Investigating Fairness in Epidemic Control

Sam Vanspringel, Alexandra Cimpean

Track: Type D (Master/Bachelor Thesis Abstracts)

Keywords: COVID-19, Fair epidemic mitigation, Multi-objective Reinforcement Learning

Abstract: Effective pandemic mitigation strategies should reduce both disease burden (e.g., hospitalizations) and societal impact. Previous work approached this challenge with a multi-objective reinforcement learning method that balances hospitalizations against the loss of social contacts (i.e., social burden). To build on this, we recognize that fairness should also be considered when designing mitigation strategies. To this end, we introduce social burden fairness, which aims for a more equitable distribution of social burden, weighted by age-specific hospitalization risk. We experimentally evaluate our fairness-based approach against a baseline that jointly optimizes hospitalizations and overall social burden, but does not treat fairness as an explicit optimization objective. We then investigate how integrating age-oriented fairness constraints influences the learned mitigation strategies.

Serve As Reviewer: ~Pieter_Jules_Karel_Libin2

Submission Number: 64