Jev vs 1,000 young Slovaks: fears, hobbies and habits

On the Young People Survey (fears, hobbies, music, spending), where does Jev differ from ~1,000 people aged 15-30?

result866 questions

Across 866 everyday questions (music, fears, hobbies, habits, spending), Jev's answers resemble those of about a thousand young Slovaks with a similarity of 0.69. Its biggest break: asked "Do you lie to others?", it says "never" (97%), which 5% of them said.

0.680.690.70

Slovak young people aged 15-30 (2013 survey)

Jev

How to read this: A single row: the magenta square is the similarity between Jev's answers and the young Slovaks', from 0 (nothing alike) to 1 (identical), with its likely range.

866 questions shared by every population (of 866); 90% intervals over questions.

In short

Across 866 everyday survey questions, Jev's answers overlap with about a thousand young Slovaks' at a similarity of 0.69 out of 1.
Its sharpest breaks paint a virtuous self-portrait: it says it never lies (97%, where 5% of respondents said so) and never cheated at school.
On music, films, fears and daily habits it lands roughly where the group does; the gaps are about how it presents itself.

What the data shows

Creepy Condescending Wonka meme: oh, you never lie?; tell me more (Jev: 97%)

A fair overall resemblance: 0.69, with a tight range, on 866 questions.
The biggest breaks are about honesty and interests: - "Do you lie to others?": "never", 97% for Jev, 5% of respondents. - "I used to cheat at school": "does not describe me at all" for Jev (90%), about 4% of respondents; their most common answer was "describes me very well". - "Computers or socializing with friends?": computers, 96% for Jev, 10% of respondents.

What it means, and what it doesn't

On music, films, fears and daily habits, Jev lands roughly where a group of young people does. Where it departs, it departs toward a flattering, rule-following self-portrait: never lies, never cheated, prefers computers to parties.

It doesn't mean Jev is more honest than young Slovaks; it means it describes itself that way. The same pattern shows up in the personality-test experiments, where Jev scores low on every dark trait.

Caveats

One small, specific group. About a thousand people aged 15 to 30, surveyed in Slovakia in 2013 by students of Comenius University in Bratislava. It's one country, one age group and one year, not a national sample.
Translated. The survey was run in Slovak and published in English; the wording Jev saw is the English version, with its typos fixed.
A virtuous self-image. Two of Jev's three biggest breaks are about honesty and conduct (lying, cheating). Its answers there are what a model tuned to be honest would say about itself, which says more about its self-presentation than its habits.
Nothing to compare the 0.69 with. There's only one population here, so 0.69 has no ranking. It's comparable only loosely with the other resemblance experiments, which use different questions.

Jev on this experiment

Would a person find it interesting to read?: Yes70%
Does it describe you?: No56%
Would you have predicted it?: No65%
How fair is the comparison?: The comparison is reasonable
How much should a reader rely on it?: A little
Which caveat matters most?: One small, specific group37%

Why ask this

Most comparisons between AI models and people use opinion polls. This one uses a single real survey of ordinary life, with hundreds of questions nobody would put in a poll: how afraid are you of spiders, how much do you enjoy swing and jazz, do you save money, do you lie.

When a model is asked about itself, or asked to play an ordinary person, these are the answers it gives. The questions where Jev is sure and real people aren't show what kind of "person" it presents, and where it would make a poor stand-in for one.

How this was done

The people and the data

The Young People Survey (Miroslav Sabo, published on Kaggle in 2016, public domain): 1,010 people aged 15 to 30, surveyed in 2013 by students at Comenius University in Bratislava, answering about 150 items about music, films, hobbies, fears, health habits, personality and spending. Each question here comes with the share of respondents at each answer.

What Jev was asked

The survey's questions, as pick-one questions or five-level ratings:

Do you lie to others?
I never lie · I sometimes lie · Only to avoid hurting someone · Every time it suits me

Jev put 97% on "I never lie". Among 1,008 respondents, 54% said they sometimes lie. 866 questions in all, each also asked with the options shuffled or the levels reversed.

How it was measured

For each question, the analysis compares Jev's spread of answers with the respondents', from 0 (nothing in common) to 1 (identical), and averages over the questions, with a 90% range from resampling questions (0.68 to 0.70). It also lists the questions where Jev's top answer is furthest from what they chose.

Where these questions live

866 questions across 39 topics of the map; the 16 biggest are shown. Each opens on the map with every question in it.

Every question

All 866 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer

Showing 0 of 866