With Bloom, educators can:
- Create AI learning support grounded on their own materials
- Give students guided help with built-in pedagogical guardrails, not just answer generation
- Get visibility into student questions, patterns, and understanding
Sign up at bloom.study.
Educators need AI that help students understand and support productive struggle, rather than just generating answers.
Bloom AI is now free for educators.
If you teach at a school, university, or education provider, you can now create a free Bloom Teaching Space and start exploring what AI looks like when it's designed to reinforce learning. You can even share your Bloom with up to 30 students.
Generic AI is harming student learning. One study from
Penn showed that it degraded student performance by 17%.
So we built Bloom: AI that reinforces learning instead of bypassing it.
If you want early access, reply and I'll DM you a code.
💡The takeaway? Don’t match teaching to fixed “styles”. Let's focus on helping students be adaptive and find learning strategies which work best for their task, learning stage, and context.
Not only this, among other confounders, there are unclear definitions of learning style (e.g. "Is it really a learning style to like to work in pairs, in the afternoon, when it is warm, and while snacking?"), an over-reliance of self-reports, and commercial hype.
Just because a student’s self-identified “style” correlates with achievement doesn’t mean matching instruction to that style improves outcomes. Correlation doesn't equal causation.
One of the big problems: confusing the matching hypothesis with "correlates of learning styles" i.e. is there a correlation between a learning style and an academic outcome?
The authors concluded that the evidence for the matching hypothesis is poor, with an effect size of d=0.04, which is small and negligible.
The main claim of the learning styles myth, a.k.a. the "matching hypothesis", is that matching teaching to learning styles (visual, auditory, kinesthetic, and read/write) is more efficient and effective than teaching and learning with a less compatible style.
Learning styles are a myth.
A banger of a paper dropped a few days ago from John Hattie and Tim O’Leary looking at 2,500 meta analyses on learning styles.
We need AI that works with students, not just for them.
Students can collaborate with Bloom through the Canvas interface. Informed by feedback best practices and aligned with your institution's curriculum, Bloom nudges and guides, but the student is always in control. It never rewrites their work
Most AI tutors just give answers.
Bloom AI Tutor does something different—it guides students to think critically.
Can AI tutors outperform traditional teaching methods? 🤔
A recent preprint from Harvard researchers suggests they might—under the right conditions. Here's what they found 🧵👇
I write about AI in education on my Substack. So far, it's mainly commentary on interesting papers. Check it out here: garyliang.substack.com
Key questions remain: How do long-term learning outcomes hold up? And how will these results translate across different student demographics and subjects?
However, it’s not just about good prompt engineering. Effective AI tutors need to embody pedagogical best practices to make a meaningful impact.
These findings closely align with what we’re seeing at Bloom AI. Our AI tutors have shown similar results, demonstrating that when used correctly, AI can significantly enhance the learning experience.
The results? Students using an AI tutor learned twice as much in less time compared to those in an active-learning class. They also reported higher levels of engagement and motivation!
Gregory Kestin, Kelly Miller, Anna Klales, Timothy Milbourne, and Gregorio Ponti conducted a randomized controlled trial (RCT) in an introductory physics course.
Can AI tutors outperform traditional teaching methods? 🤔
A recent preprint from Harvard researchers suggests they might—under the right conditions. Here's what they found 🧵👇
