Harvard group launches free guide to AI use by age and learning goal

Parents and educators now have a free way to explore where AI might support learning—and where it might take over work learners need to do themselves. The Compass Collaborative at Harvard Graduate School of Education launched AI Guidance by Age this week, tailoring recommendations to a person’s role, developmental goals and learners’ ages.

Start with the skill, not the app

The tool asks users to make three selections: their role, the area of development they want to explore, and an age range. Its nine roles include school and higher education educators, workforce educators, students, researchers and policymakers. Developmental areas include language and literacy, cognitive foundations, and social understanding and interaction.

The age options span 0 to 40, so the resource extends beyond childhood. Results explain relevant developmental processes and milestones, recommend uses of AI, and describe how the technology may help or hinder particular goals. Educators can also get assessment-design recommendations and additional resources.

This is not a directory of approved products. It offers no ratings or reviews of specific AI apps. Instead, the framework asks whether a learner still gets to practice the skill being developed, and whether AI supports interactions with other people or replaces them. Parents and teachers must make the final decision about how to apply that guidance.

The first draft stays with the student

One example shows how narrow the recommendations can be. For a higher education instructor teaching language and literacy to 18- to 25-year-olds, the matrix may suggest using AI to broaden reading, reflect a student’s writing back to them, or prepare them for a difficult text. Composing and working through the material remain the student’s job.

Compass co-director Jeff Curley told The Boston Globe that high school students should do the thinking first, whether that means drafting or brainstorming on their own. Afterward, AI can serve as an editor or a partner that gives critical feedback. His distinction leaves room for AI assistance without treating a completed assignment as evidence that the student did the thinking.

Co-director Tessa Forshaw described a different possibility for creativity: a tool might help a child bring a story idea to life through complex descriptions and a fuller narrative. In that example, she said, the technology can provide practice and preserve the child’s agency over the storytelling rather than simply take over.

There are things we can’t and won’t give direct guidance [for] if the research isn’t there to support it.

Jeff Curley, Compass co-director, speaking to The Boston Globe

A large research base, but not 5,000 AI studies

Researchers developed the matrix by reviewing more than 5,000 studies in education, psychology and human development. That figure needs an important qualification: much of the work predates generative AI, sometimes by decades. Independent research on how AI-powered learning tools affect learning outcomes remains limited.

Forshaw told Inside Higher Ed that the team combined established learning research with emerging findings to guide decisions while AI-specific evidence develops. The older studies identify consistent patterns in how people learn; the matrix uses those patterns to assess likely benefits and drawbacks in particular settings. Its recommendations are therefore research-informed guidance, not a finding that every suggested AI use improves learning.

Curley identified children’s use of AI for companionship as one area with research gaps. The team’s stated approach is to flag those gaps rather than supply a satisfying answer without support. The resource is available at compass.gse.harvard.edu/explore.