Helping Those Most In Need

Often the most vulnerable people, those who need the most help, are the ones with the least resources, and the least access to be able to gain those resources.    

A team at The University of Buffalo has received an $800,000 grant to help develop a machine learning tool with a goal of helping caseworkers and human services agencies determine the best available services for youth that age out of foster care and never reunite with their families. The number of youth that meets these criteria is over 20,000 a year and can be left without the resources to succeed after they age out of the system. 

This grant is being jointly funded by Amazon and The National Science Foundation and they are partnering on a program that is being called “Fairness in Artificial Intelligence.” The goal of this program is to address bias in and “build trustworthy computational systems that can contribute to solving the biggest challenges in modern society.”

Here’s how it will work: Over a period of three years, a multi-disciplinary team of researchers from the University of Buffalo will collaborate with the Hillside Family of Agencies in Rochester. Hillside is one of the oldest family and youth non-for-profit human services organizations in the country and will work together with the UB researchers, as well a youth advisory council which will be made up of individuals who have recently aged out of the system, all of these entities will work together to create this tool. 

These groups will also consult with experts to ensure that the work is being done properly.

UB researches in multiple disciplines will use data collected from the Administration on Children and Family Services’ National Youth In Transition Database. This information combined with input from collaborators will; help to inform the predictive model. All states participate in this database to report experiences as well as available and used services by youth in foster care. 

The goals of this program are three-fold:

1. To use the combined experiences of youth, caseworkers, and experts in the foster care system to identify the hidden biases in data that train machine learning and AI models
2. To obtain multiple perspectives on objective fairness in regards to decisions about services
3. To build a system that can more equitably and efficiently deliver those services to youths that age out of the system

This is meant to be a tool that will help those working within the system to make better, more informed decisions. 

Here are some of the statistics that make such a tool so necessary: 

• By age 19, 47% of youth who age out of the foster care system have not finished high school. 
• 20% have experienced homelessness 
• 27% of males have been incarcerated