The Equal Employment Opportunity Commission demonstrated its increasing focus on artificial intelligence at a public meeting January 31. The hearing featured 12 witnesses — academics and worker advocates — who stressed AI’s risks and advocated for more regulation. No representatives of AI developers participated in the hearing.
The panelists warned that:
- AI perpetuates existing biases.
- Current laws do not address concerns about AI.
- Policies requiring third-party audits of AI tools lack standards for the auditors.
- Tests that evaluate AI tools have limitations.
- Demonstrating the validity of selection criteria and fairness measures in AI tools is challenging
- There is a lack of accountability in AI tools managed by third parties.
The panelists recommended:
- Requiring AI creators to conduct validation testing, do disparity assessments, and conduct ongoing monitoring.
- Encouraging developers to use transparent models.
- Requiring employers to demonstrate the AI tool’s validity and prove it was built using representative and unbiased data.
- Prohibiting selection procedures that pose a high risk of discrimination.
- Requiring selection procedures to measure only essential job functions.
- Strengthening recordkeeping requirements.
- Making vendors work with certified industrial/organizational psychologists to develop selection procedures.
- Having the EEOC participate in certifying third-party vendors that offer audits.