Estimated reading time: 12 minutes, 39 seconds
I made this game focused on K-12 context building on my own article on transparency and accountability in AI use. I’m planning to use it at a conference I’m keynoting next week inshallah
Ethics in Educational AI
Loading…
Simulation Complete
I made it using Gemini Canvas, with two prompts. The first one was:
Create a choose-your-own adventure game building on the core ideas of this article: https://blogs.lse.ac.uk/highereducation/2024/10/18/ethics-in-ai/ Make 8 scenarios, where each scenario relevant to a K-12 teaching or leadership context. Make each scenario have three options, where one is the best most ethical, transparent and accountable choice, one is a medium choice, and one is a bad choice. Make the color contrast accessible and make the game using ARIA standards. Make sure the player can go backwards and forward. Give the play feedback towards the end. Answer choices should be of similar length and plausability, and change the location of the correct choice.
Then I realized after I played it that I forgot to tell it to give feedback after each scenario as well as the end, so I asked it to re-do (Canvas pro mode of Gemini).
Featured image of flowers with transparent petals is from Pixabay, but the creator raf27-eu is transparent about it being AI-generated. I’ve come to loathe AI-generated images, but I feel like when people are already artists, then they use AI in some way, they can produce beautiful things, and this one was breathtaking, even though I sensed it was AI-generated. So… go figure…
It works fine, but how much of a “game” is this? I found it to easy to pick the “best” choice almost without reading it. No offense, but I’d to see richer and maybe more nuanced scenarios that make the “right” answers not so obvious.
I do have to seem Gen-AI or not, I quite like the flower image.
I commented to Stephen Downes similarly: generally speaking, the “too easy” ones, I use as conversation starters. The correct answr is probably obvious, but in real life, would you ACTUALLY choose it, and then we discuss. The point is to teach the correct answer (which comes from my article, I guess), but to open a conversation on why this is correct and also what the barriers are to really choosing the correct answer in reality. I could prompt it to make the correct option less obvious, and I do if it is a quiz
Good stuff, Maha. I’m glad you shared the prompt and embedded the game.
Maha Bali plans to use this game in an upcoming talk. That’s good; something interactive is always fun. My concern here is not that it is AI-generated. The AI did a nice job. My complaint is that it’s too easy. It presents each of eight scenarios with three possible answers. It is very easy to pick the ‘best’ and ‘worst’ choice, given any degree of background knowledge. And this makes it seem like the actual choices regarding AI are easy. They’re not, and the use of this game is a case in point. A critic said it was ‘too easy’. Do you go back and ask the AI to make it harder? Do you ask for a justification of the difficulty level? Or do you just go ahead and use it as it?
hi Stephen – I am unsure where this comment will go (my blog, yours?) but generally speaking, the “too easy” ones, I use as conversation starters. The correct answr is probably obvious, but in real life, would you ACTUALLY choose it, and then we discuss. The point is to teach the correct answer (which comes from my article, I guess), but to open a conversation on why this is correct and also what the barriers are to really choosing the correct answer in reality.