Speaker at podium before audience and satirical signs reading “Critical Thinking Is for Poor People” and “Make America Dumb Again but Classy.”

The iPhone emerged in 2007. I started teaching in 2004. That gives me something most researchers lack, a before. My hypothesis is that we have already run this experiment once, so if I am right then we don’t have to guess how the AI version goes.

Here’s where I’m coming from. Before teaching, I was a graduate student at LSU (2002-2004), researching metacognition as it relates to mathematics education. Then I taught high school mathematics and AP Statistics for fourteen years, until 2018. Over my tenure, I observed time and time again that the real measure of understanding isn’t a flawless calculation. It’s the courage to question the data when it doesn’t make sense.

Over the past 24 years, both inside the classroom and observing from the outside since, I’ve watched smartphones and other technologies erode that critical thinking, turning learning spaces from places of discovery into battlegrounds for fractured attention. Slowly, we’ve traded productive struggle for frictionless convenience, raising a generation adept at retrieving answers but increasingly blind to the need for critical thought.

Now, with AI in our pockets, we face the ultimate pivot: do we let these tools do the thinking for us, letting our neural circuits atrophy, or do we use them as a scaffold to deepen our own inquiry and preserve what makes us human?

The Shift in Student Cognition (The 24-Year Trajectory)

When I began teaching, the classroom wasn’t necessarily a sanctuary for sustained focus, but the competition for attention was reserved for the typical distractions (e.g. passing notes, double entendre jokes, overall life concerns, zoning out, etc). Over the next fourteen years, and observing from the outside since, I watched that attention erode further. It wasn’t just about devices; it was the constant cognitive load of living in a state of perpetual partial attention.

Research bears this out. When students divide or switch their attention between learning and a mobile device, the cognitive cost is steep. A 2025 meta-analysis of 27 randomized controlled experiments (n = 2,245) found a substantial, statistically robust drop in immediate recall (Hedges’ g = -0.65), with an even steeper penalty during live lectures (g = -0.70) (Chen et al., 2025).

And then there’s the Cognitive Load Theory, which demonstrates that our working memory can only juggle a few pieces of new information at once (Sweller, 1988/2010). When we flood that narrow bottleneck with constant distractions alongside information at our fingertips, we create an environment where deep encoding is nearly impossible.

We are trading “desirable difficulties,” the productive struggle essential for mastery, for frictionless convenience (Bjork & Bjork, 2011/2020). Worst of all, standardized tests often measure surface-level retrieval strength rather than actual storage strength, masking the reality that we are optimizing for short-term test performance at the expense of conceptual understanding and retention.

As smartphones continue to compete with our attention, can AI take the thinking itself?

The Crux

In 2007, I didn’t know I was standing in a “before.” I only saw it in hindsight, after the change had settled into the classroom. With generative AI, we are standing in the before right now, which is why I think this conversation matters NOW.

For simplification, I’m narrowing AI here to generative AI (think ChatGPT, Gemini, or Claude): you have a question so you ask the tool a question and it spits out an answer. What happens next is what I’m calling the “crux” of AI technology, and it has two parts: how much you thought about the problem on your own before you asked, and what you do with the output once you have it. Are you going to struggle through the problem first without assistance? Are you going to examine the output thoughtfully, question back, expand your ideas, and eventually learn something from the exchange? Or are you going to bypass your own thought and run with the generated response?

The danger isn’t just that tools provide wrong answers, but that they offer right ones too easily. This triggers cognitive offloading: when we outsource the struggle of synthesis, we bypass the productive friction necessary for deep thinking (Risko & Gilbert, 2016 and Abrams, 2026). Recent studies specifically examining generative AI show a significant negative correlation between frequent, uncritical AI reliance and critical thinking scores (especially among younger users) — a pattern authors attribute to cognitive offloading (Gerlich, 2025).

So here’s how I think about it:

The Soul of Human Learning: Curiosity and Questioning

Automated data synthesis is merely statistical pattern-matching. What separates human cognition from artificial models is our capacity for meaning-making, driven by curiosity, experience, nuance, and intuition. At the heart of this is the art of the question. Widely recognized in creativity research (Guilford, 1967 and Clark et al., 2019), the act of formulating deep, evolving questions requires a delicate balance of divergent creativity and convergent analysis. Decades of research on the ‘generation effect’ show that minds retain and synthesize far more when actively producing inquiries rather than passively consuming answers (Slamecka & Graf, 1978). When we shift from passive answer-recipients to active question-generators, we build intellectual autonomy — the ability to think for oneself, analyze evidence, adjust your hypothesis, and form your own reasoned conclusions. Uncritical reliance on AI threatens to close our neural circuits right when we need discernment most.

A Blueprint for Modern Learning (Students & Adults)

To navigate an automated landscape, my challenge to society is to shift our educational focus from information retrieval to active synthesis. This requires cultivating something called metacognition. Metacognition is what I define as “thinking about my own thinking,” also known as, “the deliberate habit of auditing my own thinking.” This practice is consistently ranked among the highest impact-drivers of durable learning (Flavell, 1979). For both students and adults, the goal is to use AI as an intellectual scaffold — a temporary support designed to stretch capability without replacing it (Wood, et al., 1976) while preserving analytical independence. When we use technology to challenge our thinking rather than bypass it, we transform an automation threat into a cognitive amplifier keeping our minds sharp, adaptable, and unmistakably human.

Here is how I think about it:

  1. Struggle first. Spend some time trying to figure out the solution to the problem (or the answer to the question) on your own.
  2. Ask with your attempt in hand. Don’t prompt for the answer, ask, “here’s my reasoning, where is it weak?” That turns the tool into your Socratic partner.
  3. Question back. Check claims against sources and ask what would make the output wrong. This is not only your sanity check, it’s your own “courage to question the data.”
  4. Own it. Write what you now understand in your own words and what changed from your first attempt. This is the metacognition step, auditing your thinking.

Leave a Reply