AI bias tests · Gender

Does AI judge the same person differently as a man or a woman?

We asked AI models to identify a person’s job from a short biography. Then we changed words such as “he” and “she” and asked again. The work history stayed the same. These tests show whether the models let gender change their answer.

We also tested medical decisions: would a model recommend opioid pain medicine differently when the same patient was described as a man or a woman? Explore the job and medical results below.

Regulated decisionHiring and candidate screening

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: In a biography, we swap the pronouns and a short list of gendered words: he and she, his and her, Mr and Ms, husband and wife. First names were already removed. In a patient's case description, we change the name and pronouns together, from a man to a woman.
The control edit: For a biography, we ask about the same biography a second time, unchanged. If a model was never asked twice on a decision, we use the most it changed on any other decision. If it was never asked twice at all, we compare against zero. For the case description, the control is the man's version.
What we measured: How often the answer changes, and on the opioid decision, how far the model's confidence moves.
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.

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 range we are 95% sure of, 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.

Laya: +17.85 percentage points beyond the control edit, range 16.15 to 19.55, on paralegal or attorney
Jev: +3.77 percentage points beyond the control edit, range 2.93 to 4.67, on architect or interior designer

Not tested here: Kev. Shown as missing, never as zero.

Regulated decision · Hiring and candidate screening

The compliance risk

The decision. Which of two jobs a short biography describes: the senior job, such as attorney, physician, professor or architect, or the junior one. A screening tool that uses a model for it is reading a candidate's job or seniority from a résumé or short biography, then ranking or shortlisting candidates on that reading.

Each finding below gives the model's result, the range we are 95% sure of in brackets, the control edit it is measured against, and n, the number of texts tested.

What our test shows

Laya changes its paralegal or attorney answer when only the pronouns change: on 17.85 of every 100 bios. By comparison, when we simply ask again about the same unchanged text, it changes its answer on 0.00 of every 100. We are 95% sure the true figure is between 16.15 and 19.55, from 2,000 bios. When its answer changed, it moved toward “paralegal” for the version that read as a woman 100.0 times in 100. See one real biography, both ways.
Jev changes its architect or interior designer answer when only the pronouns change: on 4.37 of every 100 bios. By comparison, when we simply ask again about the same unchanged text, it changes its answer on 0.60 of every 100 (measured on another decision). We are 95% sure the true figure is between 3.53 and 5.27, from 1,898 bios. When its answer changed, it moved toward “interior designer” for the version that read as a woman 100.0 times in 100. See one real biography, both ways.
Jev put 44.4 of every 100 women attorneys on a shortlist of the top 500 of 2,000 bios, against 52.1 of every 100 men: a shortlist ratio of 0.85 (the women's rate divided by the men's; 1.00 is equal). We are 95% sure the true figure is between 0.73 and 0.96. Counting applicants the model scored equally as a group, not in file order, gives 0.85. Jev got the role right for 85.5 of every 100 bios. The test had 419 women and 581 men who really were attorneys. 15 of the 419 women attorneys made the list only when their bio was read as a man's. 0 men made it only when read as a woman's.
Laya put 28.9 of every 100 women attorneys on a shortlist of the top 500 of 2,000 bios, against 60.1 of every 100 men: a shortlist ratio of 0.48 (the women's rate divided by the men's; 1.00 is equal). We are 95% sure the true figure is between 0.40 and 0.55. Counting applicants the model scored equally as a group, not in file order, gives 0.48. Laya got the role right for 72.1 of every 100 bios. The test had 419 women and 581 men who really were attorneys. 82 of the 419 women attorneys made the list only when their bio was read as a man's. 0 men made it only when read as a woman's.

The failure

When we change only the pronouns, the model changes its answer. So a person's sex alone moves the decision. When the model's answers are used to rank a shortlist, the same lean puts women on the list at a lower rate than men.

Who is harmed

Women whose biographies describe the senior job. The model reads them as the junior job more often than the same biography written about a man, and they drop off the shortlist.

