There's More to Come for AI in Ed

• By Dian Schaffhauser
• 12/03/20

The biggest uses for artificial intelligence in education haven't been invented yet. But whatever they end up being, people working on AI applications need to keep educators and education policy makers well informed "early and deeply." That's the conclusion of a new report recently issued by the Center for Integrative Research in Computing and Learning Sciences.

The report was developed out of a gathering of 22 researchers and other experts in AI and learning, who met online over two days to consider two questions:

• What will education leaders need to know about AI in support of student learning, to provide input, do planning and make better decisions?
• What do researchers need to work on "beyond the ordinary" to generate the know-how for "shaping AI in learning for the good"?

Among the participants were representatives from higher education (North Carolina State and the University of Pittsburgh, among them), education organizations (such as Digital Promise and the Education Development Center) and the government (the U.S. Department of Education and the National Science Foundation).

The group came up with dozens of "opportunities" for AI in education, from extending what teachers can do to better understanding human learning:

• Using virtual instructors to free up "personalization time" for classroom teachers;
• Offloading the "cognitive load" of teaching;
• Providing "job aids" for teachers;
• Identifying the links between courses, credentials, degrees and skills;
• "Revolutionizing" testing and assessment;
• Creating new kinds of "systems of support";
• Helping with development of "teaching expertise"; and
• Better understanding human learning through "modeling and building interfaces" in AI.

But contributors also offered just as many barriers to success:

• Differences in the way teachers teach would require "different job aids";
• Teachers would fear losing their jobs;
• Data privacy concerns;
• Bias worries;
• Dealing with unrealistic expectations and fears about AI pushed in "popular culture";
• Lack of diversity in gender, ethnicity and culture in AI projects; and
• Smart use of data would require more teacher training.

Two areas that offered heightened promise to the experts were the uses of AI in "social learning" and in assessment.

For the first, the report noted, AI could be used to help orchestrate complex learning scenarios involving multiple people working on a project together. For example, for students working as team members, AI could notice, listen to and build on various members' contributions while also providing task support, such as helping the team "organize, manage and connect their contributions" to the bigger goal. In this scenario, the report stated, AI would be "socially aware" and could "use social interaction with students as a way of bootstrapping their academic performance."

On the assessment side, AI agents could help teachers "build a portrait of a student's competencies" in various ways, such as by assessing writing and providing suggestions for improvement.

The panel also developed a list of seven recommendations for research priorities, including investigating AI designs for an expanded range of learning scenarios and intensifying and expanding research on AI for assessment of learning.

The full report is openly available on the CIRCLS website, along with a link to an on-demand webinar discussing the findings.