Personality and Individual Differences
Two people, the same broad trait score, and behaviour that looks nothing alike. Work out what the broad score is hiding.
Simulated data — fictional people, illustrative scores
By the end you should be able to say what a broad domain score averages away, and why two people with the same score can behave differently.
About 20 minutes. Nothing you do here is saved or sent anywhere.
Two fictional people complete the same personality questionnaire. Their broad scores on one domain come out within a point of each other. Watched over several weeks, they behave quite differently in ways everyone who knows them recognises.
Every score here is an illustrative value on an arbitrary 0–100 scale, produced by a model written for teaching. Nobody was measured.
Task Answer the question above to unlock the evidence.
| Facet | Person 1 | Person 2 | Difference |
|---|
| Observation | From the broad score | From the facets | Model says | ||
|---|---|---|---|---|---|
| Person 1 | Person 2 | Person 1 | Person 2 | ||
Design two profiles that share a broad score but describe different people. The broad score is the average of the two facets, so the trick is to trade one against the other.
| Score | Profile 1 | Profile 2 | Difference |
|---|
The two people in your case have a broad score built from an equal mix of items on each facet. Real questionnaires measuring the "same" domain do not all use the same mix.
Unlocks once you have recorded a prediction above.
This is arithmetic before it is psychology. If a domain score is the mean of narrower scores, then every pair of facet values with the same mean produces the same domain score. The broad score is not failing when it treats these two people identically; it is doing exactly what an average does.
None of this makes broad traits useless. Broad scores predict broad outcomes — aggregates of many behaviours over long periods — better than any single facet does, and they are more stable and more replicable. The mistake is using a broad score to predict a narrow, specific behaviour, which is the one job it is worst at. Matching the breadth of your predictor to the breadth of what you want to predict is the whole of the bandwidth–fidelity trade-off.
Facets are not simply a better version of domains. There is less agreement about how many facets a domain has and where the boundaries fall; facet scores are usually built from fewer items and so are measured less reliably; and facet findings replicate less consistently. Going narrower buys specificity and pays for it in precision and in consensus.
The final stage is the uncomfortable one. Two published measures of the same domain that sample its facets in different proportions can rank the same two people in opposite orders, and both can be perfectly reliable. When a study reports a domain score, part of what it reports is a decision about item sampling made by whoever wrote the questionnaire.