What ethical principles should guide the use of AI in Education?

I have dear friends and colleagues who are very excited about what artificial intelligence (AI) can offer education. There's a lot of chatter about tools like ChatGPT and Khanmigo. The potential is clear: imagine personalized, patient, on-demand tutoring for every student in the world. That's just one of many potential scalable solutions AI can offer education (e.g., another exciting area involves helping teachers develop lesson plans). I certainly hope my friends and colleagues turn out to be right and that AI meets these potentials, but I have to say I'm less enthusiastic. Every time I hear a new way that AI is going to "revolutionize" education, I think this:

Unfortunately, quite a bit can go wrong due to the way many of these large language model AI systems work. For example, AI math tutors sometimes mess up basic math. And, AI can recreate societal bias and prejudice. Therefore, I think more work is needed on the ethics of using AI in education: we need guidelines to understand how to evaluate these AIs and their effects on students, teachers, and society.

That's why I was pleased to see that Şenocak et al. (2024) have conducted a scoping review of nine international AI in Education guidelines to identify 12 common principles. They found the following:

"The results of the study also show that while 9 ethical principles (i.e., “transparency, diversity and equity, accountability, privacy and data protection, security and safety, sustainability and societal well- being, democratic participation in education policy planning and AI practices, empowerment of teachers and teaching, empowerment of learners and learning”) are recognized and embraced by these nine international AIED guidelines, to our knowledge, ethical principles such as “autonomy, ethical design and commercialization” were not highlighted in all the documents examined. These findings bring us to the conclusion that in general, there is a global consensus regarding the ethical principles of AIED, and yet more emphasis should be placed on the three ethical principles mentioned above" (pp. 203-204).

In the article, they unpack each of those principles and discuss why they are important. I think this kind of review is a great step toward establishing a common set of guidelines for not only evaluating AIs in Education, but also for designing them. Ultimately, we want "ethics-guided design" AND rigorous, ongoing evaluation of AI in Education implementation and effects. These 12 principles are a great place to start and I'd be excited to see companies and governments adopt them.