Why bias audits matter
A tool can treat every candidate by the same rules and still produce uneven outcomes if those rules correlate with a protected characteristic. An audit measures outcomes, which is how such effects are found. It is required for automated employment decision tools in New York City and is good practice anywhere.
How an audit is calculated
The example uses made-up numbers. The impact ratio divides each group's selection rate by the highest group's rate. Lower ratios point to outcomes worth investigating; the law requires the figures to be published, not that a particular threshold is met.
| Category | Applied | Advanced | Selection rate | Impact ratio |
|---|---|---|---|---|
| Group A | 200 | 50 | 25% | 1.00 |
| Group B | 160 | 32 | 20% | 0.80 |
What an auditor needs
- A record of each decision the tool contributed to, with the outcome.
- Demographic data where candidates provided it, or test data where historical data is insufficient.
- A description of what the tool assesses and how its output is used.
- Dates: when the tool was first used, and the period covered.
How SynergicHire supports bias audits
SynergicHire keeps an append-only activity log, selection rates, and a row per decision for an independent bias audit, and the export itself is audited. Because every outcome is recorded by a named person, the data shows which decisions were made, by whom and when.
Questions
Does SynergicHire perform the bias audit?
No. The law calls for an independent auditor. SynergicHire provides the exports an auditor works from.
How often is a bias audit needed under Local Law 144?
The audit must have been done no more than one year before the tool is used, so in practice yearly.
What is the four-fifths rule?
A long-standing US guideline that treats a selection rate below 80% of the highest group's as a sign of possible adverse impact. Local Law 144 asks for impact ratios to be published rather than setting that threshold. This is general information, not legal advice.