Designing Defaults for School Choice
in
Workshop: Workshop on Human and Machine Decisions
Abstract
School districts employing variations on the Gale–Shapley deferred acceptance algorithm assume that households have perfect information and list their preferences over schools truthfully. However, many families submit partial preference lists either by virtue of limited available resources or a misunderstanding of the mechanism. We investigate the role of defaults in deferred-acceptance towards alleviating search costs for families.
In San Francisco Unified School District (SFUSD), 11% of the 4,713 students were assigned using distance-based defaults in 2018-19. We study nine variations of the SFUSD assignment system, focusing on how defaults are constructed and how defaults are integrated algorithmically. We observe and discuss the change in the estimated welfare for different populations under the nine variations, and seek input on how to improve and evaluate our approach.