The fastest way to help a stuck student is to show them how. It is also the surest way to have them stuck again next week.
Every parent knows the moment. Your teenager stares at a math problem and says "I don't get it." The kind thing seems to be to show them. A good tutor does something harder: they wait, ask one question, and let the student keep going.
Sophie is built to do the harder thing. Here is the research behind it, in four findings, and then what Sophie actually does.
1. Trying first makes the lesson stick
A math class usually teaches a method, then students practice it. Manu Kapur, a learning scientist now at ETH Zurich, has spent years testing the opposite order: students first try a problem nobody has taught them to solve, and only then get the lesson.
In one study, 140 ninth graders learned variance, a way to measure how spread out a set of numbers is. One group got the lesson first. The other group first tried to invent their own measure of spread, with no help. Most of their inventions did not work. Then they got the same lesson.1 Across Kapur's studies, both groups learn the formula about equally well. The students who tried first understand the idea better, and do better on new kinds of problems.2 Kapur calls it productive failure.
One study can be luck. So in 2021, Tanmay Sinha and Kapur pooled 53 studies and 166 comparisons. Trying first won on understanding and on new problems, with an effect size of 0.36. When the trying was designed well, it reached 0.58.3
Figure 1
Trying first beats being shown first, most of all when it is done well.
One honest limit: for young children, in grades 2 to 5, the lesson first did better.3 Sophie works with high school students.
2. Even a wrong guess helps
Kapur's students failed, and still learned more. Memory research explains part of why. In 2009, Nate Kornell, Matthew Hays and Robert Bjork asked people to guess answers they could not know yet, then showed them the right one. People who guessed wrong first remembered the answer better than people who simply studied it for the same amount of time.4
Trying to answer wakes up everything you already know about the question. When the answer arrives, it has somewhere to land.
Bjork has a name for this family of effects: desirable difficulties. Some things make learning feel harder and work better.5
3. Struggle feels like failure, even when it works
Here is the catch. Struggling does not feel like learning.
In 2019, a Harvard team taught the same physics lessons two ways. In some classes, students listened to a polished lecture. In others, they worked through the problems in small groups, and the teacher explained after. The students who did the work scored higher on the test. They also said they had learned less.6
The effort itself felt like a sign of not getting it. We all read a smooth explanation as learning, and hard thinking as failing.
This matters for any tutor. If you judge a session by how good the student feels at the end, you will show the answer every time.
4. Help that comes too soon backfires
Tutoring software learned this the hard way. Many programs offer a ladder of hints that ends with the answer. Researchers at Carnegie Mellon watched students rush down the ladder to the last rung, a habit they called gaming the system. Those students learned less.7
A chatbot is a ladder with one rung. In a large trial, high school students who practiced with plain ChatGPT solved more problems, then scored 17% lower on the exam.8 We wrote about it in Why Sophie is not a chatbot.
Stuck, but never alone
None of this means leaving a student stuck. Kapur's word is productive: the problem has to be one the student can get a grip on, and someone has to bring in the real idea after. Stuck alone, far from the answer, most students simply give up.
In 1976, David Wood, Jerome Bruner and Gail Ross watched a tutor help young children build a block pyramid. They called good help scaffolding: the tutor takes over only the parts the child cannot manage yet, and steps back as the child takes them on.9
How Sophie does it
Sophie is built to keep students in the good kind of stuck: working hard, close to the answer, never alone.
- The student goes first. Sophie keeps the answer and the correction to herself until the student has tried.
- She starts with what is right. Before any hint, she names the steps the student got right, precisely: "Nice, you fixed the sign."
- A hint is a question. One question in plain words that points at where to look. Never the next step done for them.
- She checks the rule before the step. If the student is missing a rule, she names it and asks if they know it. If they don't, she writes it on a short sheet, in general form, with an example from a different problem.
- A right answer still needs a reason. When a question asks for a justification and the student gives only the answer, she asks how they got it.
- Judgment, not a counter. There is no fixed number of hints before the answer appears. Sophie shows the solution only when a student is truly stuck and asks for it, or says yes when she offers. Then she goes through the correction with them, and offers one more problem like it.
At home
You don't need a tutor to use this. Next time your teen says "I'm stuck":
- Ask what they have tried before you explain anything.
- Ask one question instead of giving a step: "What is the problem asking you to find?"
- Praise the effort, not the speed. A hard evening of math is often a good one.
The hardest part of tutoring is not explaining. It is waiting. Sophie waits, then helps just enough.


