Teaching Judgment in the Age of AI

Ask an AI model to generate a picture of a nurse and a doctor standing side by side, and see which one it makes a woman. 

HIS teachers ran that exact experiment on themselves during a training session in August, then had to explain the results to each other.

Different AI approaches for different scenarios

Our teachers spent three days of this year's in-service, August 5 to 7, on AI training, working through scenarios such as a student essay that needs feedback, a parent email about a child's grades, and a rubric for a new unit.

HIS doesn't have a blanket AI policy. There are instances when teachers and students can use AI and there are situations where it's not permissible.

There is a set of questions teachers are trained to ask before every decision on whether or not to use AI: Is this predictable? Is this useful? Is this appropriate? And who gets hurt if it's wrong?

That last question came up more than any other.

The training was led by Fiona Reynolds, Chief Learning Officer at UnconstrainED, a US-based firm that works with schools on AI literacy for both staff and students. Reynolds has presented for NESA, the regional association covering international schools across the Near East and South Asia, and spent August 5 to 7 walking HIS staff through actual classroom situations rather than abstract policy.

One scenario asked whether AI-drafted feedback on student essays was appropriate. The answer, for HIS, was no. Feedback has to come from the teacher directly, because a student's growth depends on being seen by someone who knows them, not summarized by a system that doesn't.

Teachers also worked through where AI earns its place. Building a rubric that tracks a unit's progression, or checking whether a lesson plan actually teaches the standards it claims to teach, is the kind of task AI handles well: high frequency, lower stakes, and easy for a teacher to check against their own expertise. 

 

Drafting a difficult message to a family about declining grades sits at the other end. Rare, high stakes, and something a teacher should sit with rather than delegate to a bot.

The nurse-and-doctor exercise wasn't really about catching AI in a mistake. It was about showing teachers how to train their students to ask why a machine produced the answer it did, and to recognize that every model carries the assumptions of the data it learned from.

Understanding the AI context issue

The same problem shows up closer to home. A student researching Great Zimbabwe through a general-purpose AI tool will find far more written in English than in Shona, and more written from a colonial lens than a local one. 

This means that a lot of the local context is never captured and this information can therefore not be treated as a complete record. 

This fits well with a curriculum that already asks students to hold their Zimbabwean context and their international one at the same time.

It also aligns with the Diploma Programme Theory of Knowledge subject which asks students to interrogate the validity of ‘knowledge.’

Teaching students to think differently

The ultimate lesson was that none of this new technology replaces what a teacher does. If the pandemic taught schools anything, it's that a child's relationship with the adults teaching them carries more weight than any tool placed in front of them. 

Yes, AI makes certain processes more efficient. But the big change at this academic level is the kind of thinking we ask students to practice: not just finding an answer, but deciding whether the answer deserves to be trusted, and how to arrive at one that does.

That is the skill our students will need long after they leave HIS, whichever university, workplace, or country comes next.

This training is part of a broader push at HIS to build AI literacy across the school, with future sessions planned to continue the training school-wide.

At the same time, schools around Africa and the rest of the world, are tackling AI. The Government of Zimbabwe recently announced that is moving to make artificial intelligence (AI), coding, and robotics compulsory from primary through secondary school as part of a national curriculum overhaul.