Fairness
Fair by design. A human decides.
Fairness is not a setting we toggle on. It is how the read is built. Here is what we do to reduce bias, what we deliberately do not do, and where we are honest about the limits.
We reduce bias. We do not pretend to erase it.
Language models can carry bias from the data they were trained on. Any hiring tool that tells you it has removed bias entirely is selling you something. We take the opposite position, and we build around it.
Slate reduces bias in three concrete ways. It constrains the screen to job-relevant evidence, so the read is about the work and not the person. It applies one consistent rubric to every applicant, so two similar candidates get similar reads. And it attaches a verbatim quote to every reason, so a person can audit the read rather than trust it blindly. Then it stops, and hands the decision to your team.
How the read stays fair
Job-relevant evidence only
Slate is instructed to screen on evidence tied to the role and to ignore personal characteristics. What is judged is the work shown, not who the person is.
One rubric for everyone
The same must-haves, read the same way, for every applicant. Two similar candidates get similar reads, whether they applied first or last.
Every reason is auditable
Each judgment cites a verbatim line from the application. You never have to take a read on faith, and you can review any decision the screen influenced.
Honest about uncertainty
When an application is thin, Slate says the read is limited rather than inventing confidence. Absence of evidence is a gap to verify, not a mark against someone.
A human makes the decision
Slate never auto-rejects and never decides. It makes the first read fair and fast and hands the judgment, and the accountability, to your team.
What Slate does, and does not do
What Slate does
- Screens only on evidence relevant to the role
- Keeps name, age, gender, race, school and employer prestige out of the read
- Applies one consistent rubric to every applicant
- Ties every reason to a verbatim line you can check
- Says plainly when the evidence is thin or unclear
What Slate does not do
- Never auto-rejects a candidate
- Never makes the hiring decision for you
- Does not score personality or predict a person
- Does not read protected characteristics or run a background check
- Does not claim to be free of bias
A decision-support tool, not a decision maker
Slate is a screening copilot. It produces a fair first read and surfaces evidence so your team can review every applicant, not just the first ten. It never makes the hiring decision, and it never auto-rejects.
Because a person makes the decision, that person is accountable for it. You remain responsible for lawful, fair hiring in your jurisdiction, and Slate is built to support that: every read is auditable, every reason is sourced, and the same rubric is applied across a pool so you can check for consistency. If you need records for a review, the reasons and the evidence export with the candidate.
Where we keep improving
Reducing bias is ongoing work, not a finished feature. We test for consistency across a pool, we constrain what the model is allowed to consider, and we would rather ship a read that says the evidence is unclear than one that invents confidence. If you find a case where the read is unfair, tell us, and we will look at it.
A fair first read, with the reasons in view.
Start free with one open role and see the evidence-based read for yourself.