This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: A Grade 4 teacher showed her class a video of children who walk miles barefoot to school. A hand shot up: "Why are you showing us AI slop?
No way this is real." A classmate pushed back: "Bro, my grandma didn't have shoes when she walked to school." But another chimed in: "Totally AI." I watched this classroom exchange last spring during one of more than 50 classroom observations across five Ontario schools for a study on restorative justice and anti-racism. Teachers whose classrooms I visited expressed an interest in learning how to implement talking or "peacemaking" circles—exercises in which participants gather to discuss topics, share stories or resolve conflicts—and everyone except the person holding what's known as the "talking piece" listens. In the Grade 4 class where students thought they were seeing AI slop, learning about such circles hadn't yet begun.
Debates about generative AI tools in classrooms are often concerned with cheating and accuracy. My research in Canadian classrooms, as well as my own teaching practice and other emerging research about AI-supported debate, points to something else: Listening is a practice. Most teachers need help leading this, and AI might be able to give them that preparation.
In the Grade 4 classroom, most students insisted the video was fake. "Fake" is a powerful word in a classroom because it ends the conversation. If the children on screen might not be real, no one has to contemplate what they're seeing.
Once anything can be synthetic, anything can be dismissed. Listening means treating another person's account of their life as real enough to be changed by it. In our democracies, practicing listening matters more now because we are getting less exposure to perspectives that differ from our own.
Recommendation algorithms sort us into feeds of the like-minded, misinformation exploits the sorting and we finish the job by settling into networks or neighborhoods where disagreement rarely intrudes. Large language models tend to affirm, accommodate and keep users engaged. In a competitive market, the surest way for a frontier model to win loyalty is to tell people what they want to hear.
Extract — continue reading at the source.
