Chicago tests AI‑led model

Chicago is about to test one of the starkest ideas in education: what happens when software handles the teaching and adults in the room are not teachers. A new Alpha School campus is set to open in the Loop this fall for about 100 students, using artificial intelligence-powered lessons for core academics and “guides” to supervise students through the day. (governing.com) The pitch is simple and expensive. Alpha says students can finish academic lessons in two hours a day, spend the rest of the schedule in workshops like robotics and obstacle-course activities, and do it all for $55,000 a year in tuition. (governing.com) The model comes from a private-school network founded in Austin, Texas, in 2014 by education entrepreneur MacKenzie Price. Governing reports that the network has expanded to nearly two dozen locations across the United States and now enrolls more than 1,000 students. (governing.com) What makes the Chicago experiment different is not just the use of software. Schools have used digital math and reading programs for years, but Alpha pushes the idea further by removing credentialed classroom teachers from the center of daily instruction and replacing them with noncredentialed “guides.” (governing.com) That shift lands in a city whose public school system is taking a much more cautious line on artificial intelligence. Chicago Public Schools says its own approach is “responsible adoption,” with guidance focused on ethics, privacy, security, bias, and professional development for faculty and staff rather than teacher replacement. (cps.edu 1) (cps.edu 2) Chicago Public Schools’ guidance is explicit about the risks. The district warns staff and users never to enter personally identifiable information, protected health information, or confidential student records into generative artificial intelligence tools, and it says outputs must always be checked because they can contain false information and bias. (cps.edu) That is the backdrop for the harder question Alpha raises: not whether software can help with schoolwork, but whether it can carry the core work of teaching young children. For older students working independently, adaptive software can often drill skills efficiently; for younger children, the missing piece is usually not content delivery but live interaction. (governing.com) (naeyc.org) Early learning depends heavily on oral language, which means children build understanding by talking, listening, asking questions, and getting responses in real time. The Institute of Education Sciences says oral language is the foundation for later reading and also supports math, science, social development, and self-regulation. (ies.ed.gov) That matters because a good elementary classroom is not just a place where facts are delivered. It is a place where a teacher notices hesitation in a child’s face, rephrases a direction, settles frustration before it spills over, and turns a half-formed answer into a conversation the whole room can learn from. (naeyc.org 1) (naeyc.org 2) Researchers and child-development groups have been making a related point for years about screens. The National Association for the Education of Young Children says young children can learn from interactive media, but they learn more easily from real-life experiences, and the value of media depends on content, context, and the individual child. (naeyc.org) The American Academy of Pediatrics has also moved away from simple screen-time counting toward a quality question: what exactly is the child doing on the screen, and who is helping them make sense of it. That sounds close to Alpha’s promise of personalized software, but it also underlines the weakness of any system that treats adult interaction as optional. (aap.org) There is already evidence that regulators are uneasy with the teacher-light version of this model. In January 2025, Pennsylvania’s Department of Education rejected an application for Unbound Academy, a cyber charter school that would have used artificial intelligence tutors and human “guides,” saying the instructional model was untested and did not show how it would align with state standards. (governing.com) Alpha’s supporters argue that the old school model wastes time and that software can personalize instruction better than one teacher facing 25 students at once. Critics answer that efficiency is not the same thing as education, especially in kindergarten and elementary school, where children are still learning how to speak in groups, regulate feelings, follow social cues, and connect words to the world around them. (governing.com) (ies.ed.gov) So the Chicago test is less a clean replacement story than a stress test for the limits of software-first schooling. If the model works, it will strengthen the case that some academic instruction can be compressed and automated; if it struggles, the reason may be the oldest one in education, which is that young children do not just need lessons, they need adults who can teach them in the moment. (governing.com) (naeyc.org)