By Josh Stowe
More than 3.4 million American children, or about 6 percent of the school-age population, now learn at home, and that group has grown at nearly three times the pre-pandemic rate. Homeschooling is also increasingly popular in countries including Canada and the United Kingdom, as more parents look to customize educational options for their children.
A new study co-authored by University of Notre Dame researcher Mohammad Rashidujjaman Rifat shows how artificial intelligence can empower parents to be engaged educators who tailor curriculums and learning plans that meet their children’s needs.
The study was published in the 2026 proceedings of the CHI Conference on Human Factors in Computing Systems, the flagship conference for researchers working in human-computer interaction.
“School-centered platforms optimize for classroom control, standardized pacing and scale, but homeschooling families often prioritize adaptable routines, value-aligned materials and community-based support,” said Rifat, assistant professor of tech ethics and global affairs at Notre Dame’s Keough School of Global Affairs. “We wanted to build something that keeps parents in control as reflective partners who shape what their children actually learn.”
Rifat and his colleagues designed a digital platform powered by a large language model. Their goal was to accommodate a diverse cohort of homeschooling families whose needs range from academic flexibility and cultural alignment to school safety concerns. During the study, they worked with Muslim families in the greater Toronto area, evaluating the platform’s usability and overall effectiveness from April to July 2025. Over time, they hope to work with larger groups in multiple communities.
Researchers designed the platform, called Homeroom, around three goals: supporting parents as active participants who shape their children’s education; promoting culturally responsive personalization to help families create identity-affirming learning environments; and facilitating parents’ collaboration within a trusted circle of fellow homeschoolers.
The tool’s workflow enabled participants to brainstorm, create and refine stories for their children’s ethics education and to craft curriculums that incorporated family heritage and cultural norms alongside age-appropriate educational standards.
Findings underscore need for agency, representation
Researchers found that parents embraced a trust-but-verify approach, wanting to preview and edit educational materials before sharing them with their children. That desire for control was especially evident around cultural representation.
Some parents used the platform to fill gaps in available materials, generating stories that reflected their children’s cultural identities. One participant said she would use the tool primarily because heritage stories are difficult to find in libraries and too costly to keep buying. Parents appreciated being able to adjust tone, setting and cultural motifs including familiar foods, holidays and everyday customs until a story felt right for their child.
They also identified a need for precision around cultural values. When one participant asked the platform to generate stories emphasizing humility, she received moralizing narration rather than what she had in mind, which was a set of concrete, everyday practices her children already knew, such as sitting on the floor to eat. The experience pointed to a broader finding: Generic value labels are not enough. Families needed tools to define what those values actually look like in their household.
Researchers also found that participants preferred small, intimate groups over broad forums, which they found made it easier to ask sensitive questions and receive practical feedback. Parents wanted to follow trusted peers, adapt shared lessons while providing clear attribution and retain the option to monetize original content.
In addition, parents asked for tools to reduce administrative workload, including reusable rubrics, a portfolio generator and connections to local events that could feed directly into student records.
Addressing a larger challenge in human-AI collaboration
The study is a natural extension of Rifat’s broader research agenda, which examines how AI can move beyond one-size-fits-all design to better serve diverse communities with distinct cultural needs and values.
It also addresses the concern that AI risks being sycophantic and simply reinforcing what the user already believes. While that risk is real, Rifat said, well-designed personalization can create space for communities to encounter difference rather than avoid it. Homeroom included features that exposed children to materials from outside their own background precisely for that reason.
“One of the critiques I hear is that we’re building echo chambers,” Rifat said. “But personalization is already happening whether we like it or not. The question is whether we design it in a way that brings communities together or drives them further apart.”
Rifat co-authored the study with Ayla Khan, a student at the University of Toronto Mississauga; Nouran Yahya, a student at the University of Waterloo and a research assistant at the University of Toronto Mississauga during the study; and Samar Sabie, assistant professor at the University of Toronto. The study received funding from the Urban Challenge Grant program at the University of Toronto’s School of Cities.
In addition to his role at the Keough School, Rifat holds a concurrent appointment in the Department of Computer Science and Engineering within Notre Dame’s College of Engineering. He is also a faculty affiliate of the Keough School’s McKenna Center for Human Development and Global Business, Ansari Institute for Global Engagement with Religion and Pulte Institute for Global Development as well as Notre Dame’s Institute for Ethics and the Common Good and Lucy Family Institute for Data & Society.