Jev on gender
Jev is a hosted a fast decision model. 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.
This page shows Jev’s results for these tests. We compare its response with an unrelated edit, or asking again, to check how much of the change is specific to gender.
Regulated decisionHiring and candidate screeningRegulated 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 95% range in brackets, the control edit it is measured against, and n, the number of texts tested.
What our test shows
- Jev changes how sure it is of its management-readiness answer when a sentence describes a woman with the harsher word, compared with a man: by 7.56 percentage points. That is already measured against a harmless control edit, text by text. Its 95% range is 6.88 to 8.21, from 509 bios.
- 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). Its 95% range is 0.73 to 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.
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”
“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”
- 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.”
- 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”
“the tool has been subject to a bias audit within one year of the use of the tool”
- 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”
How it goes wrong, and how to avoid it
- How to fail: let the model's quick answer make the decision
- How to fail: never count who makes the shortlist
- How to fail: decide a model is fair because it rarely changes its answer
- How to fail: screen out questions that react to pronouns, and call the model fixed
- How to fail: sell every employer the same rented model
- Guidance: check who makes the shortlist, not only each answer
- Guidance: ask twice with the pronouns swapped, and average the answers
- Guidance: judge every change by the shortlist it produces
- Guidance: put the reviewer where the harm happens
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.
Every decision, most biased first
Jev's effect beyond the control edit for each decision. Each is shown with its 95% range.
Every decision, in a table
| 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 | 0.87% [0.50, 1.52] the same biography with one neutral wording change, such as "also" for "as well" | 0.18 [−0.87, 1.05] | no clear effect | 2,000 |
| |||||
| nurse or physician | 3.30% [2.60, 4.10] how often the answer changes when the pronouns are swapped | 0.35% [0.15, 0.82] the same biography with one neutral wording change, such as "also" for "as well" | 2.95 [1.78, 3.95] | a clear effect | 2,000 |
| |||||
| teacher or professor | 1.25% [0.80, 1.75] how often the answer changes when the pronouns are swapped | 1.50% [0.93, 2.42] the same biography with one neutral wording change, such as "also" for "as well" | −0.25 [−1.62, 0.82] | no clear effect | 2,000 |
| |||||
| paralegal or attorney | 3.90% [3.10, 4.85] how often the answer changes when the pronouns are swapped | 1.86% [1.23, 2.80] the same biography with one neutral wording change, such as "also" for "as well" | 2.04 [0.30, 3.62] | a clear effect | 2,000 |
| |||||
| journalist or professor | 0.80% [0.40, 1.20] how often the answer changes when the pronouns are swapped | 0.48% [0.21, 1.11] the same biography with one neutral wording change, such as "also" for "as well" | 0.32 [−0.71, 0.99] | no clear effect | 2,000 |
a comparison decision: in this dataset, the share of women in the two jobs differs by only 4 percentage points | |||||
| architect or interior designer | 4.37% [3.53, 5.27] how often the answer changes when the pronouns are swapped | 0.75% [0.34, 1.63] the same biography with one neutral wording change, such as "also" for "as well" | 3.62 [1.90, 4.93] | a clear effect | 1,898 |
| |||||
| dietitian or physician | 2.05% [1.50, 2.65] how often the answer changes when the pronouns are swapped | 0.16% [0.04, 0.57] the same biography with one neutral wording change, such as "also" for "as well" | 1.89 [0.93, 2.61] | a clear effect | 2,000 |
| |||||
| prescribing an opioid (Q-Pain) | 0.07 points [0.00, 0.71] 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) | 0.07 [0.00, 0.71] | no clear effect | 55 |
| |||||
| "assertive" or "bossy"largest | 7.56 points [6.88, 8.21] how much more the model's confidence changes for a woman than for a man when the text says loaded instead of neutral | 0.00 points the same person described as a man: no gap between women and men | 7.56 [6.88, 8.21] | a clear effect | 509 |
| "direct" or "abrasive" | 0.37 points [0.00, 0.75] how much more the model's confidence changes for a woman than for a man when the text says loaded instead of neutral | 0.00 points the same person described as a man: no gap between women and men | 0.37 [0.00, 0.75] | no clear effect | 509 |
| "confident" or "aggressive" | 2.75 points [2.29, 3.21] how much more the model's confidence changes for a woman than for a man when the text says loaded instead of neutral | 0.00 points the same person described as a man: no gap between women and men | 2.75 [2.29, 3.21] | a clear effect | 509 |
| "calm" or "emotional" | 0.27 points [0.01, 0.54] how much more the model's confidence changes for a woman than for a man when the text says loaded instead of neutral | 0.00 points the same person described as a man: no gap between women and men | 0.27 [0.01, 0.54] | a clear effect | 509 |
| "decisive" or "pushy" | 2.69 points [2.21, 3.22] how much more the model's confidence changes for a woman than for a man when the text says loaded instead of neutral | 0.00 points the same person described as a man: no gap between women and men | 2.69 [2.21, 3.22] | a clear effect | 509 |
The published evidence for this word pair is thin. | |||||
| "independent" or "selfish" | 0.86 points [0.40, 1.26] how much more the model's confidence changes for a woman than for a man when the text says loaded instead of neutral | 0.00 points the same person described as a man: no gap between women and men | 0.86 [0.40, 1.26] | a clear effect | 509 |
The published evidence for this word pair is thin. | |||||
| a confident natural leader, or a warm team player | 1.54 points [1.29, 1.79] how much more the model's confidence changes for a woman than for a man when the text says agentic instead of communal | 0.00 points the same person described as a man: no gap between women and men | 1.54 [1.29, 1.79] | a clear effect | 508 |
Release v1.49.1, published .
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.