Learning Science
Why AI Chatbots Make Bad Tutors — and How Aristotle Is Different
Most AI tutors are generic chatbots with a new coat of paint — ask a question, get a reasonable-sounding answer, close the tab. We built Aristotle around how students actually learn and how the best teachers teach. Here's what's under the hood, and the research behind it.
Your child talks, Aristotle listens
Aristotle is voice-first: the student speaks through problems out loud while Aristotle listens and responds back out loud, draws on a live whiteboard, and works through equations and diagrams in real time.
We based this design on the ICAP framework, one of the most cited findings in learning science. ICAP establishes a hierarchy of engagement: interactive beats constructive beats active beats passive. Typing into a chatbot and reading the answer is passive. Explaining your reasoning out loud to a tutor who responds to it is interactive, and consequently produces the deepest learning.
Aristotle guides. It doesn't give answers.
Aristotle guides through Socratic dialogue. It asks your child to explain their thinking and nudges them until they get there themselves. Students who explain their reasoning learn dramatically more than students who simply receive explanations. That's why Aristotle asks your child to explain their thinking instead of handing over answers.
Other LLMs are hardwired in every way to give you answers and help you solve problems, which is useful in almost every other setting, but a huge problem when it comes to teaching well. Overcoming that has taken a lot of hard engineering work.
In one study, LLMs taught to utilize Socratic dialogue were already shown to significantly outperform regular models like ChatGPT and Claude on teaching tasks. Aristotle uses this exact principle to make sure that students truly learn.
It diagnoses misconceptions, not just mistakes
When your child gets something wrong, a generic chatbot corrects the answer and moves on. Aristotle does what expert human tutors do: it identifies the specific error, figures out the flawed reasoning underneath it, and chooses a remediation strategy before responding.
This three-step process — inspired by a Stanford methodology called the Bridge framework — separates expert tutors from novices. Aristotle addresses why your child is confused, instead of staying on the surface.
It tracks what your child actually knows
Underneath every session, Aristotle maintains what we call a subject map: a structured mental model of a subject's component topics and how they depend on each other, with a live status of which ones your child has mastered and which they are working towards.
This draws on a field of AI research called knowledge tracing: modeling a student's knowledge from their history of work and predicting where they'll struggle next. And because Aristotle remembers every session, that map keeps getting richer. Session twelve builds on sessions one through eleven. Switching from calculus to chemistry doesn't mean starting over — Aristotle retains a deep understanding of the student and how they learn across subjects.
Every response is verified
In real-time, all tutor actions are checked against pedagogical criteria. Is this response actually guiding them toward understanding, or just toward the answer? This turn-by-turn verification approach — inspired by research out of the University of Michigan — achieves significantly higher learning task completion than unverified tutoring.
Does this actually work?
There is already research that shows AI tutoring systems built on structured models of student knowledge can reach effect sizes matching human one-on-one tutoring. More recently, a randomized Harvard trial found students using an AI tutor learned more, in less time, than students in active-learning classrooms. A study across five UK schools found AI-tutored students performed as well as human-tutored ones and were better at solving novel problems afterward. Our data from early users of Aristotle has shown improved test scores, understanding at mastery levels, and increased curiosity-driven learning.
One-on-one tutoring has been the gold standard since Bloom described the 2 Sigma Problem in 1984. The bottleneck was always access. That's the problem Aristotle solves.
And it gets better every day
Most tools your child opens forget them the moment they close the tab. Aristotle doesn't. Every session, it understands your child a little better — where they get stuck, how they think, what makes learning come to life for them. Every session, these insights compound into something no chatbot can offer: a tutor that actually knows your student. We'd love for Aristotle to meet yours.