Google Classroom Gets Better

Five years ago, I'd written a blog post on features that I would have loved within Google Classroom. I used to work at the Gokhale Institute back then, and we were in the Lockdown era, so a lot of that post was about how Google Classroom could be made better from an administrative perspective.

Yesterday, Google published a blog post of their own, on recent Gemini-powered advances coming to Classroom, and a lot of what I was wishing for back in 2020 is now very much here.

---

But 'Tis The Age of AI, After All

But a post written in 2020 is so...2020.

What can we unlock with AI in classrooms that will make the learning experience better for all concerned? Note that my thoughts here are that of a person who has worked in higher education, and much of what Google's blog post is about is directly relevant to students in school. But that being said, there is, of course, a significant overlap.

1. First, Gemini is now integrated into Google Classroom, and it can be used to generate a lesson plan. That first draft of the lesson plan can then be further refined (with the help of Gemini if you so prefer, of course), and Gemini can also suggest relevant videos, and other material.
   Scale, quality and customization form the impossible trilemma of non-AI education. What AI unlocks is the ability to do all three at the same time, and better relative to non-AI education. This is not a guarantee, but the possibility exists. The reason I bring this up is because I'd argue that it is possible to come up with lesson plans (plural) now.
   You know what you want to teach, of course. But what if you (and Gemini) also know what each student knows about the subject. You know what each student has learnt and has learnt well. You also know how they like to learn, and what else they like to learn. Then why not use this information to build out customized lesson plans that take all of this into account?
   One student might get an explanation of comparative advantage using equations, but no diagrams, while another might get the same topic explained, but using words and a video. Everybody will have access to all of the material, but which of these is the focus, and which of these is the additional material can and should change given learning abilities and preferences.
   Once a student is comfortable with a particular topic using their preferred "in", you can show and discuss other approaches - and here, Gemini can help. Study groups can be formed where students with different learning approaches discuss the same topic, for example. I've learnt about comparative advantage by listening to a podcast, whereas you've seen a video, and our friend here knows the equations and what they mean. Our discussion about the topic is likely to be much richer, and we all gain perspectives that our approach may not have made clear (and again, here too, Gemini can help).
   I'm very much on board with using Gemini to make lesson plans, but more can (and should) be done.

2. NotebookLM and Gems are now baked into Google Classroom. This is excellent, and a welcome development. I am well aware of the fact that this is meant for schools, but at least in higher education, it is perfectly possible to do this not as a teacher-led activity, but a student led activity. Study groups can create their own NotebookLM based study guides (complete with audio overviews) and these artifacts can be endlessly remixed. Imagine you as a student have access to all of the study material uploaded by all of the study groups in your class. You can ask for an overview of comparative advantage by focusing on examples drawn from agriculture. Or gender studies. Or historical examples. Or examples related to technology. All RAG-based, of course. Your teacher can set the condition that certain PDF's must always be included in all overviews, but the list of complementary materials is entirely up to you, and you can use your own, or mix and match. Hell, you can have meta audio-overviews! RAG based learning, where the retrieval is across material collated (and created) by the entire class, sounds like an excellent idea to me.
   And ditto for Gems, of course.
   There will be issues relating to quality, pranks being attempted, and privacy concerns. But I'd think the potential gains are worth the experimentation, and again, surely Gemini can help out here.

3. There is talk of tracking progress against learning standards and skills, and this will be launched in the coming months. AI allows you to track progress across much more than that, and we should be making good use of this opportunity. Track progress by method of learning, for example. Does over-reliance on audio overviews impact learning outcomes? For which students, for which subjects, under what method of testing? How much do we learn about the student(s) in question, and how much do we learn about the tests that are being administered?
   Is the group of topics that students are struggling with remain the same across cohorts, or is it changing? Which set of learning content is a good predictor of good learning outcomes? Again, privacy/ethics concerns are very real here, but as always, there is potential for Gemini to help out in these cases.

4. Wittgenstein's Ruler popped up in my head when I read this:
   "We know each institution requires a specific set of learning standards, some of which are custom. In partnership with 1EdTech and Common Good Learning Tools, any institution or standards-issuing body can host and publish their learning standards via CASE Network 2, enabling a scalable, interoperable solution for making your learning standards available in Classroom. In advance of the Classroom launch, we encourage institutions to add their standards to CASE Network 2 now to make them readily available."
   That is, I suspect that in the years to come, we will learn more about the standard of learning standards than we will about how schools (and students) are doing on them.

5. Recursive Google Classroom Intelligence:
   "Teachers visiting their class pages will now see a new “Analytics” tab. This shows a centralized view of student performance and engagement in Google Classroom, now with insights that help educators understand when students may need additional support, such as which students are consistently missing assignments or whose grades are improving. Actionable insights are also shown directly on the Classwork page, where teachers can see whether students have made progress with an upcoming assignment and send a quick reminder."
   Figure out a way to feed this into NotebookLM and Gems, at the individual level! When the student interacts with either of these two, they "know" when students need additional support, and they have access to these actionable insights, so that their answers can be customized accordingly. Certain material can be emphasized, or elaborated upon, or more examples can be given, as per the (ongoing) assessment of the student.

6. Educators (and presumably students too, soon) can now generate videos. Can you imagine what students will be able to create in the years to come? Imagine being able to improve your understanding by creating and sharing imaginatively done videos related to the subject at hand. WATTBA!

7. Google Classroom will soon have AI generated questions for educational YouTube videos, built right into the app. Again, if this can be piped together with NotebookLM and Gems, life will be very, very good.

8. There is no mention of this in the document, but how long before Gemini's inputs are taken into consideration while grading? Who better than Gemini to understand exactly how much the student in question has understood, both in an absolute and a relative sense (and relative to all cohorts over time, not just this current cohort)? How long before students and parents start preferring Gemini's grading?

9. The entire (rather lengthy) list is here, if you'd like to take a look.

By the way, if you are an educator, you may want to consider doing this course.