Student Reflections on AI Use Doesn’t Work

Here’s what to do instead

The Second Draft: #0070

I write two articles each week on the emerging discipline of Human + AI Work—tracing how education, expertise, and human capability evolve as AI reshapes work, knowledge, and meaning.

---

The Reflection Trap

Everyone is talking about reflection, transparency, and metacognition in the AI era.

And they’re partially right.

But here’s the problem:

the way we’re doing reflection doesn’t work. at all.

---

The Self-Referential Paradox

Here’s the fundamental trap:

Reflection asks you to think about your thinking
—using the same thinking that reflection is trying to upgrade!

In other words, you’re stuck in a closed loop.

Your reflection is limited by the very cognitive patterns you’re trying to examine.

This is why reflections always feel performative and surface-level.

Students write things like:

• “I should have started earlier”

• “I need to manage my time better”

• “I learned the importance of preparation”

These aren’t insights.
They’re placeholders.
Best guesses.

Because the student is being asked to transcend their own thinking without any external input or new perspective.

They’re being asked to bootstrap themselves into awareness using the same operating system that didn’t produce awareness in the first place!

It’s cognitively impossible and it’s why reflection adds nothing new.

You can’t think your way out of how you think.

You need different questions that break the loop.

---

Enter Solution-Focused Brief Therapy (SFBT)?

Solution-Focused Brief Therapy is a therapeutic approach developed in the 1980s by Steve de Shazer and Insoo Kim Berg.

Unlike traditional therapy models that focus on analyzing problems and their causes, SFBT focuses exclusively on solutions and what’s already working.

The core insight: You don’t need to understand a problem to solve it.

Instead of asking “Why is this happening?” SFBT asks:

• “When is this not happening?”

• “What’s different when things are better?”

• “What would one small step forward look like?”

This makes SFBT incredibly practical for learning contexts, because students don’t need to become amateur psychologists analyzing their own cognition.

They just need to notice what’s working and do more of it.

SFBT is action-oriented, strength-based, and future-focused.

Which makes it the perfect framework for learning in the AI-era.

---

Why SFBT Works When Reflection Doesn’t

SFBT works because it doesn’t ask you to analyze what you did.

It asks you to notice what’s already different when things work better, even slightly, even partially.

These are the moments when you somehow broke free from your usual pattern.

You don’t have to understand why it worked.
You just have to notice that it did.

SFBT questions bypass the self-limiting nature of reflection by:

• Focusing on exceptions, not explanations: “When has this been even slightly better?” You don’t need to analyze—just observe.

• Surfacing unconscious competence: You’re already solving parts of this problem, but you don’t realize it. SFBT makes it visible.

• Building from what works, not what doesn’t: Instead of being trapped by what you did wrong, you identify what you did right—and do more of it.

• Creating external perspective: Questions like “What would your friend notice about what’s working?” give you the outside view you can’t get from reflection alone.

In other words: SFBT doesn’t ask you to transcend your limitations through introspection.

It asks you to notice the moments when you already transcended them,
and then replicate those conditions.

That’s why it works.
That’s why reflection doesn’t.

---

The Core Assumptions of SFBT

SFBT operates from a completely different set of assumptions than traditional therapy on several key assumptions.

1. The problem is the problem

Not the person. Not their intelligence. Not their “growth mindset.” The problem is external—it’s a situation to solve, not an identity to fix.

2. The problem is not always the problem

Meaning: You don’t always experience the problem. There are times when it’s better, when it doesn’t show up, when you navigate it successfully. These exceptions are gold—they reveal your existing solutions.

3. You already have solutions

You’ve already solved versions of this problem before, even if you don’t realize it. The goal isn’t to discover some new magic—it’s to make visible the competence you already possess and use it more deliberately.

4. Small changes lead to bigger changes

You don’t need to overhaul everything. Identify one thing that’s working and do more of it. Identify one small next step and take it. Momentum compounds.

5. The future is created, not predicted

Instead of analyzing what went wrong, we ask: What do you want to be different? What would that look like? What’s already moving you in that direction?

6. Clients are the experts in their own lives

Students know more about their own learning context than we do. Our job isn’t to diagnose and prescribe—it’s to help them surface and amplify what already works for them.

---

Reframing “Problem” for Learning Contexts

SFBT focuses on solutions, but it is rooted in problems.

In SFBT “problem” has a specific therapeutic meaning,
it’s the issue that brings someone to seek help.

So in learning contexts, we need to be more precise about what we mean by “problem.”

The “problem” isn’t necessarily what went wrong.

It’s the gap between where the student is and where they want to be.
It’s the learning they’re trying to consolidate.
It’s the challenge they’re navigating.

We use reflection and transparency prompts because we believe growth comes from metacognition.

We want students to “think about their thinking” because we assume that awareness = improvement.

But here’s where the paradox returns:

If the “problem” is that a student can’t see their own thinking clearly, asking them to reflect on their thinking uses the same unclear thinking they’re trying to improve.

So in our context, when we talk about “the problem,” we mean:

• The pattern the student wants to shift (not a deficit to fix)

• The learning challenge they’re navigating (not a failure to analyze)

• The gap between current and desired competence (not something broken inside them)

We don’t need students to explain what they did.
We need them to notice that they do it and then do more of it.

---

Why This Matters for AI + Learning

Students are producing work effortlessly with AI.

So we ask them to reflect on it.
To be transparent about their process.

But, as discussed, those learning moments are really just performance,
students guessing on what the instructor wants to hear.

Reflection and transparency prompts ask students to report what they did. But reporting is not the same as unlocking the mental models they’re building.

SFBT helps students surface the actual cognition happening as they work— the mental frames, the heuristics, the decision points, the strategies they’re using (often unconsciously) as they engage with (or without) AI.

When students use AI to produce work, they’re navigating complex decisions:

• How to evaluate AI output

• What to prompt for and when

• When to accept, reject, or modify suggestions

• When they’re done vs. when they need to keep iterating

• How to integrate AI-generated content with their own thinking

These are the heuristics and mental frames we want students to develop and we want to make them aware of through our reflection and transparency ideas.

But if we ask them to “reflect on their process,” we get surface-level reports:
“I used AI to help me brainstorm ideas.”

If we use SFBT questions, we help students surface real insight.

That’s what this series is about.

---

What’s Next

In the next article, we’ll unpack the 10 core SFBT question types. You could ask GPT in the meantime, but I’ll provide some more context for how they fit into our assessment and feedback practices in the AI-era.

Stay tuned . . .