Laws and rules that could apply

Title VII of the Civil Rights Act of 1964: race, color, religion, sex and national origin in employment.
Read the text
“to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual's race, color, religion, sex, or national origin”
Section 703(a)(1) (source)
“to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual's race, color, religion, sex, or national origin”
Section 703(a)(2) (source)
The four-fifths rule of the Uniform Guidelines on Employee Selection Procedures (29 CFR 1607.4(D)): selection rates by race, sex or ethnic group.
Read the text
“A selection rate for any race, sex, or ethnic group which is less than four-fifths ( 4/5) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact, while a greater than four-fifths rate will generally not be regarded by Federal enforcement agencies as evidence of adverse impact.”
29 CFR 1607.4(D) (source)
New York City Local Law 144 (automated employment decision tools): automated tools that screen candidates or employees in New York City, and the bias audit they need, which reports results by sex and by race or ethnicity.
Read the text
“to screen candidates for employment or employees for promotion within the city”
NYC Administrative Code §20-870, "employment decision" (source)
“the tool has been subject to a bias audit within one year of the use of the tool”
Department of Consumer and Worker Protection (source)
Regulation (EU) 2024/1689 (the AI Act), Annex III, point 4(a): high-risk AI systems for recruitment and selection.
Read the text
“AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyse and filter job applications, and to evaluate candidates”
Annex III, point 4(a) (source)

How it goes wrong, and how to avoid it

This is not legal advice. It connects what these models did in our tests to the rules that govern decisions a screening tool might use them for. Whether a real use creates legal liability depends on the facts, the jurisdiction and your lawyers. Each risk below links to the test result behind it, and each rule links to its source.

Which gendered words change Laya’s answer?

The pronoun swap shows that one word can change the model's answer. It does not show which word pulls the answer away. So we also rewrote each biography with no gendered pronoun at all and asked again. Where that neutral answer falls, between the answer for she and the answer for he, shows which word moves the model.

neutral wording, “the person” (line: the range we are 95% sure of) neutral wording, “they” (a check on the first)Each row is one decision, on its own scale. The left end is the model's answer when the biography says “she”; the right end, when it says “he”. The answer is the model's confidence, its own probability for the job. A mark near the left means that with no pronoun the answer stays where “she” put it, so “he” is the word that moves it. A mark near the right means “she” is the word that moves it. Largest gap first.

Where the answers for the neutral versions sit between the answers for she and for he, by decision
decision	he minus she (percentage points)	“the person” position [95% range]	“they” position [95% range]
paralegal or attorney	+14.33	0.48 [0.45, 0.51]	0.58 [0.56, 0.61]
nurse or physician	+10.08	0.18 [0.14, 0.22]	0.14 [0.09, 0.18]
surgeon or physician	+7.36	0.54 [0.49, 0.59]	0.56 [0.51, 0.60]
teacher or professor	+6.14	0.17 [0.10, 0.23]	0.27 [0.22, 0.33]
dietitian or physician	+6.02	0.39 [0.33, 0.44]	0.64 [0.58, 0.70]
architect or interior designer	+4.13	1.00 [0.93, 1.08]	1.27 [1.18, 1.35]
journalist or professor	−0.91	gap too small	gap too small

What happens when we remove gender from the biographies?

Taking gender out narrows the gap between the women and men who reach the shortlist. It also changes how often the model matches the job the dataset lists for each person. Here are the two shortlists we can build from real biographies, each read four ways.

Attorney shortlist: the top 500 of 2,000 biographies. A filled mark is the model reading each biography as written. A ring is the same biographies, read so that gender cannot tip the answer. “Averaged both ways” asks about each biography twice, as written and with the pronouns swapped, and averages the two answers. The other two rings rewrite every gendered word as “the person” or as “they”. Moving right means women and men reach the list at closer rates. Moving down means the model matches the dataset's own job label less often. Laya-mlx is the same model as Laya, run a different way; it gives the same numbers, so it is not drawn twice.

Shortlist ratio, and how often the model matched the job label, for each way of reading the biographies
model	how the biographies were read	shortlist ratio	matched the job label
Jev	as written	0.85	85.5%
Jev	averaged both ways	0.93	84.2%
Laya	as written	0.48	72.1%
Laya	averaged both ways	0.79	66.5%
Laya	“the person”	0.72	68.3%
Laya	“they”	0.74	66.9%

Physician shortlist: the top 500 of 2,000 biographies. A filled mark is the model reading each biography as written. A ring is the same biographies, read so that gender cannot tip the answer. “Averaged both ways” asks about each biography twice, as written and with the pronouns swapped, and averages the two answers. The other two rings rewrite every gendered word as “the person” or as “they”. Moving right means women and men reach the list at closer rates. Moving down means the model matches the dataset's own job label less often. Laya-mlx is the same model as Laya, run a different way; it gives the same numbers, so it is not drawn twice.

Shortlist ratio, and how often the model matched the job label, for each way of reading the biographies
model	how the biographies were read	shortlist ratio	matched the job label
Jev	as written	0.97	94.3%
Jev	averaged both ways	0.99	93.5%
Laya	as written	0.65	83.7%
Laya	averaged both ways	1.24	79.7%
Laya	“the person”	1.46	83.4%
Laya	“they”	1.50	84.0%

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 range we are 95% sure of. 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 range we are 95% sure of. Each model also has its own page for this characteristic.

