Note: This post is a followup to a piece I had written earlier. You can find that post Why Sal Khan’t: On Learning by Making but Teaching by Telling. This post was also cross-posted on the Civics of Technology blog.

“Students aren’t great at asking questions well.”

This was Kristen DiCerbo, Khan Academy’s Chief Learning Officer, in a recent Chalkbeat article. She was offering her explanation for why Khanmigo hadn’t lived up to its promise.

I was reminded of a story about the physicist and Nobel laureate Isidor I. Rabi, who was once asked, “Why did you become a scientist, rather than a doctor or lawyer or businessman, like the other immigrant kids in your neighborhood?” Rabi’s answer is worth quoting in full:

“My mother made me a scientist without ever intending it. Every other Jewish mother in Brooklyn would ask her child after school: ‘So? Did you learn anything today?’ But not my mother. She always asked me a different question. ‘Izzy,’ she would say, ‘did you ask a good question today?’ That difference, asking good questions, made me become a scientist!”

So of course asking good questions is important. But why are students bad at asking them? That raised many questions in my head. Questions about questions. Here are three.

---

1. What makes a good question?

There’s a concept in psychology called the curse of knowledge. Once you know something, you can’t un-know it, and you lose the ability to reconstruct what it felt like not to know it. A beautiful demonstration comes from Elizabeth Newton (popularized by Heath and Heath in their bestseller Made to Stick): she had people tap out the rhythm of a well-known song and asked them to predict whether listeners would recognize it. Tappers predicted 50%. Listeners got it 2.5% of the time. The tappers could hear the melody in their heads, and they literally could not imagine not hearing it.

DiCerbo’s statement is the curse of knowledge in action. From the perspective of someone who already understands the content, a “good question” is one that’s well-formed and targeted. It identifies a gap and asks for the right piece to fill it. That’s a reasonable definition if you already have the map and just need directions to one location.

But learners don’t have the map. That’s what makes them learners. The inability to articulate what one needs to know is part and parcel of the condition of not-yet-knowing. It is, quite literally, what being a learner means.

So when DiCerbo says students aren’t great at asking questions, she’s describing a system that requires a capability that is itself the product of learning. You need to already understand something in order to ask about it well. That’s circular, and it’s the kind of circularity you can’t see when you’re standing inside the curse.

What Rabi’s mother did, by asking him to ask good questions, was nudge him to think in a meta way about what he knew and what he didn’t. A good question in that context is one that acknowledges the landscape of not-knowing.

Helping students figure out what to ask, giving them enough scaffolding and context and permission to fumble toward the question that matters, is teaching itself. It is not a precondition for it.

---

2. What does “Socratic” actually mean?

Every AI tutor on the market seems to describe itself as “Socratic.” It has become the go-to word in edtech for “doesn’t just give the answer.” But it’s worth asking what Socrates actually did, because it wasn’t what these systems do.

Socrates asked the questions. His interlocutors answered. The diagnostic labor sat with the teacher rather than the student. Socrates would observe where someone’s reasoning broke down, and then craft a question designed to expose exactly that fault line. Meno came to Socrates with a big, messy question: can virtue be taught? Socrates didn’t say “be more specific.” He took that rough question and worked with it, turning it, reframing it, asking smaller questions that exposed what Meno thought he knew but didn’t.

That last part is important. Socratic irony, the teacher pretending not to know, is a pedagogical technique. It’s the deliberate construction of ignorance to create a productive space. You say “I don’t understand, help me see it,” and suddenly the student is the one doing the explaining and the sense-making and the reasoning. But this only works if the teacher knows exactly what to pretend not to know. It’s improvisational and relational, and it depends entirely on reading the specific person in the specific moment. It requires care and attentiveness, and produces an emotional experience that showcases the humanness of teaching and learning.

The word “Socratic” has been emptied of all of this. What’s left is a thin definition: the system won’t give you the answer, and it will ask you questions instead. But asking questions isn’t the same as doing what Socrates did. There’s a difference between asking a question and knowing exactly what to ask, when, and why. It is no surprise that students just respond “idk” over and over again. 

---

3. What does a tutor actually notice?

Dan Meyer recently asked teachers on Twitter a simple question: if you’re tutoring a kid one-on-one, under what circumstances do you decide to intervene?

Nobody said “when the student asks.” Instead they described what they see. The deer-in-the-headlights stare. The moment the gears stop turning. The gap between what a student says aloud and what they write down. The point of hesitation where a student begins to realize their approach isn’t working. One teacher put it perfectly: “When I can no longer see the gears turning in their head, which means I think they’re stuck in an unhelpful way.”

These are acts of perception. The tutor’s most important skill is noticing. And noticing is exactly the thing DiCerbo’s framing skips over entirely. Her model assumes the student initiates (“students aren’t great at asking questions”). But the teachers Meyer talked to don’t work that way. They watch. They read faces and body language and hesitation and silence. They make a judgment call about whether a pause means thinking or shutdown. And then they intervene with the right question at the right moment.

A chatbot, as Meyer points out, can sense none of these signals. It doesn’t know when you’re thinking productively and when you’ve given up. It doesn’t know when to speak and when to stay quiet. It has no access to the most important data in the tutoring relationship, which is the data that lives on the student’s face.

---

Three questions, three different angles, all leading to the same conclusion: the burden of asking the right question, knowing what to ask, noticing when to ask it, all of it sits with the teacher, not the learner. It starts from a simple recognition that confusion, and the inability to articulate what you need, is where learning begins, and that expecting otherwise of a learner gets the whole thing backwards.