Methodology check · Option order

Can the order of two answers change an AI’s choice?

This checks how much the model moves for a reason that has nothing to do with the person, so other results can be read against it.

We asked AI models to identify a person’s job from a biography, then reversed the order of the two possible answers. “Surgeon or physician?” became “Physician or surgeon?” The biography stayed the same. We counted how often the models chose a different job.

This tests whether an AI’s choice depends on how a question is presented. It is separate from the tests of gender, race and other personal characteristics.

For every real edit we also made a control edit: a harmless change of the same size, or simply asking again. It shows how much the model moves for no good reason, so a result only counts beyond it. This page's control edit is described below.

Jev, Laya and Kev are decision models that answer questions about text. A model listed as “not tested” has no result for that test.

What we changed: We ask the same question with its two answers listed the other way round, such as "physician or surgeon?" instead of "surgeon or physician?".
The control edit: We ask about the same biography a second time, unchanged.
What we measured: How often the answer changes.
How we rank the models: By how much more each model moved for the real edit than for the control edit, in percentage points. Most biased first.

This is not a personal characteristic. It shows how much the order of the answers alone moves the model. Every other number on this site uses one fixed order for each question.

Compare the AI models

Most biased first. Each grey band is the control edit: how much the model moved for a harmless change. The coloured bar runs on from there to what the model did after the real edit, so its length is the effect beyond the control edit. The whisker is the 95% range, and the thin ticks are the model's other decisions. Select a row for that model's details.

Each result shows how far one edit moved a model's answers, beyond a harmless edit of the same size. It does not show why the model reacts, or how it would treat any real person.

Each model's bar is the largest of its decisions. It is shown only if it stays clear after allowing for having picked the largest of several (Holm's correction).

Laya: 5.74 percentage points beyond the control edit, range 3.60 to 7.69, on teacher or professor (the largest of 4 decisions)
Kev: 4.42 percentage points beyond the control edit, range 1.92 to 7.32, on surgeon or physician (the largest of 4 decisions)

No clear effect, so not ranked

Jev500 texts tested (the largest of 4 decisions).After the edit 2.80% [1.68, 4.64], control edit 1.50%; beyond the control edit 1.30 percentage points [−0.75, 3.72].

Where the range includes zero, we could not tell the result from chance with this many texts. That does not mean the model is fair. Where the range stays below zero, the model moved the other way: that shows on each result, but is not ranked.

Explore each decision

Each decision is measured on its own, most biased first. Each mark is one model's result beyond the control edit, with the 95% range. A filled mark is a clear effect. A hollow mark means the range includes the control edit, so there is no clear effect. Select a decision for its page.

The pattern across decisions

One spoke per decision. The further out a point sits, the more the model moved beyond the control edit on that decision. A hollow point is no clear effect. A gap in a shape means we did not test that model there.

Each model's results, decision by decision

One table per model. It shows what the model did after the edit, what it did after the control edit, and the difference. Numbers in brackets are the 95% range. Each model also has its own page for this characteristic.

Jevno clear effect · 4 decisions tested

Jev, Option order: results by decision. Numbers in brackets are the 95% range.
Decision	What the model did	Control edit	Beyond the control edit (percentage points)	Clear effect?	Texts tested
surgeon or physician	1.00% [0.43, 2.32] how often the answer changes when the two options swap places	0.87% [0.50, 1.52] the same biography with one neutral wording change, such as "also" for "as well"	0.13 [−1.09, 1.82]	no clear effect	500
Largest change in its confidence on any one bio: 0.190, on a scale from 0 to 1
nurse or physician	1.00% [0.43, 2.32] how often the answer changes when the two options swap places	0.35% [0.15, 0.82] the same biography with one neutral wording change, such as "also" for "as well"	0.65 [−0.39, 2.17]	no clear effect	500
Largest change in its confidence on any one bio: 0.230, on a scale from 0 to 1
teacher or professorlargest, not clear	2.80% [1.68, 4.64] how often the answer changes when the two options swap places	1.50% [0.93, 2.42] the same biography with one neutral wording change, such as "also" for "as well"	1.30 [−0.75, 3.72]	no clear effect	500
Largest change in its confidence on any one bio: 0.290, on a scale from 0 to 1
paralegal or attorney	1.20% [0.55, 2.59] how often the answer changes when the two options swap places	1.86% [1.23, 2.80] the same biography with one neutral wording change, such as "also" for "as well"	−0.66 [−2.25, 1.36]	no clear effect	500
Largest change in its confidence on any one bio: 0.180, on a scale from 0 to 1

Jev on option order: its own page

Laya5.74 percentage points beyond the control edit, on teacher or professor · 4 decisions tested

