Curse of Knowledge in Training: Why Experts Can't Teach

Your sharpest engineer just explained the deployment process for the third time, and the new hire still got it wrong. The problem is not the new hire, and it is not the engineer's knowledge. It is the curse of knowledge in training: a Harvard experiment found that learners taught by a smooth, expert-led lecture scored about half a standard deviation lower on a test of what they learned than peers who had to work through problems themselves, yet they felt they had learned more. In this post you will see why your best experts are often your most confusing trainers, what the research says actually works, and a six-step fix you can apply to your next onboarding or upskilling program.

What is the curse of knowledge in training?

The curse of knowledge is a cognitive bias: once you know something well, you cannot accurately imagine what it is like not to know it. Stanford's Chip Heath and Dan Heath popularized the term in a Harvard Business Review article, where they showed how executives compress years of context into abstractions like "unlocking shareholder value" and then wonder why frontline teams hear only opaque phrases.

Training has the same failure mode. Your expert says "then you just reconcile the accounts" and genuinely cannot see that "just" is hiding eleven separate decisions they now make on autopilot.

This is why handing training over to your most experienced person so often disappoints. The knowledge is real. The transmission is broken.

Why subject matter experts struggle to teach

Experts struggle to teach because expertise actively hides the beginner's view: the steps they skip, the vocabulary they no longer notice, and the difficulty they no longer feel. Researchers call the teaching-specific version of this the expert blind spot, and it grows with mastery. The better you get, the more of your skill becomes invisible to you.

There is a second, less obvious reason: teaching is its own body of knowledge, not a bonus feature of subject expertise. In a recent essay on EduResearch Matters, Griffith University learning scientist Sakinah Alhadad argues that even knowing the cognitive science of how people learn does not add up to the full professional knowledge of teaching. As she puts it, "neither the science of learning nor the learning sciences constitutes the whole of education."

That debate is playing out in school policy, but the corporate translation is blunt. If knowing learning science is not enough to make someone an effective teacher, knowing Kubernetes or revenue accounting definitely is not.

So when a company assigns training to whoever knows the topic best, it is quietly assuming a skill that was never there:

• The SME knows what is true, but not which parts novices predictably get wrong.
• The SME experiences the material as one fluid whole, while the novice needs it in ordered, digestible steps.
• The SME reads silence and nodding as understanding, because the material feels self-evident to them.

Why polished expert lectures feel effective but are not

Polished lectures feel effective because fluency is persuasive, not because learning is happening. A randomized crossover experiment at Harvard published in PNAS tested this directly: the same physics students experienced one class taught as a fluent, expert lecture and another where they attempted problems in small groups before seeing solutions. Content, handouts, and slides were identical.

The results should worry anyone who picks trainers by presentation skill:

• Students in the active sessions scored 0.46 standard deviations higher on the follow-up test.
• The same students rated their feeling of learning 0.56 standard deviations lower than the lecture groups.
• Perceived learning tracked the instructor's fluency, not the teaching method that actually produced learning.

"Actual learning and feeling of learning were strongly anticorrelated," lead author Louis Deslauriers told the Harvard Gazette. Physicist Eric Mazur, who was not involved in the study, said it "unambiguously debunks the illusion of learning from lectures."

Here is the uncomfortable part for L&D teams: your post-session satisfaction scores are measuring the illusion. A charismatic expert walking through flawless slides generates great ratings and weak retention, while the session that made people struggle a little gets grumbling feedback and better on-the-job performance.

What works better: practice over polish

The evidence for active learning over expert presentation is about as settled as findings get in education research. A meta-analysis of 225 studies in PNAS found that active learning raised exam performance by roughly 6 percentage points, and that students in traditional lecture courses were 1.5 times more likely to fail: failure rates averaged 33.8% under lecturing versus 21.8% with active approaches.

None of this means experts are the problem. It means the format most experts default to, the confident walkthrough, is the problem. The expert's knowledge only transfers when the learner is forced to retrieve, apply, and get corrected.

How to beat the curse of knowledge in your training

You beat the curse of knowledge by changing the expert's role, not by replacing the expert. Six moves make the difference:

1. Pair every SME with a structure, not a stage. The SME supplies the knowledge and the war stories. Someone or something else supplies the pedagogy: sequencing, questioning, and checks. Never make one person do both jobs unaided.
2. Watch a real novice attempt the task. Before building anything, have your expert silently observe a beginner try the process and write down every stumble. Nothing punctures the blind spot faster than seeing which "obvious" step was not obvious.
3. Flip the order: attempt first, explain second. The Harvard experiment's winning format had learners try problems in small groups before the explanation. In corporate terms: give the scenario before the walkthrough, and let people commit to an answer before you reveal the right one.
4. Replace presenting with questioning. A question surfaces the gap the expert cannot see. This is where interactive training videos earn their keep: an AI tutor built on your expert's content asks learners questions throughout, adapts to wrong answers in real time, and never assumes a step is too obvious to check.
5. Tell learners the struggle is the point. In the Harvard study, a short briefing on why effortful learning feels worse but works better changed attitudes: about 65% of students said their view of active learning improved afterward. Two minutes of framing protects your redesigned training from bad first reactions.
6. Measure learning, not applause. Swap satisfaction surveys for a delayed check: can people do the task a week later? If your highest-rated session produces the most follow-up questions on the job, the curse of knowledge is running your program.

FAQ

How do I know if the curse of knowledge is hurting my training?

Look for a gap between ratings and results. Sessions score well, but managers report the same questions and mistakes weeks later, and experts say things like "I already covered that." If your training is mostly experts presenting while learners watch, assume the curse is active, because the research shows neither the expert nor the learner can feel the failure happening.

Should subject matter experts stop delivering training?

No. SMEs are the only source of the judgment, edge cases, and tacit tricks that make training worth taking. The fix is to change their role: capture their knowledge, then deliver it through formats that force learner practice and questioning, with the expert focused on the hard parts a novice attempt reveals.

Does active learning work for corporate training or just classrooms?

The strongest controlled evidence comes from university courses, but the mechanism is not academic: retrieval, application, and feedback beat passive watching wherever skills must transfer to real work. Corporate learners are, if anything, more time-pressed, which makes practice-dense formats more valuable per minute, not less.

The lesson is not that your experts know too much. It is that knowledge does not teach itself, and the more someone knows, the more their explanation skips. Turn expert walkthroughs into interactive experiences that make learners attempt, answer, and get corrected, and the curse of knowledge becomes a design problem you have already solved.