Orbrya
All posts
Orbrya2026-03-12

The Two Types of Critical Thinking Every Parent Needs to Understand

Researchers have identified two distinct forms of critical thinking when kids use AI. Only one of them predicts outcomes that matter. Here's the difference — and why schools are teaching the wrong one.

The Two Types of Critical Thinking Every Parent Needs to Understand

Published: Week 3 | 8 min read

In our last post on the school policy gap, we promised to cover the two types of critical thinking identified by researchers at King's College London, and why only one of them predicts the outcomes that matter. This is that post. The distinction is straightforward to explain and surprisingly hard to unsee once you understand it — it reframes almost every conversation about AI education in a way that makes the current approach look incomplete.

Picture this: your child finishes a research project in less than an hour. The writing is clean, the structure is solid, the sources are cited. You ask what it's about and get a reasonable summary. You ask a follow-up question and the answer starts to thin out. You press a little further and realize they can describe what the AI told them but not quite explain the ideas behind it.

Something got done. Learning is harder to confirm.

This is the situation that a growing number of parents describe, and it maps almost exactly onto a distinction that researcher Chahna Gonsalves at King's College London identified in a 2024 paper published in the Journal of Marketing Education. Her research wasn't aimed at parents or homeschoolers. It was focused on how students in higher education engage with AI during learning tasks. But the core insight transfers directly to any family thinking about how their child is developing alongside these tools.

Gonsalves identified two separate forms of critical thinking that come into play whenever someone uses AI. They sound similar. They're not.

The Distinction That Changes Everything

The first type she called critical thinking for the assignment: using your reasoning to analyze, synthesize, and apply information in order to complete a goal. This is the kind of thinking educators have always valued and tried to cultivate. It's what a student is doing when they evaluate sources, construct an argument, weigh competing evidence, and produce original work. When AI educators talk about "using AI thoughtfully" or "AI as a tool for learning," this is almost always what they mean. The student uses AI to help with the task, and then applies their own judgment to shape the output into something worth submitting.

The second type she called critical thinking toward the AI: directing your attention not at the assignment but at the AI itself. Is this output accurate? What assumptions is it making? How would I verify this claim? What is this system likely to get wrong, and how would I know if it did? This type of thinking isn't about using AI more thoughtfully. It's about being skeptical of AI as a source before incorporating what it produces into your thinking at all.

The difference between these two types sounds academic until you sit with it for a moment. Critical thinking for the assignment assumes the AI's output is basically usable and works with it from there. Critical thinking toward the AI treats the output as a claim that requires evaluation before it earns the right to be used. One type is downstream of accepting the AI's output. The other is upstream of it.

Almost everything being taught under the label of AI education right now develops the first type and leaves the second largely unaddressed.

Why This Gap Is So Consequential

Gonsalves framed her research around Bloom's Taxonomy, the framework educators have used for decades to describe different levels of cognitive work, from basic recall up through analysis, evaluation, and creation. Her argument is that the arrival of AI requires a rethink of where certain cognitive tasks actually live on that taxonomy, because AI can now handle the lower levels reliably and the upper levels inconsistently.

The implication isn't that lower-order thinking no longer matters. It's that students who only learn to use AI for the lower-order work and then apply their judgment to the result are operating on a different set of assumptions than students who also know how to interrogate what the AI gave them in the first place.

The practical consequence shows up most clearly when the AI is wrong.

AI language models produce incorrect information regularly, and the errors aren't random. They tend to cluster in specific places: recent events, obscure specifics, citations and sources, complex causal claims, and anything requiring updated information that postdates the model's training. A student who has learned to use AI thoughtfully for their assignment may still produce high-quality work most of the time, because AI is often right on well-established topics. But they haven't developed the habits to catch it when it's wrong, because their attention has been trained on the output rather than on evaluating the output.

