Bias audit

How we approach bias.

The honest version: no AI screening system is bias-free. The right question is whether the system is structured enough that bias is measurable, reviewable, and correctable. Voxxhire is built around that question.

What we do

• Apply the same structured rubric to every candidate for a given role.
• Score against the rubric using evidence quotes from the transcript, not pattern-matching on candidate identity signals.
• Require human review on every advancement decision.
• Capture transcripts, scorecards, reviewer notes, and decisions in one audit trail.
• Surface structured outcome data customers can use for their own bias audits (e.g. NYC Local Law 144).

What we do not do

• No facial analysis.
• No emotion detection.
• No personality prediction.
• No demographic inference from voice characteristics.
• No autonomous hiring decisions.

How we test

We test the scoring engine against known-stable transcripts to confirm that identical answers from differently-labelled candidate profiles produce consistent scores. This is a guard against drift, not a guarantee of zero disparate impact — disparate impact depends on the customer's candidate pipeline, role rubric, and reviewer behaviour, all of which Voxxhire does not control.

We are explicit about that scope. Bias-free hiring is not something any vendor can credibly claim.

NYC Local Law 144

For customers using Voxxhire for candidates resident in NYC, the customer is responsible for the candidate notification and annual bias audit required by Local Law 144. We provide:

• AI-disclosure copy you can show candidates pre-interview.
• Structured outcome data per role / time window suitable for an independent auditor.
• Documentation of the scoring methodology to share with the auditor.

What you do

The customer owns the rubric, the reviewer behaviour, and the final decision. Voxxhire surfaces the structured evidence; you and your team make sure the rubric is fair, the reviewers are calibrated, and the outcomes are audited.