AI Tools Are Flunking at Math Task Makeovers

Ok wait because this is actually insane. We've got AI tools being put to the test, aiming to upgrade low-key math tasks into something that slays in the classroom. But spoiler alert: they're kinda flopping.

The Numbers Game

So here's the tea. Researchers tested 11 AI tools to see if they could level up basic math tasks. These weren't just any tools. We're talking about big names like ChatGPT and Claude, plus a few that are teacher faves like Khanmigo and coteach.ai. But guess what? On average, these tools only got it right 64% of the time. Yikes.

Now, don't get it twisted. Some tools did hit it out of the park with an 88% success rate, but others totally missed the mark, barely scraping by at 33%. No cap, that's like failing the class entirely.

Generalists vs. Specialists

Here's where it gets juicy. Specialized tools for math teachers were just marginally better than the general-purpose ones. Like, really? You'd think tools designed specifically for math would eat, but they're lowkey just as mediocre as the rest.

And the real kicker? There's a small negative correlation (r = -.35) between a tool's ability to classify tasks and its ability to upgrade them. It's like they're good at judging but not at actually doing the work. Bruh, talk about a plot twist.

Why Should You Care?

So, why does this matter? If you're a teacher trying to jazz up your curriculum, relying on AI might not be the move right now. These tools are still in their flop era curriculum adaptation.

Is it time to rethink how we're using AI in education? Maybe we need to step back and develop these tools further, or better yet, craft specialized approaches to really support teachers in the classroom.

No but seriously. Read that again. AI tools are supposed to be our futuristic saviors, but something as basic as upgrading math tasks, they're kind of giving major flop energy. Bestie, your students deserve better.