In 1943, the American military had a problem with its bombers. Too many were being shot down. The obvious fix was more armour, but armour is heavy, so it had to go only where it was needed most.
Engineers studied the planes returning from raids and mapped where they were riddled with bullet holes: largely along the wings, the fuselage, and the tail. That, clearly, was where they needed to add the armour.
However, a statistician named Abraham Wald stopped them. The planes they were studying, he pointed out, were the ones that made it back. The bullet holes showed where a bomber could be hit and survive. Armour the clean spots, he said.
This is survivorship bias, and it’s one of the many ways that intelligent, well-informed people reach confident, evidence-based conclusions that also happen to be wrong.
In learning and development, we spend a great deal of energy diagnosing the biases that stop learners learning. However, we are far less practised at spotting the ones that distort our own decisions, about what to build, what to buy, and what we choose to believe is working.
That blind spot is expensive. When an L&D team’s bias misfires, it wastes precious budget, ships the mistake to thousands of people, and calls it a success. In this article we’ll be turning the mirror around. It’s time for some serious reflection.
Why is L&D Blind to Its Own Biases?
There’s a good reason L&D is quick to see bias in learners and slow to see it in itself. It’s called the bias blind spot, and it’s one of the most reliably documented findings in the field.
In a 2002 study, Emily Pronin and colleagues found that people readily recognise cognitive biases in others while often failing to spot the same bias in themselves. Across their surveys, people consistently rated themselves as less biased than the average person.
In one sample of over 600 people, more than 85% believed they were less biased than average. Exactly one person thought they were more biased than average. And when the effect was explained to them, most simply insisted their own self-assessment was the accurate one.
You might expect training and intelligence to help. They don’t. In a follow-up study, Richard West and Keith Stanovich tested whether cognitive sophistication reduced the bias blind spot. It didn’t. If anything, the more cognitively able people showed a slightly larger blind spot.
This is the uncomfortable premise of everything that follows. L&D is a profession of thoughtful, evidence-minded, well-read people, but none of that provides us with additional cognitive armour. The biases listed below are not a list of people’s failings.
They’re recognition that capable professionals can act in good faith and still reach confident and costly conclusions.
What Are The Main Cognitive Biases in L&D?
Before the deep dives, here is the full set. Each is a well-documented bias. We’ve also indicated where these biases tend to show up in the decisions learning teams make.
| Bias | What It Is | How It Shows Up in L&D |
|---|---|---|
| Confirmation bias | Seeking and favouring evidence that fits what you already believe | Choosing the metrics and feedback that make a programme look successful, and quietly discounting the ones that don’t |
| Survivorship bias | Judging by the successes you can see and ignoring the failures you can’t | “Our top performers all did this programme”, without asking about everyone it didn’t help |
| Sunk cost bias | Sticking with something because of what you’ve already put into it | Keeping a costly legacy programme or platform alive because scrapping it would waste the investment |
| Curse of knowledge | Being unable to imagine not knowing what you already know | Experts and designers building content that assumes knowledge the learner simply doesn’t have yet |
| Authority bias | Over-weighting the view of a prominent or credentialed source | Adopting a model or method because a big name or popular framework endorses it, not because the evidence does |
| Novelty bias | Overvaluing something simply because it’s new | Chasing VR, AI, or the latest trend without evidence that it actually improves learning |
These six biases share a single root. Because we experience our own decisions as objective and evidence-based, the bias feels like sound judgement from the inside. This is what makes them so hard to catch in the moment.
The rest of this article looks in depth at the most damaging of them, and then at how to design them out of your learning programmes.
Confirmation Bias: Measurement That Flatters
Of all the biases that distort L&D decisions, confirmation bias is the most pervasive, because it hides inside the very thing we use to prove we’re being objective: measurement.
Confirmation bias, as Raymond Nickerson set out in his definitive review, is the tendency to seek and interpret evidence in ways that favour what we already believe. And in L&D, what we already believe is usually that the programme we built, chose, and championed is working.
You’ll have seen this play out before. A team launches a new initiative that they’re proud of. When the results come in, they reach, unconsciously, for the numbers that make it look good. Typically this means:
- Enrolment numbers: How many people signed up or turned up.
