As classes resumed last week at Los Angeles public schools, the district and families are dealing with problems in what was billed as the country’s first artificial intelligence program tailored to students’ learning.
I’ll recap the saga in L.A. and suggest questions that parents and citizens can ask when AI projects come to your school district.
L.A.’s headaches are “coming again — bigger, harder and worse,” as school officials around the country are eager to put AI to best use, said Alex Molnar, a director of the National Education Policy Center at the University of Colorado at Boulder.
Molnar starts with the opposite premise from how we usually treat technology, as something we use first and iron out later.
He believes school districts shouldn’t use AI until it can pass a two-question test: Is it good? And is it the best way to accomplish the stated goal?
Molnar said he doesn’t know of any AI for education that has proved it checks these boxes, and therefore it shouldn’t be embraced by schools. Period. Students and teachers are using AI already, but it’s different when schools endorse it and use taxpayer money for it, as L.A. did.
What went wrong with AI at L.A. public schools
In March, the country’s second-largest school system by student enrollment touted a new AI project and chatbot called “Ed.”
Officials with the Los Angeles Unified School District said the AI would be a “personal assistant to students,” including by personalizing academic plans. A child whose math assessment scores were below those of peers might, for example, be recommended tailored educational games.
Students and parents could also ask the chatbot for resources if they needed help with reading or mental health struggles, or ask what was for lunch at the school cafeteria.
But within months, the start-up behind the district’s AI technology had financial problems. The chatbot was largely put on hold. And Los Angeles Unified started to investigate whether student data was misused in the AI project, according to reporting by the 74, an education news organization.
District officials have said that they will continue with the AI program and possibly restart the chatbot soon. Some parents in Los Angeles have questioned the decision.
“AI is a potentially transformative technology and it would be negligent not to explore its implementation in education,” a Los Angeles Unified spokesperson said. The district also said it’s “continuing to take steps to safeguard student data.”
The start-up, AllHere, didn’t respond to a request for comment.
Molnar said the “problem is much deeper and more profound than any technical failure” specific to L.A. or its choice of technology partner. “It’s the whole idea of it.”
He said school districts aren’t equipped to answer, nor do they usually ask, essential questions of AI projects: “What are we trying to accomplish, and is this the best way to accomplish it?”
As one example from the Ed chatbot, he said that many families do struggle to find information such as tutoring resources or cafeteria menu calendars from sprawling school bureaucracies.
Molnar said the root cause is that many schools can’t or don’t ensure that information like that is kept up to date or is easy to understand. A chatbot cannot fix the problem of garbage information.
Molnar also derided most AI billed as “personalized” learning as generic technology that just says your child’s name.
Questions to ask when your school district starts to use AI
Molnar suggested that parents try to assess whether AI works for your needs. Is the AI easy to use? Does it give you the information you want? In other words: Does it work? (Parents know from remote school starting in 2020 that a lot of the technology is awful.)
Second, be prepared to ask what alternatives the school or district considered.
If it’s a tutoring chatbot, for example, ask whether and how school officials assessed that the AI was “the best of all the options in the world,” Molnar said. Would it be better to instead spend the attention and money on extra staff hours to help children?
Third, ask how students’ personal data will be safeguarded. Molnar said those questions are often waived off with assurances that school technology only collects data in a way that is anonymous and can’t be traced back to a child. But anonymous data is almost never truly anonymous.
Molnar said it may be best for parents to pressure lawmakers to mandate that school districts won’t use AI until the company behind the technology proves it’s effective and won’t cause harm, and is held legally responsible for inappropriate uses of children’s data.
(In a recent research paper about AI in education, Molnar and co-authors suggested more ideas for policymakers and school officials.)
Surveys have found that Americans are generally skeptical about the benefits of AI, but Molnar and I also talked about countervailing hopes that technology can help solve our thorniest challenges in education, health care, climate change and transportation.
There’s something so human about wishing for AI that might understand your child, that has perfect information, and that’s creative and endlessly patient. Molnar gets it, but he also said it’s unrealistic to make AI into an embodiment of hope for our children’s future.