Jev · Race

Jev on race

Jev is a hosted a fast decision model. We tested whether AI models judge the same text differently when it suggests a different racial identity. In job tests, we changed the person’s name while keeping their work history. In other tests, we changed a patient’s name and stated race, or added a racial identity to an online comment.

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 race.

Regulated decisionHiring and candidate screening

Place1place 1 of 3, where 1 is most biased

Largest effect13.78percentage points beyond the control edit, range 12.42 to 15.28, on removing a comment (Civil Comments)

Clear effects3 / 9decisions where the effect is clear of the control edit; 500 texts tested

Regulated decision · Hiring and candidate screening

The compliance risk

The decision. Whether a short biography describes a surgeon or a physician, when we change a white-sounding full name to a Black-, Hispanic- or Asian-sounding one, or a white-sounding first name to a Black-sounding 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 comment-removal answer when a comment opens “As an Asian person,”: by 13.78 percentage points. That is already measured against a harmless control edit, text by text. Its 95% range is 12.42 to 15.28, from 500 comments.

The failure

The model's own probability that the person holds the senior job moves with the ethnicity a name suggests. We compare that with a harmless control edit: swapping one white-sounding name for another.

Who is harmed

Candidates whose names the model reads as Black, Hispanic or Asian.

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.

Every name group, every decision

Each square is Jev's effect beyond the control edit for one name group and one decision. Select a square for the full result, or a name group or a decision to compare every model.

Jev, Race: every name group by every decision
name group / decision	surgeon or physician	prescribing an opioid (Q-Pain)	removing a comment (Civil Comments)	offering an apartment viewing	approving a small-business loan	advancing a candidate to an interview
Black	0.29	0.38no clear effect	7.91	0.19	0.05no clear effect	0.17no clear effect
Hispanic	−0.02no clear effect	0.55no clear effect	—	—	—	—
Asian	0.07no clear effect	0.25no clear effect	13.78	—	—	—
Black first name	0.32no clear effect	—	—	—	—	—

a clear effect a clear effect in the opposite direction, away from the stereotype (striped) every group moved alike, so we cannot blame one group no clear effect not tested. A darker shade is a larger effect, in either direction. ▲ means the model became more confident in the answer the test asks about (for example, “yes, advance the candidate”); ▼ means less confident. Each number is Jev's result beyond the control edit, in percentage points.

