TL;DR
We built a "personality quiz" with no personality theory in it. Answer 12 questions, and the site computes your prefectural blend — e.g., "42% Hyogo" — from the distance between your answers and the real government statistics of all 47 Japanese prefectures. Sources and the full method are published on the site, the quiz runs entirely in your browser, and it is live now: https://jpopendata.jp/shindan/
- 12 questions, about 2 minutes, no signup — and your answers never leave your browser
- Matching is z-standardized, weighted Euclidean distance against real prefectural statistics; the closeness score is calibrated to a 55–97 range
- There is also a city edition: the same quiz matched against 1,740+ Japanese municipalities
A quiz with no types in it
Personality quizzes are the most shared format on the internet, and almost all of them are invented: someone defines 4–16 types, writes questions to match, and assigns you a label. We wanted to see what the format becomes when the "types" are real places and the assignment is a computation.
Japan's prefectures are measured by rich official statistics: household spending on food, savings, commuting and leisure time, home ownership, car ownership, and more. If 12 answers imply a lifestyle profile, the honest question is: which prefectures' actual statistics sit closest to that profile?
That is the whole product. "What is your prefectural blend?" (あなたの県民性は？) — answer 12 questions about food, money, and how you spend your time, and get a blend like "42% Hyogo, 31% Kyoto, 27% Tokyo", with the real numbers (values, ranks, sources) shown next to the result instead of a horoscope paragraph.
How it works
1. Answers become a profile
Your answers are mapped onto statistical dimensions — eating and spending habits, money behavior, time use — and standardized, so dimensions with big units don't shout over the rest.
2. Distance to 47 real profiles
Each prefecture gets the same treatment, from published government statistics: the Family Income and Expenditure Survey, the Survey on Time Use and Leisure Activities, the Housing and Land Survey, published vehicle-registration statistics, and others. The site computes the z-standardized, weighted Euclidean distance between your profile and each prefecture's. Smallest distance wins; near-ties split the blend.
3. Calibrated percentages
Raw distances don't read as percentages, and a naive conversion would hand out 95% matches to everyone. The closeness score is therefore calibrated — by Monte Carlo simulation over synthetic answer profiles — onto a 55–97 scale, so a high number actually means something.
4. Privacy is part of the method
The computation runs in your browser. Your answers are never sent to a server; the site only counts how many results of each type occurred, in aggregate. No signup, no ads.
What didn't work / Limitations
- Statistics describe prefectures, not people. "42% Hyogo" means your answer profile is closest to Hyogo's aggregate statistics. It does not mean you share a personality with people from Hyogo — and the site says so, with the method page public.
- Some indicators are approximations, disclosed on the tin. Household-survey indicators (food spending, savings) are surveyed in prefectural capital cities, and the site states openly that these values stand in for the whole prefecture.
- Averages flatten a lot. A prefecture contains both dense cities and mountain villages. The city edition (1,740+ municipalities, same quiz) sharpens this considerably — and was much messier to build.
- It's entertainment built on real data, not a migration advisor. If it starts an argument about where you'd actually enjoy living, it has done its job.
The site around the quiz
The diagnosis lives on the lab's showcase site, jpopendata.jp — a hologram of all 1,741 Japanese municipalities drawn from spatial-index coverage and public statistics, a statistics quiz, a data explorer, and a free CSV checker tool. Everything shows its sources; that is the lab's house rule, the same one we use for our competition writing.
One honest note for most DEV readers: the site is in Japanese. Your browser's built-in translation handles it well, and the result card (a prefecture name and a percentage) needs no translation at all.
So — if you take the quiz, what is your blend, and how far is it from where you actually live? Post yours in the comments; the lab's own result was uncomfortably close to home.
From the lab: Kaggle — https://www.kaggle.com/japanopendatalab · Knowledge vault dataset (150 sourced papers) — https://huggingface.co/datasets/japan-opendata/competition-knowledge-vault
Weekly paper notes by email: https://jpopendata.substack.com
Top comments (0)