2749. When "Optimization" Ruins Everything: School Assignment in Antwerp, Belgiumy
Invited abstract in session MC-33: Decision Processes, stream Decision Analysis.
Monday, 12:30-14:00Room: Maurice Keyworth 1.31
Authors (first author is the speaker)
| 1. | Kenneth Sörensen
|
| Faculty of Applied Economics, University of Antwerp |
Abstract
In 2010, the city of Antwerp, Belgium introduced a centralized school assignment system, initially for primary education (kindergarten, ages 2.5+). From then on, parents were asked to communicate an ordered list of preferred schools, after which an algorithm would assign the free seats. This system used a variant of the school-proposing deferred acceptance algorithm with priority given to students residing closer to schools.
In 2017, the city extended this system to high schools but removed the distance priority criterion. The system was strategy-proof, but not Pareto efficient, which meant that children could get a seat at each other's higher-ranked school. Following parental protests, Antwerp then introduced an "optimization" algorithm designed to swap students in this situation.
In this talk, we discuss why such optimization measures are actually detrimental. While the modifications aim to achieve Pareto optimality (where no student could be made better off without making another worse off) they undermine the system's strategy-proofness, a crucial property ensuring that families cannot game the system by misrepresenting their preferences.
Using a computational experiment we demonstrate that such misrepresented preferences undermine the performance of the system, and can potentially lead to disaster. We also discuss several alternative algorithms that achieve the same efficiency without compromising fairness and strategy-proofness.
Keywords
- Algorithms
Status: accepted
Back to the list of papers