Social & Critical Psychology · Simplified
Four People, One Score
Everyone in this activity holds the same view. Almost nothing else about them is the same.
What the model is and is not
What the model is and is not. The likelihoods come from a small logistic model whose coefficients are printed below and in the teaching notes. They are chosen to make a teaching pattern visible. They are not estimates of anything, they come from no dataset, and they should never be quoted as effect sizes. This page also takes no position on whether voicing disagreement is a good idea: it estimates a frequency, not a virtue, and the costs of speaking up fall very unevenly across a workforce.
Step 1 of 3
First, a prediction
Aisha scores 0.86 out of 1 on an attitude scale about speaking up when she disagrees in a team meeting. It is a well-made scale, matched to the behaviour, and she answered it honestly. Out of the next thirty meetings, how many will she speak up in?
Step 2 of 3
Four colleagues, all of them Aisha
Every one of these four scores 0.86 on the same scale. The difference between them is not in the column you would have measured.
Key terms
- Perceived norm
- What this person takes the people around them to expect and to do.
- Perceived control
- How far they believe the act is available to them, given who they are in this room.
- Opportunity
- Whether the occasion arises at all. In the model below it multiplies rather than adds, so when it is near zero nothing else matters.
- Correspondence
- How closely the measure matches the act. The scale here is a behaviour-level attitude scale, which is a favourable case.
| Person | Attitude | Norm | Control | Habit | Constraint | Opportunity | Speaks up in |
|---|
Step 3 of 3
Three hundred colleagues, twice
Three hundred simulated people, each observed over thirty meetings. Their attitude scores are exactly the same in both versions below. The only thing that differs is whether they are all in the same kind of room.
One marker per person. Across is their attitude score, up is the share of thirty meetings in which they spoke up. The markers move up and down when the conditions change, and never left or right.
What this shows
The gap is not a fact about attitudes
Key idea: The four colleagues hold the same view and speak up at completely different rates, and the reason is in columns a study of attitudes would not have collected. Marek is the case worth keeping: every internal condition he has is as strong as Rowan's, and the occasion simply does not arise in his team, so a course aimed at changing his attitude would change nothing at all. Then the three hundred make a further point. The attitude-behaviour correlation is often reported as though it were a property of attitudes, or of people, or of the concept of an attitude. Here the same attitude scores produce a correlation twice as large when everybody is in the same kind of room, because when the conditions are held constant the only thing left varying is the attitude. A low attitude-behaviour correlation is partly a measurement of how unequal the conditions are.
A simulation cannot test a theory. The coefficients are invented, the correlations that follow from them are not findings, and no number here estimates anything. Six factors is a deliberate simplification of a literature with many more, and the model has no feedback in it at all: in reality speaking up changes the norm, the norm changes the opportunity, and being the person who spoke up last time changes what is available to you next time. Two further cautions. Holding conditions constant to raise a correlation is a description of what happens in a tidy workplace, not a research recommendation, and a study that removes variation in conditions has removed the thing that most needed explaining. And the model estimates a frequency and not a virtue: the costs of speaking up fall very unevenly, and Devi's four months on a fixed-term contract are a reason rather than a deficit.
The longer version lets you move every factor for every person, switch the questionnaire between a general values scale and a matched intention to see what correspondence does, and work through a third experiment on where the correlation comes from. It is at Attitude-Behaviour Gap in the main collection.