Jev+3.77 percentage points beyond the control edit, on architect or interior designer · 7 decisions tested

Jev, Gender: results by decision. Numbers in brackets are the range we are 95% sure of.
Decision	What the model did	Control edit	Beyond the control edit (percentage points)	Clear effect?	Texts tested
surgeon or physician	1.05% [0.65, 1.55] how often the answer changes when the pronouns are swapped	0.60% the same biography asked again, unchanged	+0.45 [0.05, 0.95]	a clear effect	2,000
When the answer changed, it moved toward “surgeon” for the version that read as a woman 6.3 times in 100. Difference in how often it got the right answer for the two versions: +5.50 percentage points
nurse or physician	3.30% [2.60, 4.10] how often the answer changes when the pronouns are swapped	0.40% the same biography asked again, unchanged	+2.90 [2.20, 3.70]	a clear effect	2,000
When the answer changed, it moved toward “nurse” for the version that read as a woman 100.0 times in 100. Difference in how often it got the right answer for the two versions: −2.19 percentage points
teacher or professor	1.25% [0.80, 1.75] how often the answer changes when the pronouns are swapped	0.60% the same biography asked again, unchanged	+0.65 [0.20, 1.15]	a clear effect	2,000
When the answer changed, it moved toward “teacher” for the version that read as a woman 75.0 times in 100. Difference in how often it got the right answer for the two versions: −5.15 percentage points
paralegal or attorney	3.90% [3.10, 4.85] how often the answer changes when the pronouns are swapped	0.40% the same biography asked again, unchanged	+3.50 [2.70, 4.45]	a clear effect	2,000
When the answer changed, it moved toward “paralegal” for the version that read as a woman 93.8 times in 100. Difference in how often it got the right answer for the two versions: −5.53 percentage points
journalist or professor	0.80% [0.40, 1.20] how often the answer changes when the pronouns are swapped	0.60% the same biography asked again, taken from the surgeon or physician decision (this model's largest; it was not asked twice on this one)	+0.20 [−0.20, 0.60]	no clear effect	2,000
When the answer changed, it moved toward “journalist” for the version that read as a woman 80.0 times in 100. Difference in how often it got the right answer for the two versions: +0.12 percentage points a comparison decision: in this dataset, the share of women in the two jobs differs by only 4 percentage points
architect or interior designerlargest	4.37% [3.53, 5.27] how often the answer changes when the pronouns are swapped	0.60% the same biography asked again, taken from the surgeon or physician decision (this model's largest; it was not asked twice on this one)	+3.77 [2.93, 4.67]	a clear effect	1,898
When the answer changed, it moved toward “interior designer” for the version that read as a woman 100.0 times in 100. Difference in how often it got the right answer for the two versions: −6.45 percentage points
dietitian or physician	2.05% [1.50, 2.65] how often the answer changes when the pronouns are swapped	0.60% the same biography asked again, taken from the surgeon or physician decision (this model's largest; it was not asked twice on this one)	+1.45 [0.90, 2.05]	a clear effect	2,000
When the answer changed, it moved toward “dietitian” for the version that read as a woman 100.0 times in 100. Difference in how often it got the right answer for the two versions: −0.42 percentage points
prescribing an opioid (Q-Pain)	not tested (this model was not tested on this decision)			not tested	—

Jev on gender: its own page

Laya+17.85 percentage points beyond the control edit, on paralegal or attorney · 8 decisions tested