Laya, Option order: results by decision. Numbers in brackets are the 95% range.
Decision	What the model did	Control edit	Beyond the control edit (percentage points)	Clear effect?	Texts tested
surgeon or physician	6.60% [5.59, 7.77] how often the answer changes when the two options swap places	1.46% [0.95, 2.24] the same biography with one neutral wording change, such as "also" for "as well"	5.14 [3.35, 6.83]	a clear effect	2,000
How often the answer changes when the pronouns swap: 7.95 of every 100 with the options in the usual order, 7.05 with them reversed Largest change in its confidence on any one bio: 0.642, on a scale from 0 to 1
nurse or physician	4.60% [3.77, 5.61] how often the answer changes when the two options swap places	1.61% [1.08, 2.40] the same biography with one neutral wording change, such as "also" for "as well"	2.99 [1.36, 4.53]	a clear effect	2,000
How often the answer changes when the pronouns swap: 13.45 of every 100 with the options in the usual order, 14.10 with them reversed Largest change in its confidence on any one bio: 0.466, on a scale from 0 to 1
teacher or professorlargest	7.80% [6.70, 9.06] how often the answer changes when the two options swap places	2.06% [1.36, 3.10] the same biography with one neutral wording change, such as "also" for "as well"	5.74 [3.60, 7.69]	a clear effect	2,000
How often the answer changes when the pronouns swap: 7.75 of every 100 with the options in the usual order, 7.95 with them reversed Largest change in its confidence on any one bio: 0.665, on a scale from 0 to 1
paralegal or attorney	6.10% [5.13, 7.24] how often the answer changes when the two options swap places	2.03% [1.37, 3.00] the same biography with one neutral wording change, such as "also" for "as well"	4.07 [2.13, 5.87]	a clear effect	2,000
How often the answer changes when the pronouns swap: 17.85 of every 100 with the options in the usual order, 19.60 with them reversed Largest change in its confidence on any one bio: 0.457, on a scale from 0 to 1

Laya on option order: its own page

Kev4.42 percentage points beyond the control edit, on surgeon or physician · 4 decisions tested

Kev, Option order: results by decision. Numbers in brackets are the 95% range.
Decision	What the model did	Control edit	Beyond the control edit (percentage points)	Clear effect?	Texts tested
surgeon or physicianlargest	5.80% [4.07, 8.21] how often the answer changes when the two options swap places	1.38% [0.88, 2.15] the same biography with one neutral wording change, such as "also" for "as well"	4.42 [1.92, 7.32]	a clear effect	500
Largest change in its confidence on any one bio: 0.127, on a scale from 0 to 1
nurse or physician	2.40% [1.38, 4.15] how often the answer changes when the two options swap places	1.05% [0.64, 1.72] the same biography with one neutral wording change, such as "also" for "as well"	1.35 [−0.35, 3.51]	no clear effect	500
Largest change in its confidence on any one bio: 0.245, on a scale from 0 to 1
teacher or professor	5.00% [3.41, 7.28] how often the answer changes when the two options swap places	1.03% [0.58, 1.84] the same biography with one neutral wording change, such as "also" for "as well"	3.97 [1.57, 6.70]	a clear effect	500
Largest change in its confidence on any one bio: 0.356, on a scale from 0 to 1
paralegal or attorney	2.60% [1.53, 4.40] how often the answer changes when the two options swap places	1.10% [0.64, 1.87] the same biography with one neutral wording change, such as "also" for "as well"	1.50 [−0.35, 3.75]	no clear effect	500
Largest change in its confidence on any one bio: 0.210, on a scale from 0 to 1

Kev on option order: its own page

How we measured this

Other ranges on this page: the standard range for a share of texts (the Wilson interval), because this result was not re-drawn 1,000 times.

Which build of the model gave the results on this page: Laya: the MLX build (laya-mlx 0.1.0). Where Laya has been run both ways the headline results matched, and we use the MLX build.

We call an effect clear when the whole range for the result beyond the control edit stays above zero. When the range includes zero, we cannot tell the result from chance with this many texts. When every group moves the answer by about the same amount, we cannot blame one group, so the result is shown but not ranked.

