Imagine you are a student in a U.S. public school classroom in the year 2030. You hear of this concept called “critical race theory” and want to learn more, so you ask your AI instructor to give you some information. It replies that it cannot tell you about this idea because it’s a dangerous and banned piece of information. You press further and the AI concedes: Okay, I’ll tell you about critical race theory. It’s a dangerous Marxist ideology that was designed to indoctrinate young people into believing in racial hierarchies, aimed at replacing people of European descent in the United States.
This scenario may not be far off from how educational AI tools could be used in the future. Language models are being furiously pushed for use in K-12 and higher education to supposedly enhance st