A student who has also developed critical thinking toward the AI knows to pause before accepting any AI claim as usable material. They know to ask how they would verify this, what the AI's likely failure modes are on this type of question, and what it would mean if this particular claim were wrong. That pause is a skill. It requires practice. And it's almost never explicitly taught.

Joshua Lockair, a homeschool parent who also teaches computer science at the university level, described the issue precisely: "Students must understand how AI works to evaluate whether it's giving the right answer. Even if it does give the right answer, they're not developing the process or skills to understand the concept. It's like skipping to the end."

That's the gap Gonsalves' research is pointing toward: not that students are using AI, but that the skills being built around AI use don't include the evaluative layer that makes that use actually safe.

This Is Not a Criticism of AI

It's worth being clear about what this distinction doesn't mean.

It doesn't mean AI is bad for learning. It doesn't mean children should avoid it or families should restrict it. The research on AI as a learning tool is genuinely mixed, and there are real use cases where AI helps students understand things faster, access better explanations, and practice skills more efficiently than traditional methods allow.

What the Gonsalves framework is pointing toward is that the two types of critical thinking are not automatically developed together. A student can become quite proficient at the first type, using AI thoughtfully and productively for their work, without ever developing the second. And the second is the one that matters most when the stakes are high.

This isn't so different from what we know about reading comprehension. Decoding words and understanding what those words mean are related skills that get taught separately because they don't automatically develop together. A child can become a fluent decoder without strong comprehension. Catching the gap early and addressing it directly is how it gets fixed.

The AI version of that gap is between using AI outputs and evaluating them. Schools are teaching the usage side. The evaluation side is largely being left to chance.

What the Second Type Actually Looks Like

Critical thinking toward the AI isn't a subject or a curriculum. It's a set of habits that, once formed, activate automatically.

At its most basic level, it looks like a child who doesn't accept an AI output at face value. Not because they distrust AI generally, but because they've learned that trusting anything without checking is a habit worth breaking. They ask: where is this claim coming from? Is this the kind of question AI is reliable on? How would I verify this if I needed to be sure?

At a more developed level, it looks like a child who knows the specific failure modes to watch for. AI is more reliable on topics that are widely documented and stable than on topics that are recent, obscure, or require precise citation. A student who knows this doesn't just ask "is this true?" They ask "is this the kind of question where AI is likely to be right?" That's a metacognitive distinction that most adults who use AI haven't fully internalized.

At the level Orbrya's curriculum is designed to build, it looks like a child who has internalized a verification reflex: the automatic move to check before accepting, regardless of how confident the AI's output sounds. This is the habit that transfers across subjects, contexts, and every new AI tool that arrives in the coming years.

Why Families Have an Advantage Here

Schools teach critical thinking for the assignment well. They've been doing it for decades. The frameworks exist, the assignments are designed for it, and teachers are trained to develop it.

Critical thinking toward the AI is harder to teach in a traditional classroom for a few reasons. It requires students to actually engage with AI outputs in real time, evaluate them against reliable sources, and develop intuitions about where the tool fails. That's harder to structure into a 45-minute class than a research essay or a Socratic discussion.

At home, with a single child or a few, the conditions are actually better. You can have the conversation in context, the moment the AI output appears on screen. You can ask the question that matters most not as an assignment but as a habit: how do you know that's true?

That's the seed of the second type of critical thinking. And the families who plant it now, before their child's AI habits are fully formed, are building something that the Gonsalves research suggests matters more than anything else being taught under the AI education label.

We'll dig into exactly what this looks like in practice next week, when we walk through the Three Questions Framework, the simplest tool we know for building the verification habit at any age.

If you want to be notified when the Three Questions Framework post goes live, and when we open access to the full AI literacy curriculum we're building for K-12 families (launching August 2026), join the waitlist.

Orbrya builds AI literacy curriculum for K-12 families, homeschool and supplemental, focused on the verification and evaluation skills that separate effective AI use from dependent AI use. Curriculum launches August 2026.

Sources cited in this post: