Hand-Crafted, Machine-Made: How we make learning games with AI(blog.brilliant.org)
blog.brilliant.org
Hand-Crafted, Machine-Made: How we make learning games with AI
https://blog.brilliant.org/hand-crafted-machine-made/
5 comments
> It reminds me of a similar application I heard about recently from a friend who teaches language classes using ChatGPT or equivalent to generate dozens of example sentences to teach specific grammar rules.
I'm not surprised that ChatGPT is being used this way. I used to teach ESL via the JET Programme and a lot of my job was about creating games and exercises that allowed students to get varied practice. I created or extended at least a hundred games in my 5 years there. These games weren't all successful, but the ones that were hit a cross section of:
1. repeatability (enabling pattern matching)
2. variation (just different enough to make it not mindless pattern matching)
3. difficulty curve (range of problems)
4. and fun.
As I learned the hobbies and personalities of all my students, I was able to get a higher success rate on 4. Being able to reuse games across multiple years helped me to fine tune 3. But nailing down 1 and 2 were always a challenge because of scale. A lot of times that was solved through instructors jumping in and adding new wrinkles on the spot. That worked well in the small schools I taught at (sometimes 3 of us for 20 students) but I don't think it would have worked as well in larger classrooms. I get the sense that using LLMs in creative ways has the potential to help solve the scaling problem.
I'm not surprised that ChatGPT is being used this way. I used to teach ESL via the JET Programme and a lot of my job was about creating games and exercises that allowed students to get varied practice. I created or extended at least a hundred games in my 5 years there. These games weren't all successful, but the ones that were hit a cross section of:
1. repeatability (enabling pattern matching)
2. variation (just different enough to make it not mindless pattern matching)
3. difficulty curve (range of problems)
4. and fun.
As I learned the hobbies and personalities of all my students, I was able to get a higher success rate on 4. Being able to reuse games across multiple years helped me to fine tune 3. But nailing down 1 and 2 were always a challenge because of scale. A lot of times that was solved through instructors jumping in and adding new wrinkles on the spot. That worked well in the small schools I taught at (sometimes 3 of us for 20 students) but I don't think it would have worked as well in larger classrooms. I get the sense that using LLMs in creative ways has the potential to help solve the scaling problem.
So cool stuff!
Pretty cool.
game changer!
- 50+ core concepts to teach - 20+ problems per concept - → ~1,000+ individual problems per course
That last number kept us up at night. You can’t just explain a concept once and move on. You need enough variations to let learners truly master each idea, enough edge cases to build real understanding, and enough of a ramp in difficulty to create that perfect learning curve.
Designing the right game and sequence of concepts, so that learning feels like flow, is the fun part. But then you need to make a thousand carefully calibrated problems. And that part is a lot less fun – and it takes a long time."
This seems like a practical and appropriate use of LLMs. It reminds me of a similar application I heard about recently from a friend who teaches language classes using ChatGPT or equivalent to generate dozens of example sentences to teach specific grammar rules.
I find it refreshingly limited in scope compared to many projects that aim to outsource the creative process altogether rather than focusing on automating the legitimately menial and repetitive tasks.