It's a tall task for government agencies and policymakers to determine fair and just resource allocation for people who experience homelessness. But Sanmay Das, professor of computer science in the College of Engineering and Computing's School of Computing, has received two grants from the National Science Foundation (NSF) totaling nearly $760,000 to use computing to develop a new method of optimizing these resource allocation decisions. 

In collaboration with Patrick Fowler, a colleague at Washington University in St. Louis, Das is analyzing the effective allocation of scarce resources like access to shelters, food assistance, and more, to help households experiencing homelessness. “When you have scarce resources like those used to help people who are homeless, you can’t rely on the free market to achieve societal goals, so these are often effectively governed by society,” says Das. 

Using sociological data, Das and Fowler use artificial intelligence to create simulations to estimate the outcomes of allocating specific resources one way or another. “Think of these resources as being in a basket. How does one go about making these allocation decisions? For example, if someone has one resource right now, should they allocate it then, or hold onto it and wait for someone who may be in even greater need?” asks Das. And those are the types of questions Das, Fowler, their colleagues, and students are answering. 

Both of their NSF grants focus on developing these novel decision-making methods for allocating resources in an efficient and fair way, and one looks explicitly at providing targeted assistance to prevent child maltreatment.

Das and Fowler use data from a major urban area to measure the efficacy of different programs by monitoring people’s re-entry into homelessness programs. For example, suppose someone receives an intervention and doesn’t need to use the resources again in the next two years. In that case, the intervention is considered a success according to some standard evaluation metrics, says Das. 

Monitoring effectiveness is central to the long-term success of their work. “There has been limited evaluation of the effectiveness of current practices in preventing and combatting homelessness, which means we don’t know how well these systems are working or how much room for improvement there actually is, and we want to find out,” Das says. 

At the end of their analysis, their work will be used to advise policymakers on how to modify their systems to maximize the impact of available resources. “Fundamentally, it is a challenging problem, but there are gains to be had. So, it’s imperative to study these problems with all of the tools we can bring to bear on them,” says Das. 

https://cec.gmu.edu/news/2021-07/using-artificial-intelligence-combat-homelessness