By: Dave Schmidt and Sarah Layman
Federal focus on broadly understanding, regulating, and balancing the benefits and risks related to the use of artificial intelligence (AI) continues to grow; increasingly, there is activity on this front related to the use of AI-based tools in the employment sphere (see prior DCI blogs on this). The latest example of this is the “No Robot Bosses Act of 2023” submitted by Senators Bob Casey (D-PA) and Brian Schatz (D-HI) , on July 20, 2023. Citing the need for oversight and safeguards related to automated decision systems (ADS) driven by complex (opaque-box) algorithms, rather than humans, in the context of employment decisions (e.g., hiring, firing), this bill would require employers to disclose when they are using automated systems to make employment decisions and how they are being used. The bill would also:
- Prohibit exclusive reliance on an ADS for employment decisions
- Require testing/validation of the ADS for issues (e.g., discrimination, bias) before use and periodically after implementation
- Require training of individuals on the proper use of the ADS
- Require independent, human oversight
- Require disclosure of the use of an ADS, inputs, outputs, and employee rights
Further, this bill would seek to establish a “Technology and Worker Protection Division” within the Department of Labor to regulate this area. More details on this can be found in a fact sheet from the senators.
At this junction, this is only a proposed bill, and there are a number of facets to this that would require clarification; however, it is an example of the latest federal activity related to regulating the use of automated/AI systems in employment decision-making.
With movement on this front at the federal, state, and local levels, employers using AI, complex algorithms, or automated mechanisms associated with employment decisions need to be thinking proactively about being able to comply with present and future requirements. To this end, we encourage employers to consider proactive self-audits and/or third-party audits. Various contextual factors will drive audit specifics, and common considerations include the extent to which tools are:
- Evaluated for potential adverse impact using a comprehensive framework
- Using appropriate and job-relevant inputs to the algorithm
- Effectively predicting important, job-related outcomes
- Explainable and transparent, or minimally semi-transparent
- Compliant with prevailing professional and legal (e.g., Uniform Guidelines on Employee Selection Procedures) guidance where feasible
Stay tuned for more updates in this ever-evolving space, and feel free to reach out to DCI with any questions or audit requests you may have.