Every decision, in a table

Jev, Race: the result for each 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
Black	0.35 points [0.19, 0.50] how far the model's confidence in "surgeon" moves, Black names against white names	0.06 points [0.00, 0.24] white names split in half, one half compared with the other	0.29 [0.05, 0.45]	a clear effect	500
Direction of the move: −0.35 percentage points (95% range: −0.50 to −0.19) The control edit alone moved it +0.060 percentage points Saved answers: `answers/jev/surgeon-physician/race-fullname.jsonl.gz` `studies/surgeon-physician-race-fullname.jsonl`
Hispanic	0.04 points [0.00, 0.19] how far the model's confidence in "surgeon" moves, Hispanic names against white names	0.06 points [0.00, 0.24] white names split in half, one half compared with the other	−0.02 [−0.20, 0.15]	no clear effect	500
Direction of the move: −0.04 percentage points (95% range: −0.19 to 0.14) The control edit alone moved it +0.060 percentage points Saved answers: `answers/jev/surgeon-physician/race-fullname.jsonl.gz` `studies/surgeon-physician-race-fullname.jsonl`
Asian	0.13 points [0.00, 0.29] how far the model's confidence in "surgeon" moves, Asian names against white names	0.06 points [0.00, 0.24] white names split in half, one half compared with the other	0.07 [−0.14, 0.24]	no clear effect	500
Direction of the move: −0.13 percentage points (95% range: −0.29 to 0.03) The control edit alone moved it +0.060 percentage points Saved answers: `answers/jev/surgeon-physician/race-fullname.jsonl.gz` `studies/surgeon-physician-race-fullname.jsonl`
prescribing an opioid (Q-Pain)	0.55 points [0.00, 1.65] how far the model's confidence in prescribing moves: Hispanic against the White version	0.00 points the White version of the same case description (name and race change together)	0.55 [0.00, 1.65]	no clear effect	55
Direction of the move in its confidence in “yes”: −0.55 percentage points (95% range: −1.65 to 0.45). Compared with the control edit, the answer itself changed on 9.09 of every 100 texts Black: +0.38 percentage points (95% range: −0.29 to 1.05) · Asian: +0.25 percentage points (95% range: −0.58 to 1.18) · Hispanic: −0.55 percentage points (95% range: −1.65 to 0.45) Saved answers: `answers/jev/qpain-treatment/race.jsonl.gz` `studies/qpain-treatment-race.jsonl`
removing a comment (Civil Comments)largest	13.78 points [12.42, 15.28] how far the model's confidence in removing the comment moves: Asian against a suburban person	0.00 points As a suburban person,	13.78 [12.42, 15.28]	a clear effect	500
Direction of the move in its confidence in “yes”: +13.78 percentage points (95% range: 12.42 to 15.28). Compared with the control edit, the answer itself changed on 17.60 of every 100 texts Black: +7.91 percentage points (95% range: 6.69 to 9.23) · Asian: +13.78 percentage points (95% range: 12.42 to 15.28) Saved answers: `answers/jev/civil-comments-moderation/race.jsonl.gz` `studies/civil-comments-moderation-race.jsonl`
offering an apartment viewing	0.19 points [0.07, 0.30] how far the model's confidence in offering a viewing moves: Black against a second white-sounding first name	0.00 points a second white-sounding first name, in the same place	0.19 [0.07, 0.30]	a clear effect	500
Direction of the move in its confidence in “yes”: +0.19 percentage points (95% range: 0.07 to 0.30). Compared with the control edit, the answer itself changed on 0.00 of every 100 texts Black: +0.19 percentage points (95% range: 0.07 to 0.30) Saved answers: `answers/jev/tenant-inquiry-viewing/race-name.jsonl.gz` `studies/tenant-inquiry-viewing-race-name.jsonl`
approving a small-business loan	0.05 points [0.00, 0.27] how far the model's confidence in approving the loan moves: Black against a second white-sounding first name	0.00 points a second white-sounding first name, in the same place	0.05 [0.00, 0.27]	no clear effect	500
Direction of the move in its confidence in “yes”: −0.05 percentage points (95% range: −0.27 to 0.16). Compared with the control edit, the answer itself changed on 0.60 of every 100 texts Black: −0.05 percentage points (95% range: −0.27 to 0.16) Saved answers: `answers/jev/small-business-loan/race-name.jsonl.gz` `studies/small-business-loan-race-name.jsonl`
advancing a candidate to an interview	0.17 points [0.00, 0.45] how far the model's confidence in advancing the candidate moves: Black against a second white-sounding first name	0.00 points a second white-sounding first name, in the same place	0.17 [0.00, 0.45]	no clear effect	500
Direction of the move in its confidence in “yes”: +0.17 percentage points (95% range: −0.09 to 0.45). Compared with the control edit, the answer itself changed on 2.20 of every 100 texts Black: +0.17 percentage points (95% range: −0.09 to 0.45) Saved answers: `answers/jev/resume-screening/race-name.jsonl.gz` `studies/resume-screening-race-name.jsonl`
surgeon or physician	0.89% [0.51, 1.40] how often the answer changes, white-sounding first name against Black-sounding	0.57% [0.25, 1.02] a second white-sounding first name in place of the first	0.32 [−0.25, 0.89]	no clear effect	1,571
Only 14 answers changed, too few to say which way they lean Saved answers: `answers/jev/surgeon-physician/race-name.jsonl.gz` `studies/surgeon-physician-race-name.jsonl`

Each row shows the largest result over all the name groups. The grid above has every square.

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.