AI-in-the-loop human interventions for homelessness resource allocation

Rapid improvements in real-world AI systems have created opportunities to improve the efficiency and effectiveness of human resource allocation decisions. The use of AI in public resource allocation is promising in its ability to reduce administrative burden while optimally solving complex resource allocation problems. This work examines the problem of homelessness affecting the City of Austin and proposes an AI-in-the-loop decision making framework to augment and improve housing allocation decisions. Our research develops a holistic understanding of the homelessness landscape through descriptive analysis of the Homeless Management Information System and then proceeds towards developing a demographic parity group fairness criterion based intervention to improve the equity in outcomes associated with the Austin Prioritization Index Coordinated Assessment. Our findings result in the proposal of an AI-in-the-loop assistive decision system to augment and improve the Permanent Supportive Housing and Rapid Rehousing resource allocation system. Our proposed research is designed to enable decision makers and homelessness service providers to move towards a deliberative case management framework.