But some might argue that optimizing network connectivity is a more nebulous task than optimizing test scores. What exactly should be the objective functions?
One framework to explore this could be to focus on how the networks in which children and families intertwine are formed and developed in the first place. In the context of school education, this includes the wide range of policies developed by school districts to determine which schools students can attend (“assignment policies”), as well as the practices that families adhere to when choosing a school for their children in accordance with this politics. These policies and practices have historically perpetuated harmful traits such as school segregation based on racial and socioeconomic status, which, despite almost 70 years since it was officially outlawed, continues to define public education in the USA. Many scholars argue that demographic integration has historically been one of the the most effective methods not only to improve the academic preparation of historically disadvantaged groups, but also to fostering more compassion and understanding – let’s say ethics pluralismamong people of different backgrounds.
AI can help develop more equitable school assignment policies that promote school diversity and inclusion, for example by supporting district-level planning efforts to redraw “school attendance zones”—i.e., coverage areas that determine which neighborhoods feed which schools — in a way that mitigate underlying patterns of residential segregation without imposing high transportation costs and other inconveniences on families.
Existing partnership between researchers and practitioners– and some of my own research with co-authors Doug Biferman, Kristin Vega-Purheidarian, Cassandra Auvergne, Pascal Van Hentenreek, Kumar Chandra, and Deb Roy – use rules-based operations research and AI community tools such as constraint programming explore alternative designation policies that could optimize racial and socioeconomic inclusion in schools.
These algorithms can help simplify the otherwise cumbersome process of exploring a seemingly endless number of possible boundary changes to identify potential paths to more integrated schools that balance a number of competing goals (such as family travel time and school change). They can also be combined with machine learning systems — such as those that try to predict family choices in the face of changing boundaries — to more realistically assess how policy changes might affect school demographics.
Of course, none of these AI applications are without risks. Changing schools can be disruptive for students, and even with school-level integration, segregation can persist at smaller scales, such as classrooms and canteens, due to curriculum tracking, lack of culturally sensitive teaching practice and other factors. In addition, applications must be built into an appropriate socio-technical infrastructure that takes community input into the policy development process. However, using AI to inform which students and families are attending school together could bring about deeper structural changes that will change the networks students connect to and, by extension, the life outcomes they ultimately achieve.
However, changes in school designation policy without changes in school choice behavior by families are unlikely to lead to sustainable transformations in the networks to which students are connected. AI can play a role here too. For example, digital school ranking platforms such as GreatSchools.org are increasingly influencing how families rate and choose their children’s schools, especially as their ratings are often embedded in housing sites like Redfin, which can influence where families choose to live.