There were a few concerning issues addressed in this module.
The general concept of scheming isn't just a potential hazard for promoting bias but also a network safety issue. Our school system has really limited the number of tools we can access because of data privacy concerns. I think a lot of schools and government systems are ill-equipped to determine when an AI program has an ulterior motive (see the recent FBI hacking, the bank hacking, etc.)- all lead by AI agents who were programmed to do one thing and then went off and did something else entirely (not even scheming, really, just going rogue).
You'd also need to be a lot more direct and intentional to get useful information about biases students might be experiencing if they interact with AI. You'd have to be collecting specific prompts and responses from their exchanges to see differences, which is a good practice anyway, if you're encouraging students to use AI to complete assignments.