- Completion rates: How many people finished the course.
- Satisfaction scores: How much learners enjoyed it.
Naturally, the one metric that didn’t move, actual behaviour on the job, is quietly reframed as “hard to measure” or “too early to tell”. They’re not lying. But a filter is being applied, giving the flattering evidence a pass and holding the unflattering evidence to impossible standards.
Smile Sheets: The Perfect Confirmation Bias Instrument
The smile sheet is where this bias lives most comfortably. Post-course happiness scores are easy to collect, reliably positive, and almost entirely disconnected from whether anyone learned or changed.
They tell you what you want to hear, and they feel like evidence. Choosing to measure learner reaction over learner behaviour isn’t a neutral methodological choice. It’s often confirmation bias selecting the kindest possible test.
It’s also why Donald Kirkpatrick’s evaluation framework has four separate levels.
If your programme’s wins are proven with headline numbers while its failures are always explained away as circumstance, timing, or measurement difficulty, you’re not evaluating the programme. You’re defending it.
Genuine evaluation should always feel slightly uncomfortable. After all, you’re looking for evidence that the thing you’re proud of didn’t work.
Survivorship Bias: Learning From The Wrong Success Stories
Let’s return to Wald’s bombers, because the mistake he caught is one that L&D makes constantly.
The military studied the planes that came back and drew conclusions about all planes, forgetting that the ones with fatal damage weren’t in the room to be studied. This is survivorship bias: judging solely by the success you can see, while failures stay silent and invisible.
L&D is full of similar success stories:
- “Our top performers all went through this leadership programme.”
- “Everyone who completed the LMS pathway got promoted.”
- “Our best managers swear by it.”
Each is offered as proof the programme works. However, you’re only looking at the people the programme seemed to help: the ones still visible, still with the company, still succeeding.
The people it failed are missing planes. They’re the ones who took the same programme and left, stalled, or just got nothing out of it. They’re no longer in your data because they’re no longer around to be counted.
The correction we need to apply is Wald’s question, turned on our own programmes. Not “what do our successes have in common?” but “what happened to everyone who didn’t succeed, and did they really do anything different?”
Sunk Cost Bias: Clinging to Programmes That Don’t Work
Every L&D team has one: the programme that everyone privately knows isn’t working, that somehow never gets cut. The flagship academy that took two years to build. The authoring tool the department standardised on. The blended curriculum with someone else’s name attached to it.
The evidence that it isn’t delivering keeps arriving, and it still survives, because too much has gone into it to stop now.
This is the sunk cost bias in action, and it’s one of the most robust findings in decision science. In their defining 1985 study, Hal Arkes and Catherine Blumer showed that people persist with a course of action in proportion to what they have already invested in it, even when a clear-eyed look at the future says stop.
The money, time, and effort already spent are gone either way, and rationally should have no bearing on what to do next. But they weigh on us anyway, because abandoning the project means admitting the investment was wasted.
Two of their findings are especially relevant to L&D:
- We Persist So We Don’t Appear Wasteful: Arkes and Blumer traced the effect to the desire not to look as though we’ve squandered resources, which is an important consideration for a team that has championed a major programme to the wider business. Killing it feels like confessing to a costly mistake.
- Investment Inflates Our Judgement: More uncomfortably, they found that people who had sunk costs into a project raised their own estimate of how likely it was to succeed. The more you have poured into a failing programme, the more you convince yourself it’s about to turn the corner.
The bias, in other words, doesn’t just keep you committed. It quietly rewrites your judgement to justify staying.
Novelty Bias: Chasing The Next Shiny Thing
L&D has a weakness for the new. Every couple of years a technology arrives: eLearning, mobile learning, gamification, VR, and now generative AI. With it comes a rush to adopt before anyone has established whether it actually improves learning.
This is novelty bias: overvaluing something because it is new, and treating “innovative” and “effective” as synonymous.
The tell is a business case built on excitement rather than evidence. A platform gets bought because it demos impressively, because competitors are using it, or because “we need an AI strategy”, not because there’s a good reason to believe it will help people to learn better.
