Goals:
- Avoid over-weighting participants whose preferences the model predicts with higher confidence
- Reduce susceptibility to Strategic manipulation and Winner’s curse effects.
Method:
- Ranked-choice aggregation: capturing the ordering of preferences rather than just top choice (reduces gaming)
- Shulze method: identifies statement that would win all pairwise contests against others (if possible), producing a Condorcet winner (avoids winneres that most people dislike)
Why this works:
| Voters | 1st Choice | 2nd | 3rd |
|---|---|---|---|
| 35% | A | B | C |
| 33% | B | C | A |
| 32% | C | B | A |
In this case, majority vote winner would be A, even though 65% prefer B or C over A. In other words, majority vote fails to account for second or third preferences, while Shulze method does.