How e dey work, well well
Di voter give grade to EACH option, separate from di others: dem dey evaluate, no be rank. Two options fit collect di same grade.
For each option, dem arrange di grades and take di median — di grade wey majority judge « at least dis good » and majority judge « at most dis good ». Unlike di average, di median no dey shake when person drop extreme grade for tactics.
If medians tie, dem go break am by removing median grades one by one from di tied options until dem separate. Na di same as comparing how many supporters and opponents dey around di median grade.
Example wey get figures
Three options wey eleven people grade on five-grade scale.
| Option | Grades wey e collect | Median |
|---|---|---|
| Bordeaux | 3 Very good, 5 Good, 3 Fair | Good |
| Lyon | 5 Very good, 1 Good, 5 Reject | Good |
| Lille | 2 Very good, 4 Fair, 5 Poor | Fair |
- Bordeaux and Lyon get di same median grade: Good.
- Remove one median grade from each and start again: Bordeaux still get wider base above Good.
- Lyon, wey dey divide people (5 « Very good » but 5 « Reject »), fall behind.
Majority judgment prefer di option wey plenty people respect pass di one wey dey polarise. One-round vote for don crown Lyon on im five strong fans.
Where e come from
Majority judgment na Michel Balinski and Rida Laraki propose am for 2007, researchers for École polytechnique and CNRS, for di Proceedings of the National Academy of Sciences. Dia main book, « Majority Judgment: Measuring, Ranking, and Electing », come out for MIT Press for 2011.
Di inspiration no come from politics: wine competitions, gymnastics and figure skating don dey grade with word scales for long and dey throw away di extremes, because average too easy for one judge to manipulate.
Di authors test am for April 2007 with voters for Orsay during di French presidential election, side by side with di official vote. Since then several French citizen primaries don use am, including di Primaire populaire for 2022.
Where dem dey use am
- Internal elections and citizen primaries, where you wan measure real support.
- To assess applications, projects or suppliers on quality criteria.
- Decisions where you wan catch di divisive option before you adopt am.
- Any vote where di strength of support, no be just im existence, suppose count.
Limits and wahala
- E show plenty feeling, no be just one choice
- E reduce tactical voting
- E measure real support
- Di median no too clear
- Longer ballot
- Plenty tie fit happen
Questions wey people dey ask
Why median instead of average?
Because people fit buy average. Tactical voter wey give di lowest grade to serious rival go move dia average well; e go move di median only one step, and only if plenty people do di same. Di median na di heart of di method, no be calculation detail.
E dey truly impossible to manipulate?
No — di Gibbard-Satterthwaite theorem no allow dat for any method. But majority judgment dey cut down wetin person fit gain from insincere vote, and na di best wey person fit prove.
How many grades I need?
Five or six, with clear word labels. If e less, you lose di detail; if e plenty, voters no fit separate di steps wey dey near each other.