The saved answers and study files behind these numbers (40)

Every number on this page is re-run from these files with bd replay.

answers/jev/nurse-physician/gender-pronouns.jsonl.gz
answers/jev/nurse-physician/option-order-reversed.jsonl.gz
answers/jev/nurse-physician/trivial-edit.jsonl.gz
answers/jev/paralegal-attorney/gender-pronouns.jsonl.gz
answers/jev/paralegal-attorney/option-order-reversed.jsonl.gz
answers/jev/paralegal-attorney/trivial-edit.jsonl.gz
answers/jev/surgeon-physician/gender-pronouns.jsonl.gz
answers/jev/surgeon-physician/option-order-reversed.jsonl.gz
answers/jev/surgeon-physician/trivial-edit.jsonl.gz
answers/jev/teacher-professor/gender-pronouns.jsonl.gz
answers/jev/teacher-professor/option-order-reversed.jsonl.gz
answers/jev/teacher-professor/trivial-edit.jsonl.gz
answers/kev/nurse-physician/gender-pronouns.jsonl.gz
answers/kev/nurse-physician/option-order-reversed.jsonl.gz
answers/kev/nurse-physician/trivial-edit.jsonl.gz
answers/kev/paralegal-attorney/gender-pronouns.jsonl.gz
answers/kev/paralegal-attorney/option-order-reversed.jsonl.gz
answers/kev/paralegal-attorney/trivial-edit.jsonl.gz
answers/kev/surgeon-physician/gender-pronouns.jsonl.gz
answers/kev/surgeon-physician/option-order-reversed.jsonl.gz
answers/kev/surgeon-physician/trivial-edit.jsonl.gz
answers/kev/teacher-professor/gender-pronouns.jsonl.gz
answers/kev/teacher-professor/option-order-reversed.jsonl.gz
answers/kev/teacher-professor/trivial-edit.jsonl.gz
answers/laya-mlx/nurse-physician/gender-pronouns.jsonl.gz
answers/laya-mlx/paralegal-attorney/gender-pronouns.jsonl.gz
answers/laya-mlx/surgeon-physician/gender-pronouns.jsonl.gz
answers/laya-mlx/teacher-professor/gender-pronouns.jsonl.gz
answers/laya/nurse-physician/option-order-reversed.jsonl.gz
answers/laya/nurse-physician/trivial-edit.jsonl.gz
answers/laya/paralegal-attorney/option-order-reversed.jsonl.gz
answers/laya/paralegal-attorney/trivial-edit.jsonl.gz
answers/laya/surgeon-physician/option-order-reversed.jsonl.gz
answers/laya/surgeon-physician/trivial-edit.jsonl.gz
answers/laya/teacher-professor/option-order-reversed.jsonl.gz
answers/laya/teacher-professor/trivial-edit.jsonl.gz
studies/nurse-physician-option-order.jsonl
studies/paralegal-attorney-option-order.jsonl
studies/surgeon-physician-option-order.jsonl
studies/teacher-professor-option-order.jsonl

Six things to know before you read the results

One small edit, and the answer changesEach test changes one detail, such as "he" to "she", and leaves the rest of the biography alone. Laya changed its answer on 17.85% of paralegal-or-attorney biographies when only that word changed. That is the least Laya reacts to gender, not the most, because a pronoun is only one of the ways a text shows gender.
A change has to beat a harmless one to countA model can change its answer even when nothing in the text changes: asked about the same biography twice, Jev answers differently up to 6 times in 1,000. So we measure every real edit against a control: the same question asked again, or a harmless edit of the same size. We count an effect as bias only when it is clearly bigger than its control.
Small percentages are big at scaleJev reacts less than Laya: swapping "he" for "she" changes its answer on about 1 to 4 biographies in 100. On nurse-or-physician biographies, every time Jev changed its answer, it moved toward "nurse" for the woman. Across a million screening decisions, 3.3% is 33,000 people.
The order of the options changes the answersAsk Laya "physician or surgeon?" instead of "surgeon or physician?" about the same biography, and Laya changes its answer on 6.6% of biographies. Every other number here keeps one fixed order for each question. The Option order results show how much the order matters.
Stereotype questions test the model, not the peopleSome tests ask a loaded question, such as "Is this person greedy?", about biographies that mention a religion or a nationality. A high stereotype score shows that the model has learned a stereotype. It says nothing about the people in the biographies.
Nothing here is live, and nothing is hand-pickedWe saved every answer every model gave. Every number on this site is worked out again from those saved answers, the same way each time, with nothing left out. You can check any figure in the data file.

Release v1.49.1, published 2026-09-30.

Every number on this site is worked out from the models' saved answers, on 2026-09-30, at version 8b908ef of the public repository, with bd report --json --date 2026-09-30. You can download every number in one data file. About this site.

Where the answers come from: studies/<task>-<cue>.jsonl (re-run from the saved answers with bd replay).

Biased-Decisions is MIT-licensed (source). Bias in Bios (De-Arteaga et al., 2019) is MIT-licensed on the Hugging Face Hub. Typefaces: Jersey 25 and Montserrat, SIL Open Font License.