Abraham Gale,Amélie Marian,David M. Pennock
Abstract:Real-life applications of deferred-acceptance (DA) matching algorithms sometimes exhibit errors or changes to the matching inputs that are discovered only after the algorithm has been run and the results are announced to participants. Mitigating the effects of these errors is a different problem than the original match since the decision makers are often constrained by the offers they already sent out. We propose models for this new problem, along with mitigation strategies to go with these models. We explore three different error scenarios: resource reduction, additive errors, and subtractive errors. For each error type, we compute the expected number of students directly harmed, or helped, by the error, the number indirectly harmed or helped, and the number of students with justified envy due to the errors. Error mitigation strategies need to be selected based on the goals of the administrator, which include restoring stability, avoiding direct harm to any participant, and focusing the extra burden on the schools that made the error. We provide empirical simulations of the errors and the mitigation strategies.
What problem does this paper attempt to address?
### What problems does this paper attempt to solve?
This paper aims to solve the error - correction problem of the **Deferred Acceptance (DA) matching algorithm after errors occur in practical applications**. Specifically, after the DA matching algorithm has been run and the results have been announced, errors or changes in the input data may be discovered. These errors may include resource reduction, additive errors, and subtractive errors, etc., resulting in some students being wrongly assigned or not assigned to schools. Since these errors are discovered only after the matching results are announced, re - running the matching algorithm is neither practical nor fair.
Therefore, the paper proposes several models and strategies to mitigate the impact of these errors, ensuring that the negative impact on students is minimized without re - running the entire matching process, and restoring the stability and credibility of the system.
#### Specific problem descriptions:
1. **Resource Reduction**: For example, a school closes after matching, leaving students originally assigned to this school without a school to go to.
2. **Subtractive Errors**: For example, a school loses some application materials, causing some students who should have been admitted not to be considered.
3. **Additive Errors**: For example, some students' score calculations are wrong, causing them to be wrongly ranked higher, thus affecting the matching results of other students.
#### Error - correction goals:
- **Avoid direct harm**: Do not cause additional harm to students who are not directly affected by the errors.
- **Retain issued admission notices**: Students who have already received admission notices should not lose their opportunities due to error - correction.
- **Minimize resource increase**: If new resources (such as new degrees) need to be added, the number of new additions should be minimized and fairly distributed.
- **Focus on the party responsible for the error**: If the error is caused by a specific party, the error - correction measures should make this party bear more responsibility as much as possible.
Through these models and strategies, the paper hopes to effectively deal with and correct the errors that occur during the matching process while maintaining the stability and fairness of the system.