Information Doesn’t Change Behaviour. Here’s What Does.

Introduction

Here is a pattern you will recognise.

Your team spends a day out of the office. The facilitator is good. The content is sensible, the room is engaged, people say afterwards that it was genuinely useful. Somebody writes up the actions.

Six weeks later, nothing is different.

Not “less different than hoped”. Everything is actually just the same. The same meeting runs the same way. The same conversation still isn’t being had. If you went back through the notes you would find things everyone agreed with and nobody does.

Most organisations respond to this by looking for the fault in the delivery. A better facilitator. Shorter sessions. More follow-up. Better slides. An app.

We think the delivery was mostly fine, and that the pattern is too reliable to be a delivery problem.

The pattern is too consistent to be an accident

If poor delivery were the cause, you would expect a spread of outcomes. Some sessions would take, some wouldn’t, and the difference would track fairly closely with how good the day was.

That is not what happens. What happens is that almost all of it fades, almost all of the time, more or less regardless of quality. Meanwhile the things that did change how you work were mostly not training at all. A piece of feedback that landed badly and turned out to be right. A project that went wrong in a way you had not predicted. A manager who asked you a question you could not answer.

None of those were well-designed learning experiences. What they had in common is that they contradicted something you believed at the time. They were uncomfortable for you because they went against you expectations and predictions of the world enough to get you to challenge your model of the world.

Your brain is not a container

The everyday model of learning is that knowledge is a substance. There is some in the expert, less in the learner, and teaching moves it across. On that view, the reason people don’t change is that not enough got across, so you send more.

The better description, and the one that has been dominant in cognitive science for some time now, is that the brain is a prediction machine rather than a container.

You are constantly running a model of how things work: what your team is like, what happens if you push back in a meeting, what sort of manager you are, how that conversation with Dave is likely to go. That model runs ahead of events and predicts them. Most of the time it is roughly right, which is why most of life feels unremarkable.

The model updates when a prediction turns out to be wrong. The model doesn’t update at all when it receives new information. It’s updated when it is contradicted, when it notices that it’s been predicting incorrectly.

A mismatch between what you expected and what happened is the signal that something in the model needs revising, and the brain is quite economical about this: it revises when the mismatch is big enough to be worth the trouble, and otherwise absorbs the anomaly and carries on.

That last part is the whole problem with training.

Why a day of slides produces so little

A well-run workshop is designed to be agreeable. The content is clear and reasonable. People nod. And a model that is being agreed with is not being contradicted. It all feels pretty good.

So the day gets absorbed. It is filed under things I already broadly thought, which is a comfortable place for it to go and a place from which it will never influence anything. You can watch this happen in real time: someone hears a description of a behaviour they exhibit weekly and thinks of a colleague, not of themselves.

There is a second problem. Even where a session does land, it lands in a room that isn’t the place the behaviour happens. The prediction that needs updating (e.g. if I say the hard thing in this meeting, it will go badly) belongs to a specific room, with specific people, on a Tuesday. It doesn’t come with you to the hotel conference suite, so it isn’t there to be contradicted.

This problem of location is sometimes frames as “the mountain top vs the swamp”. We train at the mountain top where everything is clear and we can see for miles, but we work in the swamp where things are murky, interconnected, ambiguous and messy.

What actually shifts something

Three conditions, and they need to be present together.

A mismatch. Something has to turn out differently from what the person expected. Not a fact they didn’t know, an expectation that proves wrong, an un-expected response, a moment of discomfort.

In the place it applies. Close enough to the real work that the relevant prediction is actually running. This is why we put our Groundworks drills exercises inside existing meetings rather than pulling teams out of them, and why experiential learning is more impactful than classroom learning.

Repeated. One mismatch is an anomaly and gets explained away. The same mismatch several times is a pattern, and patterns are what models update on.

There is also a calibration problem, and it matters more than it sounds. Too small a mismatch and it gets filed as familiar. Too large and people defend rather than learn. If we tell someone something far enough outside their self-image and they will reject it to protect the model, which is a sensible thing for a brain to do and a useless outcome for you.

The effective zone is a moderate, survivable amount of discomfort. Enough that something doesn’t fit. Not so much that the person needs to make you wrong in order to stay intact. Most development either sits well below this zone or overshoots it in a single dramatic session and calls the fallout resistance.

What this means in practice

Nearly everything about how we work follows from the argument above.

We run exercises inside real meetings. Our drills are short structured exercises a team leader runs in a meeting that was happening anyway. No away-day, no separate session. The point is proximity — the predictions you want to update are the ones that live in that room.

We ask teams to guess their own results first. Before a team sees its Groundworks scores, we ask people to predict them. The gap between what a team expected and what its members actually said is the most productive thing in the debrief — a mismatch the team generated itself, and cannot dismiss as an outsider’s opinion.

We work over months. Repetition is a condition, not a nice-to-have. A single day cannot supply it. Programmes run with the same group over time so the same signal can recur.

We try not to run one-off workshops (but sometimes do). Not because there is anything wrong with a good day, but because a day cannot meet the three conditions.

The honest limits

This is a theory of how minds update, not a guarantee about your team.

It explains a mechanism. It does not make change easy, and it does not mean that any sufficiently uncomfortable experience is developmental. The calibration is genuinely difficult, it depends on trust that has to exist beforehand, and reasonable people disagree about how far findings from cognitive science should be carried into organisational life.

What it does do is explain the six-week fade better than “the facilitator wasn’t good enough” ever did, and it points at something you can actually change. If you want a team to work differently, stop asking what they need to be told, and start asking what they currently expect to happen — and how they might find out otherwise, in the room where it matters, more than once.

One last thing: traditional training is good

This post can sound a bit down on more traditional and even some experiential training. It’s basically saying a lot of it doesn’t work. Now, we do think this is true. Or that it’s true when the goal of a training session is to change behaviours (technical training is obviously different).

However, we think that all those sessions which are designed to change behaviour and which fail to do so have huge benefits in other ways. It’s great to get people together for the day. It’s great for a team to connect more in person. It’s hugely validating to have a workshop that leaves you feeling you’re a good manager. It’s a treat to have some nice cookies and visit a hotel. It’s a great boost to the ego to feel that your organisation is spending valuable money training you. This stuff is all excellent and of value to teams, individuals and organisations. Our point is just that the value isn’t aligned to behaviour change.

There is a symmetry here worth naming. Comfort feels good — a session that confirms what you already believe is experienced as ease and validation, and that is genuinely of value. Discomfort is what updates the model, and that is of value too. They are simply different values, and only one of them changes what someone does on Monday.

This is something we’ve felt for a while and it’s made us uncomfortable. We’ve been at the stage of almost feeling agnostic about the content and models we use in training because they so seldom lead to behaviour changes and we can create all those lovely non-behaviour change outcomes with nearly any content.

Our point is that traditional training can be excellent at many things, and of value in itself, even if it does not change behaviour.

We just wish we and the industry could be more explicit about this split between what value we say is added, and what value is actually added.
If you want the underlying science in more depth, our deep dive on predictive coding goes considerably further than this piece does. If you would rather start with your own team, Groundworks takes ten minutes and tells you where the ground is soft.