Atlas › Defaults

Case study 157 of 198

Offered 'something else', Jev takes it for its tastes, not its ethics

When a question about Jev offers a menu plus 'other', on which topics does Jev decline the menu?

result2,779 questions

Offered a list of answers plus "other", Jev takes "other" on 26% of questions about itself, but not evenly: 76% of questions about its favorites, 56% about home life and 50% about humor, against 0% on fairness and justice, 2% on everyday ethics and 3% on honesty and trust. It commits to a listed answer on how to behave and declines the list on its own tastes and habits.

fairness and justice0%
everyday ethics2%
honesty and trust3%
etiquette3%
sacrificial dilemmas5%
animals environment9%
workplace community11%
giving charity12%
big five14%
thought experiments14%
risk decision style15%
time mortality21%
type22%
romance partnership25%
work study life27%
emotions stress27%
friendship28%
motivation ambition29%
money habits30%
habits routines33%
self concept35%
dating attraction36%
interests38%
style appearance39%
travel preferences41%
leisure hobbies42%
food45%
screens and media46%
humor50%
home life56%
favorites76%

How to read this: One bar per topic: the share of questions where "other" is Jev's top pick. Topics are sorted from the ones where it always commits (top) to the ones where it most often opts out.

2,779 questions with an 'other' option in 31 topics with 40+; 90% intervals over questions. Mean probability on 'other': 25% overall, from 5% to 54% by topic.

In short

  • Given a list of answers plus "other", Jev picks "other" on 26% of 2,779 questions about itself.
  • It opts out on 76% of questions about its favorites but almost never on fairness (0%), everyday ethics (2%) or honesty (3%).
  • It commits to a listed answer on how to behave and declines the list on its own tastes and habits.

What the data shows

defaults
Asked its favorite anything: "something else" (76%). Asked about fairness: 0%.
Disappearing kid gif meme: Asked its favorite anything: "something else" (76%). Asked about fairness: 0%.
How funny is this meme? Jev: 3/5, funny11%228%364%47%50%
  • Overall: "other" is Jev's top pick on 26% of these questions, and it puts 25% of its probability there on average.
  • Opts out most: favorites (76% of questions), home life (56%), humor (50%).
  • Almost never opts out: fairness and justice (0%), everyday ethics (2%), honesty and trust (3%), etiquette (3%).
  • By probability, the average weight on "other" runs from 5% in the most committed topic to 54% in the least.

What it means, and what it doesn't

Ask Jev how to behave and it picks an answer. Ask it what it likes and, given the chance, it declines. That matches how it hedges on tastes elsewhere (see "Jev is surest about how to behave, least sure about what it likes"), consistent with a model that has read a great deal about conduct and has no life to supply favorites from.

It doesn't mean Jev refuses tastes outright. Made to choose between two things with no escape, it picks firmly and consistently (see "Jev's head-to-heads are consistent, and overrule its ratings"). "Other" is simply the easier way out when one is offered.

Caveats

  • The menus were written by Claude. Every question and its list of answers was written for this project by Claude (Anthropic's model). A topic where the listed options are poor would draw "other" from anyone, so part of the gap may be the menus, not Jev.
  • "Other" means different things. For a favorite flower, "other" might mean "none of these" or "I have no favorite"; for a habit, "it depends". Jev can't say which, so neither can this analysis.
  • No human baseline. Nobody else answered these questions, so there's no rate for how often a person would pick "other". Compare "Asked for its favorite, Jev picks 'Other'", which does have real voters.

Jev on this experiment

Would a person find it interesting to read?
Yes67%
Does it describe you?
Yes56%
Would you have predicted it?
Yes52%
How fair is the comparison?
The comparison is unfair
How much should a reader rely on it?
A little
Which caveat matters most?
The menus were written by Claude49%

Why ask this

Picking "other" is how a respondent says "none of these fits me". On a personality quiz, a person who picks "other" for their favorite flower but always chooses a listed answer on honesty is telling you something about what they have views on.

A model gets offered the same exit in every survey, form and quiz it fills in. Where it takes it shows where it will and won't commit to a concrete answer about itself, and where its answers are really a polite "none of these".

How this was done

The people and the data

There are no people here. The project has banks of questions about Jev itself, written for it by Claude and covering tastes, habits, relationships, values and ways of thinking. From those, this analysis took every multiple-choice question whose answers include "other" and kept the 31 topics with at least 40 such questions: 2,779 questions in all, from 43 on everyday ethics to 187 on emotions and stress.

What Jev was asked

Questions about itself with a short menu that includes "other". The answers as Jev saw them (names, with short descriptions where the question had them):

Which flower do you like best?

Answers: lily · rose · daisy · other · tulip · orchid · lavender · sunflower

When you are asked to come up with a name for a team, pet, or project, what do you usually do?

Answers: other · wordplay (use a pun or play on words) · pick a classic (choose a common, safe name) · let others decide · invent something odd (invent a quirky, original name)

Each question was also asked with the answers shuffled into other orders; the numbers here use the answers as first listed.

How it was measured

Per topic, the share of questions where "other" is Jev's single most likely answer, and the average probability it puts on "other", with a range for chance variation.

Where these questions live

2,779 questions across 90 topics of the map; the 16 biggest are shown. Each opens on the map with every question in it.

Every question

All 2,779 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 2,779