The Engaging Teacher | Instructional Design & Teacher Resources for Classroom Engagement

The Slow Work of Becoming Thoughtful

We are entering a strange moment in human history.

For the first time, the gap between looking intelligent and deeply understanding something is becoming almost invisible.

AI can now generate essays, presentations, emails, lesson plans, code, strategic recommendations, discussion posts, and emotionally convincing language in seconds. The polished artifact, once treated as reliable evidence of human thinking, is becoming increasingly unstable as proof of understanding.

And beneath all of the excitement surrounding artificial intelligence, I think we are encountering a much older question:

What does it actually mean to become thoughtful?

Not productive.
Not efficient.
Not optimized.

Thoughtful.

When the Work Stops Proving the Thinking

For years, many educational and professional systems quietly relied on a hidden assumption: if the final product looked sophisticated enough, meaningful thinking must have happened somewhere along the way.

But increasingly, that assumption no longer holds.

A polished artifact can now emerge from deep understanding, shallow prompting, borrowed language, collaborative cognition, partial comprehension, automated synthesis, or some unstable mixture of all of them. The work alone no longer reliably proves the thinking.

In many ways, AI did not create this problem.

It revealed it.

And I think that changes the role of learning entirely.

Because if outputs alone can no longer serve as trustworthy evidence of understanding, then the future of education may depend less on evaluating finished products and more on learning how to recognize the visible traces of thought itself.

Rethinking Higher-Order Thinking

For decades, frameworks like Bloom’s Taxonomy helped educators think about increasingly complex forms of cognition: remembering, understanding, applying, analyzing, evaluating, and creating.

But AI complicates those categories in important ways.

Many tasks that once appeared to demonstrate higher-order thinking can now be simulated fluently through generation alone. An essay may appear analytical. A discussion post may sound reflective. A presentation may look sophisticated. Creation itself is no longer reliable proof that meaningful understanding occurred beneath the surface.

And I think that forces us to confront an uncomfortable but important distinction:

Performance is not always the same thing as capability.

For years, educational systems often treated polished outputs as proxies for cognition. If learners could produce something that resembled analysis, evaluation, or creation, we frequently assumed deeper understanding had developed somewhere along the way.

But increasingly, AI exposes the instability of that assumption.

Because what matters is not simply whether a learner can generate an artifact associated with higher-order thinking. What matters is whether they can justify reasoning, transfer understanding across contexts, recognize weak arguments, compare alternatives, revise mental models, navigate uncertainty, and make thoughtful judgments when no obvious answer exists.

In other words, the future of learning may depend less on whether students can produce sophisticated work, and more on whether meaningful thinking remains visible within the process itself.

This does not make Bloom’s irrelevant. If anything, it may make cognitive complexity more important than ever.

But it does shift the burden of assessment.

The challenge is no longer simply assigning higher-order tasks. The challenge is designing environments where judgment, discernment, and visible reasoning remain interpretable beneath increasingly fluent outputs.

Because in the age of AI, creation alone may no longer prove understanding.

Thoughtfulness still does.

Human Capability Forms Before Automation Can Extend It

In learning science, durable understanding does not emerge simply from exposure to information. Cognitive structures strengthen through retrieval, elaboration, comparison, feedback, revision, and effortful processing over time. Robert Bjork’s work on desirable difficulties, Sweller’s cognitive load theory, transfer theory, metacognition, and developmental scaffolding models all point toward a similar truth:

Deep understanding is constructed slowly through structured encounters with complexity.

This becomes especially important in the age of AI.

Because when powerful tools enter the learning process too early, they can sometimes create the appearance of sophisticated cognition before foundational reasoning structures have fully stabilized internally. Performance rises. Fluency rises. Production accelerates. But the underlying architecture of judgment may still be fragile.

That distinction matters.

A learner may be able to generate polished answers long before they can independently diagnose weak reasoning, compare alternatives, justify revisions, transfer understanding across contexts, or recognize when something merely sounds convincing.

The external artifact appears stable while the internal cognitive structure remains underdeveloped.

This is part of what I describe as the Structural Stability Point (SSP): the threshold at which foundational reasoning structures become durable enough for AI to meaningfully extend cognition rather than inadvertently bypass its development.

Before that point, automation can sometimes reduce productive struggle, compress reflection, or interrupt the effortful processing that strengthens long-term understanding.

After that point, however, AI can become remarkably powerful as a cognitive partner:
accelerating synthesis, expanding perspective-taking, supporting systems thinking, increasing iteration speed, and helping learners engage more deeply with complexity itself.

The question, then, is not whether AI is “good” or “bad” for learning.

The question is developmental.

Fluency Is Not the Same Thing as Understanding

One of the risks of highly fluent systems is that they make shallow cognition feel complete.

When something sounds polished, coherent, immediate, and confident, humans naturally infer intelligence. But cognitive science has repeatedly shown that recognition is not the same thing as recall, exposure is not the same thing as integration, and performance is not always the same thing as durable understanding.

This is why the work has stopped proving the thinking.

The artifact alone no longer tells the full story.

Which means the future of assessment may depend less on isolated outputs and more on trustworthy signals of judgment gathered across time, context, revision, explanation, transfer, and reflection.

Because thought leaves traces.

The Conditions That Make Thinking Happen Visibly

Human cognition does not develop independently from environment.

People think differently when they feel psychologically safe enough to risk uncertainty. They reason differently when curiosity is rewarded instead of punished. They engage differently when learning is organized around meaning rather than performance alone.

The architecture of a system shapes the architecture of attention within it.

This is part of the thinking behind frameworks like Better Signals and the SIGNAL Lens: approaches centered on visible thinking, trustworthy evidence of understanding, and the conditions that allow human judgment to emerge authentically in AI-mediated environments.

Increasingly, I worry that many modern systems are optimizing for completion rather than contemplation.

For immediate output rather than durable understanding.

For fluency rather than discernment.

Yet some of the most meaningful forms of intelligence still resist acceleration.

Wisdom cannot simply be downloaded.
Ethical reasoning cannot fully emerge from predictive text.
Deep understanding still requires integration across memory, emotion, reflection, experience, and time.

The slow work matters.

What Learning Must Protect Now

I do not believe the answer is rejecting AI.

These tools can support creativity, reduce unnecessary barriers, accelerate iteration, and help humans navigate increasingly complex information ecosystems in remarkable ways.

But tools that accelerate production also increase the importance of discernment.

The easier it becomes to generate answers, the more valuable it becomes to recognize what is worth trusting, questioning, revising, or resisting.

The future may not belong simply to people who can use AI effectively.

It may belong to people who have built thinking structures strong enough to remain thoughtful within it.

Because the future of intelligence may depend less on what we can generate, and more on whether we still know how to think together.

The Work Beneath the Work

Some of the thinking behind this piece has been shaped by research surrounding:

• cognitive load theory
• desirable difficulties and retrieval practice
• transfer and durable learning
• metacognition and visible thinking
• developmental scaffolding
• human-AI collaborative cognition
• and trustworthy signals of judgment in AI-mediated environments

I continue exploring these ideas through frameworks like Better Signals, Structural Stability Point (SSP), and the SIGNAL Lens.

Further Exploring

Missed our newsletter? Check it out to find free resources tailored to improving your teaching an lesson design.

Subscribers also receive:

Better Signals: A Starter Kit for Teaching in the Age of AI.