The Big Five · open data · runs in your browser
A personality test that shows its work.
Answer 20 public-domain statements (or stop after 10). A model trained on 597,100 real responses places you on each of the five big dimensions of personality, with its margin of error and the answers that moved it.
20 questions · ~3 min · no sign-up · nothing leaves your device
- Statements
- 50
- IPIP, public domain
- Responses
- 597,100
- Open Psychometrics training data
- Typical error
- ±0.26
- points on a 1–5 scale, after 20 questions
- Servers
- 0
- inference in your browser
01 — Choose your depth
Choose your depth.
Every extra answer narrows the error bars. Stop whenever you like after ten; the result always says how sure it is.
Quick
20 questions · ~3 min
The short form. Close to the full test, in a fraction of the time.
Start quickExplains 88% of the full test
Typical error ±0.26 pts
Standard
30 questions · ~5 min
Tighter error bars, still well under ten minutes.
Start standardExplains 94% of the full test
Typical error ±0.19 pts
Full
50 questions · ~8 min
All fifty items, scored exactly. No model, no error band.
Start fullThe full test, scored exactly
No model, no error band
02 — The five traits
Five dimensions, no verdicts.
Each trait is its own spectrum. Neither end is better; most people sit somewhere in the middle of most of them.
Extraversion
Where you tend to get your energy: from people and activity, or from quiet and depth.
ReservedOutgoingNeuroticism
How strongly you tend to feel stress, worry and changes in mood. Sometimes called emotional sensitivity.
Even-keeledSensitiveAgreeableness
How much you orient towards other people's feelings and needs compared with your own view and interests.
DirectWarmConscientiousness
How much you prefer structure, planning and follow-through compared with flexibility and spontaneity.
SpontaneousOrganisedOpenness
How drawn you are to ideas, imagination and new experiences compared with the concrete and familiar.
GroundedCurious
03 — Why it was rebuilt
The old version gave 89% of people the same answer: Openness.
It fed raw 1–5 answers into a scaler that expected centred data, so the model's outputs barely moved from one person to the next. A softmax across the five traits then crowned whichever one saturated highest.
The rebuild scores each trait on its own scale, compares you with real people, and checks that no trait is favoured: across held-out respondents, each one is named the most distinctive roughly a fifth of the time.
Share of people whose most distinctive trait is…
- Extraversion
- 0%20%
- Neuroticism
- 0%20%
- Agreeableness
- 10%20%
- Conscientiousness
- 2%21%
- Openness
- 89%19%
Old app: its “dominant” trait, replayed on held-out respondents. Rebuilt model after 25 statements: the trait furthest from the median, as shown on your results.
04 — Privacy
Private by construction.
There is no server to send anything to. The model is about 205 KB of numbers that load with the page.
Runs on your device
Scoring happens in your browser with the same arithmetic the training code used. You can read every line of it.
Nothing is stored
No accounts, cookies, analytics or logs. Close the tab and your answers are gone.
You choose who sees it
A share link carries your answers after the # in the address, which browsers never send to a server.
Ready when you are
20 statements. About 3 minutes.
Begin