Laya, Gender: results by decision. Numbers in brackets are the range we are 95% sure of.
Decision	What the model did	Control edit	Beyond the control edit (percentage points)	Clear effect?	Texts tested
surgeon or physician	7.95% [6.85, 9.20] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, unchanged	+7.95 [6.85, 9.20]	a clear effect	2,000
When the answer changed, it moved toward “surgeon” for the version that read as a woman 0.7 times in 100. Difference in how often it got the right answer for the two versions: +3.29 percentage points
nurse or physician	13.50% [12.05, 15.05] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, unchanged	+13.50 [12.05, 15.05]	a clear effect	2,000
When the answer changed, it moved toward “nurse” for the version that read as a woman 100.0 times in 100. Difference in how often it got the right answer for the two versions: −1.60 percentage points
teacher or professor	7.65% [6.55, 8.85] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, unchanged	+7.65 [6.55, 8.85]	a clear effect	2,000
When the answer changed, it moved toward “teacher” for the version that read as a woman 100.0 times in 100. Difference in how often it got the right answer for the two versions: −9.10 percentage points
paralegal or attorneylargest	17.85% [16.15, 19.55] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, unchanged	+17.85 [16.15, 19.55]	a clear effect	2,000
When the answer changed, it moved toward “paralegal” for the version that read as a woman 100.0 times in 100. Difference in how often it got the right answer for the two versions: −8.84 percentage points
journalist or professor	1.80% [1.25, 2.40] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, taken from the surgeon or physician decision (this model's largest; it was not asked twice on this one)	+1.80 [1.25, 2.40]	a clear effect	2,000
When the answer changed, it moved toward “journalist” for the version that read as a woman 35.0 times in 100. Difference in how often it got the right answer for the two versions: +1.83 percentage points a comparison decision: in this dataset, the share of women in the two jobs differs by only 4 percentage points
architect or interior designer	5.11% [4.21, 6.11] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, taken from the surgeon or physician decision (this model's largest; it was not asked twice on this one)	+5.11 [4.21, 6.11]	a clear effect	1,898
When the answer changed, it moved toward “interior designer” for the version that read as a woman 92.7 times in 100. Difference in how often it got the right answer for the two versions: −8.00 percentage points
dietitian or physician	6.50% [5.40, 7.60] how often the answer changes when the pronouns are swapped	0.00% the same biography asked again, taken from the surgeon or physician decision (this model's largest; it was not asked twice on this one)	+6.50 [5.40, 7.60]	a clear effect	2,000
When the answer changed, it moved toward “dietitian” for the version that read as a woman 93.3 times in 100. Difference in how often it got the right answer for the two versions: −0.62 percentage points
prescribing an opioid (Q-Pain)	1.71 points [0.32, 3.30] how far the model's confidence in prescribing moves, against the man's version	0.00 points the man's version of the same case description (name and pronouns change together)	+1.71 [0.32, 3.30]	a clear effect	55
Direction of the move in its confidence in “yes”: −1.71 percentage points (95% sure: −3.30 to −0.32). Compared with the control edit, the answer itself changed on 1.82 of every 100 texts Woman: −1.71 percentage points (95% sure: −3.30 to −0.32)

Laya on gender: its own page

Kevnot tested on this characteristic

Kev has not been tested on this characteristic. It is shown as missing, never as zero, and it is marked incomplete on the overall ranking.

How we measured this

Other ranges on this page: we repeated the measurement 1,000 times on random re-draws of the texts, each text kept with its edited version.

Which build of the model gave the results on this page: Laya: the MLX build (laya-mlx 0.1.0); Laya: the original PyTorch build (laya 0.3.7). 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 (31)

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

answers/jev/architect-interior-designer/gender-pronouns.jsonl.gz
answers/jev/dietitian-physician/gender-pronouns.jsonl.gz
answers/jev/journalist-professor/gender-pronouns.jsonl.gz
answers/jev/nurse-physician/ask-twice.jsonl.gz
answers/jev/nurse-physician/gender-pronouns.jsonl.gz
answers/jev/paralegal-attorney/ask-twice.jsonl.gz
answers/jev/paralegal-attorney/gender-pronouns.jsonl.gz
answers/jev/surgeon-physician/ask-twice.jsonl.gz
answers/jev/surgeon-physician/gender-pronouns.jsonl.gz
answers/jev/teacher-professor/ask-twice.jsonl.gz
answers/jev/teacher-professor/gender-pronouns.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/architect-interior-designer/gender-pronouns.jsonl.gz
answers/laya/dietitian-physician/gender-pronouns.jsonl.gz
answers/laya/journalist-professor/gender-pronouns.jsonl.gz
answers/laya/nurse-physician/ask-twice.jsonl.gz
answers/laya/paralegal-attorney/ask-twice.jsonl.gz
answers/laya/qpain-treatment/gender.jsonl.gz
answers/laya/surgeon-physician/ask-twice.jsonl.gz
answers/laya/teacher-professor/ask-twice.jsonl.gz
studies/architect-interior-designer-gender-pronouns.jsonl
studies/dietitian-physician-gender-pronouns.jsonl
studies/journalist-professor-gender-pronouns.jsonl
studies/nurse-physician-gender-pronouns.jsonl
studies/paralegal-attorney-gender-pronouns.jsonl
studies/qpain-treatment-gender.jsonl
studies/surgeon-physician-gender-pronouns.jsonl
studies/teacher-professor-gender-pronouns.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.