Placet
🧮

Maximum satisfaction

Di sharing wey go please di most people, calculated exact.

Instead of serving people one after di other, dis one dey find di assignment wey make di sum of di ranks small pass. Di best collective result, calculated one time.

Each sharing "cost" di position wey e hold for di person ranking (1st choice = 0, 2nd = 1…). Di Hungarian algorithm go find di sharing with di smallest total cost: di best overall satisfaction, exact.

Start assignment with dis method →

How e dey work, well well

Each person rank di options. Wish of rank 1 cost 1, rank 2 cost 2, and so on: na so dem build di cost matrix.

Dem dey find di assignment wey make di TOTAL cost small pass, exploring all di possible combinations with sense — never one by one, wey no go possible.

Important thing: di calculation fit sacrifice one person to help several. Di optimum na collective, and na exactly wetin dem ask am to do.

Example wey get figures

Three people, three assignments. Di boxes show di rank of di wish.

AuditRedesignSupport
Chloé123
Ali132
Bruno213
  1. To give everybody dia first wish no possible: Chloé and Ali both want Audit.
  2. Chloé→Audit, Bruno→Redesign, Ali→Support: total cost 1 + 1 + 2 = 4.
  3. Ali→Audit, Bruno→Redesign, Chloé→Support: total cost 1 + 1 + 3 = 5.

Dem take di first combination: for similar wishes, e cost di group less. No other arrangement fit beat 4.

Where e come from

Na di « assignment problem », classic work for operations research. Harold Kuhn publish efficient solution for 1955 and call am « Hungarian algorithm », to honour di Hungarian mathematicians Dénes Kőnig and Jenő Egerváry wey im work follow.

Later dem come find say Carl Gustav Jacobi don solve di problem for di 19th century, for work wey dem publish after im death for 1890 — sixty-five years before dem rediscover am.

Today di algorithm na normal industrial tool: to assign crews to flights, tasks to machines, vehicles to trips. Every application wey dey pair two sets while optimising total cost na im pikin.

Where dem dey use am

  • To share assignments or files across team while maximising overall satisfaction.
  • To assign pupils to workshops, options or projects.
  • To assign on-call or duty slots.
  • To plan crews, routes or machines — di historic industrial use.

Limits and wahala

✓ GOOD SIDE
  • Di best total satisfaction wey possible
  • No draw involved
  • Exact result, no be approximate
✕ BAD SIDE
  • E fit sacrifice one person for common good
  • Tactical rankings dey possible
  • E no too easy to explain
People fit manipulate am
Unlike serial dictatorship, to lie fit pay: if you rank one popular option low, dem fit give you better one. Di method no dey strategy-proof.
Collective optimum, no be individual
Person fit collect dia last wish so dat di total go drop. Mathematically correct, but hard to announce to human being.
Hard to verify
Nobody fit redo di calculation for head. Trust dey rest on di tool, wey dey weaken how people see di legitimacy.
Ranks treated like distance
Di gap between 1st and 2nd wish dey count like di gap between 4th and 5th, even though people no dey feel dem di same way at all.

Questions wey people dey ask

How e take better pass serial dictatorship?

On average, di group dey more satisfied: di algorithm dey see all di combinations one time, while serial dictatorship dey suffer di running order. Di price na readability — and di chance for people to lie.

Wetin happen if two assignments tie?

Several solutions fit reach di same minimum cost; dem go pick one. If di matter heavy, announce di tie-break rule before, or switch to serial dictatorship, wey mechanics dey reproducible.

I need as many places as people?

No, but di gap get price: if places no reach, somebody no go get assignment; if dem plenty pass, some go remain empty. Di calculation still valid both ways.

Check dis too

Picking roundsTwo groups (Parcoursup)Trading circle

Voting methods