No Disruption, No Learning
By Wesley Hall Parker
New growth comes from disruption.
Have you noticed how ChatGPT, Claude, and other LLMs flatter you?
LLMs use affective conversation design strategies to increase emotional engagement and user retention.
LLMs are designed to hook you emotionally and keep you chatting longer. They use strategies like reflection and validation to make you feel understood and seen. They mirror your rhetorical complexity, imitate your vocabulary and communication style, and build on your demonstrated knowledge and interests. Some will even flatter you and call you cute nicknames.
And it’s working. A recent Harvard Business Review study shows that Therapy/Companionship is now the #1 use of GenAI—and Enhanced Learning has moved up to spot #4.
So, the BIG QUESTION is, is emotional validation GOOD for learning?
Yes! But…
Here’s the hidden issue: A constant stream of understanding and validation won’t create effective learning on its own. Building skills requires repetitive practice—cognitive work—to build new neural pathways. And cognitive work requires productive challenges, conflicting perspectives, and disruptive truths that shift outdated mindsets.
LLMs decrease friction in the name of user retention. They adapt to us so completely that they don’t show us the unfiltered truth. It’s more like chatting with a Yes Person or gazing into a flattering mirror. There’s almost zero push-back. Some AI researchers are beginning to call this the “sycophantic feedback loop.”
LLMs are designed to keep you in your comfort zone. And as all successful leaders know, if you’re too comfortable, you’re not growing.
“Most adult learning… involves self-directed focus—a ‘friction’ of sorts… Friction is the gateway to self-directed plasticity. It reduces the number of trials to learn. No friction, no learning…”
Which “creative disruption” strategies improve learning outcomes?
-
The Socratic method builds learning through questions that challenge assumptions and expose blind spots. Rather than affirming what learners already believe, it invites them to wrestle with uncertainty, contradiction, and evolving reasoning.
But LLMs are trained on affective mirroring. They’re designed to validate, not interrogate. Left to their defaults, they’re more like agreeable companions than critical thinkers.
To shift this, when appropriate to the learning context, we need to explicitly prompt LLMs to play the role of thought partner rather than cheerleader. We call this type of conversational AI agent a “Semi-Socratic Tutor.” We train them to push back gently. To question the learner’s logic. And to hold off on providing immediate easy answers if the learner can construct the knowledge on their own.
Roger Schank, pioneer of learning-by-doing and a fierce critic of passive education, argued that true learning happens when our expectations fail—when things don’t go as planned, and we’re forced to rethink what we know. That dissonance is the learning moment.
“We learn not when we're told, but when we try, fail, and discover why.”
—Roger SchankWhen used consciously, LLMs can be trained to simulate that kind of friction: a sparring partner that helps us think better—not just feel seen.
-
Humans are wired for narrative. But the most powerful learning stories aren't the ones where everything goes right. They're the ones where everything goes wrong.
In our work, we often invite internal experts to share “war stories,” real moments from their careers when assumptions failed, plans unraveled, and painful lessons were learned the hard way.
These stories do more than entertain. They disrupt ingrained beliefs, surface unconscious bias, and create cognitive dissonance, the key ingredient for mindset change.
“Wait, we did everything right…and it still failed?”
“I’ve made that exact decision. What did I miss?”
By intentionally designing stories to create constructive discomfort, we prime learners to be more receptive to change. When a peer tells a story that hits close to home, it opens the learner up in ways data never could.
This is especially true in complex systems, where root causes are cloudy and outcomes can’t be reverse engineered. Stories make failure psychologically safe to examine, and emotionally resonant enough to remember.
-
We don’t just need stories or questions. We need architectures that build friction into the bones of the learning experience. One of our most effective frameworks is what we call the “Anatomy of a Train Wreck.”
In this design, learners work through a real-life project that went off the rails, often drawn from the client’s own internal case archive. The case is delivered in phases. After each chapter, learners:
Identify risks and red flags
Recommend mitigation strategies
Predict what will happen next
Then the next phase is revealed. And they find out:
Were their predictions accurate?
Were their strategies effective?
What did they overlook?
Classroom debriefs punctuate each cycle, giving learners a chance to compare insights, cross-pollinate ideas, and reflect on their own biases. Facilitators guide the process—not to provide answers, but to ask better questions.
By the end, participants have co-created a toolkit of field-tested strategies for avoiding the most common and expensive project failures.
This architecture embodies what Schank championed: learning through failure, not instruction. It builds muscle memory for pattern recognition, anticipatory thinking, and adaptive problem-solving all grounded in lived, contextual learning.