VR is an instructive case. Studies repeatedly find a novelty effect: learners are captivated by the new medium at first, and that initial fascination can actually impede learning while they attend to the technology rather than the content.
Worse, immersive environments are prone to a familiar trap. The engagement they generate gets mistaken for learning, with learners rating their own knowledge gains higher than tests bear out.
That’s the real danger of novelty bias. It puts the technology first and the learning second, when the evidence consistently says the reverse. The delivery format matters far less than whether the design respects how people actually learn.
Why Do Cognitive Biases in L&D Matter?
A learner’s bias costs them an afternoon (providing they manage to catch it in the first place). An L&D team’s bias costs the organisation, and at scale. A single biased decision, to keep a failing programme, buy the wrong platform, or trust a flattering metric, has serious consequences.
What’s more, the biases in this article don’t operate in isolation. Instead, they reinforce one another into a closed loop.
- Confirmation bias selects the metrics that hide a programme’s failure.
- Survivorship bias supplies the success stories that seem to prove it works.
- Sunk cost keeps it running long after the doubt sets in.
- And novelty bias points the next budget at the next shiny thing.
Each bias makes the others harder to catch, and together they can keep a whole L&D strategy aimed in the wrong direction indefinitely.
Worse still, these biases help learning myths to survive. Discredited ideas persist in L&D not because the evidence is unclear, but because biased decision-making protects them. Learning styles is the textbook case for this.
The idea that matching teaching to a learner’s preferred “style” improves outcomes has never held up in the controlled tests designed to confirm it. Yet, it endures, because it feels intuitively right (confirmation bias), because well-known figures endorse it (authority bias), and because teams who have built training around it are reluctant to change course (sunk cost).
Put together, this is why the stakes are higher on our side of the screen. When L&D fools itself, it doesn’t just waste money. It entrenches the very practices the evidence tells us to drop.
How to Reduce Cognitive Bias in Learning Decisions
You cannot delete these biases. The research on the bias blind spot is clear that awareness alone doesn’t fix them. But you can build decision-making processes that catch them, by adding structure that forces the questions bias usually makes us skip.
Keep the following four practices in mind.
- Look For Disconfirming Evidence: Confirmation bias is defeated by deliberately hunting for the case against. Before declaring a programme a success, ask what evidence would show it had failed, then go and look for it. Make “what would change our minds?” a required question.
- Measure Behaviour, Not Reactions: Most of the biases feed on soft metrics that flatter. The antidote is to measure what people actually do differently on the job, not whether they enjoyed the course or passed a same-day quiz. It’s harder, but it’s also more honest.
- Pilot and Compare: Survivorship bias thrives when you only study the people a programme reached. Wherever you can, compare against a group who didn’t take it. A small pilot with a comparison group tells you far more than a room full of success stories.
- Run a Pre-Mortem: Before committing to a programme or platform, use Gary Klein’s pre-mortem: imagine it’s a year later and the initiative has failed completely, then have the team write down why. This approach helps you to identify real issues before they arise.
That last practice is particularly effective. Imagining that an event has already happened, rather than that it might, increases people’s ability to identify the reasons for it by around 30%.
Underneath all these approaches is one key shift. It’s the need to move from decisions that feel right to decisions that are built to be tested. Treating your judgement as something to check rather than trust is the best way to overcome our bias blind spots.
Final Words
The uncomfortable thread running through this article is that good intentions and professional expertise are no defence against bias. Thoughtful, experienced L&D teams reach confident, costly, wrong conclusions, not through carelessness, but because that is how human judgement works.
This is why neurogogy applies as much to the people designing learning as to the people receiving it. Designing learning around how the brain actually works means accepting that our own brains, the ones making the calls about what to build, buy, and believe, run on the same flawed hardware as everyone else’s.
Building learning that works means designing not just around your learners’ biases, but around your own. So stop armouring the bullet holes and spend your budget on what actually changes behaviour.
Ready to build learning that lasts?
- Book a demo to see how the Impact Suite turns evidence-led learning design into measurable results.
- Explore the Impact Suite and the neuroscience behind every feature.
- Download the Learning Theories guidebook for the full picture of the science that makes learning stick.
