Perturbed Culture model, finding sharp phase transitions where manipulation succeeds with high probability below a critical concentration threshold and fails above it. The model reveals natural families: Maximin, Ranked Pairs, Schulze, and Young share identical thresholds, as do Baldwin, Nanson, Kemeny, and Dodgson. Tested on Netflix and FairVote datasets, it accurately predicts vulnerability ranking based solely on number of candidates.
Key Points
- Maximin, Ranked Pairs, Schulze, and Young form a family with identical resilience thresholds.
- Baldwin, Nanson, Kemeny, and Dodgson form another family based on strengthened Condorcet notions.
- Model validated on Netflix and FairVote data, accurately ranking vulnerability by candidate count.
Why It Matters
Offers a simple, data-backed method to choose voting rules resistant to manipulation in elections and AI aggregators.