Researchers at SRI International developed an AI-powered YouTube filter that helps parents and teachers find educational videos for kindergarten and prekindergarten children. This tool, called Assisting Parents to Review Online Videos for Education (APPROVE), uses audio and visual content to find YouTube videos that teach early math and literacy.
Digital media: One of today’s biggest parenting challenges
SRI researchers are solving one of parents’ most pressing problems: how to support young children’s healthy media use. Young children in the U.S. spend more time watching videos online than any other way. Parents are struggling. Here’s what they told us:
“I want to be able to put YouTube on and feel comfortable leaving the room.”
“It’s too easy to go down a rabbit hole into inappropriate content. The recommendations aren’t good.”
“There is a lot of educational content out there, so I want them to be exposed to that.”
Clearly, parents want their children to enjoy watching online videos, but they worry about what their child might be exposed to. Parents know there are educational videos on YouTube, but it can be hard to know which ones are best or to encourage kids to watch them.
Groundbreaking solutions from artificial intelligence and education
SRI education and computer science researchers have developed an AI-powered tool to help parents filter online videos for kids. We call this tool APPROVE: Assisting Parents to Review Online Videos for Education. Using machine learning algorithms developed by SRI computer scientists, APPROVE automatically filters YouTube videos to find only the ones with age-appropriate literacy and math content.
“This is groundbreaking,” said lead researcher Claire Christensen. “Until now, if we wanted to know what kids were seeing online, humans had to watch it.” It’s almost impossible for humans to review every video on YouTube; 500 hours of YouTube content are uploaded every minute. But APPROVE can review an unlimited number of videos automatically.
APPROVE can identify videos aligned to the Common Core State Standards for pre-kindergarten and kindergarten, and aligned to the Head Start Early Learning Outcomes Framework for prekindergarten. We received SRI research and development funding to train APPROVE to detect literacy content. And we received NSF funding to train APPROVE to detect math content.
The future of video streaming
APPROVE is not yet available for home use. SRI is looking for potential partners interested in integrating APPROVE with existing video streaming sites. We envision a future, publicly available version of APPROVE that would allow parents to select the educational topics they want their child to watch online. APPROVE would then filter the child’s online video search results to include the keywords they entered as well as the educational content their parent selected. For example, a child might search for videos about horses, and with APPROVE’s AI-powered filters, all of the search results would be math or reading videos that also include horses.
Parents are excited to use APPROVE!
“I love that it filters through the [garbage] … it makes me feel a lot less guilty to leave them unattended on YouTube if they are learning something. I really like the idea that they can still watch something they are interested in, they’re just going to get some educational sprinkles thrown in there too.”
An interdisciplinary collaboration
The research team is led by Dr. Claire Christensen in SRI’s Education Division and Dr. Anirban Roy in SRI’s Information and Computer Science division. Claire is an expert in what makes media for young children educational. She has evaluated educational videos and games for many media developers, including the PBS KIDS show Molly of Denali as part of SRI and EDC’s evaluation of PBS’s Ready to Learn Initiative. Anirban is an advanced interdisciplinary researcher at the cross-section of computer vision, machine learning, and deep learning.