Optimization-based Approaches to Housing Allocation
By Aida Rahmattalabi
Sunday morning started with a series of great talks. The session “Optimization-based Approaches to Housing Allocation,” hosted four talks, each addressing the problem of housing allocation from a different angle. The first presentation was given by Sanmay Das, an associate professor in computer science and engineering at the University of Washington in St. Louis. The talk was centered on the issue of homelessness and how one can use analytic tools in order to improve the allocation of homelessness services. He posed an interesting problem on using historical data to predict intervention outcomes which can be further used in the design of better interventions.
The second talk approached the problem of housing allocation for homeless youth from the policy makers’ perspective. Duncan C. McElfresh, PhD candidate in applied mathematics at the University of Maryland, discussed how policy makers often have different preferences and what challenges one faces to gain the trust of policy makers and social workers. Their proposed solution is to elicit their preferences and to take them into account in the design of new policies. In dealing with large policy spaces, they develop a novel and robust methodology that uses a small set of queries to efficiently learn the policy makers’ preferences.
Next, Nick Arnosti from Columbia Business School gave a talk on allocation of public housing where online housing resources must be matched to people on a waiting list. He described interesting analytical results that show existing policies lead to suboptimal allocations. He then discussed strategies to mitigate this problem. Specifically, he proposed a common lottery for all housing resources and he argued that by giving individuals more choices we can ultimately improve the quality of the matches.
The last talk was given by Aida Rahmattalabi, PhD student in computer science at the University of Southern California. She revisited the issues surrounding homelessness, but this time from a mental health perspective. She discussed how this vulnerable population often suffer from high rates of suicide, and she proposed an optimization model of the suicide prevention programs that strategically targets the individuals for intervention. In her talk, she highlighted the importance of fairness, and that how existing algorithms can inadvertently result in discriminatory intervention outcomes. She concluded her talk with empirical results of their algorithm on real-world social networks of youth experiencing homelessness, where she demonstrated that their algorithm is able to provide fairer interventions.