Section 1 - Description
Section 2 - Scope
Section 3 - Requirements
Section 4 - Penalties
Section 5 - Status
MA Bill 1873 is a comprehensive set of rules related to worker protections that covers four main pillars:
- Automated decision systems (ADS)
- Productivity tracking
- Electronic monitoring
- Worker data collection
For this guide we’ll consider how the bill approaches ADSes.
The bill targets employers with workplaces in Massachusetts who use automated decision systems (ADS), which it defines as a computational process used to make or assist decision making in employment. The scope of employment decisions is comprehensive and covers hiring, promotion, termination, setting wages, performance evaluation, and more. This bill sets requirements for the users of an ADS, but vendors who sell ADSes to employers are obligated to provide their customers with all information required for the customer to meet the requirements posed in this bill.
- Algorithmic impact assessments:
Employers who develop ADSes are required to complete algorithmic impact assessments (AIAs), which must be conducted by an independent auditor. AIAs must be completed at the beginning of the development or procurement of the system, and they must be updated when the system changes materially. Employers must retroactively complete AIAs for ADSes already in use.
Similar to other bias analysis requirements, the AIA must assess discrimination against protected classes (but the bill is not prescriptive on methodology). Employers must evaluate the error profile of the system in terms of false negatives and false positives. But the AIA must also evaluate AI for harms against workers’ legal and privacy rights, the risk of physical or mental harm, and other ways the ADS could have a negative impact on workers.
Additionally, the assessment must identify risks of the ADS and explain what actions will be taken to mitigate these risks.
- Notice requirements:
Employers must provide workers with advanced notice of at least 10 days before adopting an ADS. This notice must cover the nature and purpose of the system, the system outputs, the types and sources of input data, and who built and operates the system. Employers must also provide advance warning of changes to employees of any significant changes to an ADS or its usage.
- Inventorying ADSes:
Employers must maintain a list of all ADSes currently in use and annually notify employees of all active systems and the details of their usage.
- Banned ADS usages:
Employers may not use ADSes in certain manners, including if they violate labor laws, infringes on workers’ rights, or predict the likelihood of an employee using their legal rights. Facial recognition, sentiment prediction, and similar biometric systems are not allowed.
Additionally, employers may not use an ADS as the sole input to decisions related to hiring, promotion, termination, discipline, or setting wages. Such decisions are required to include meaningful human oversight and the employer is obligated to send the affected employee/candidate with details about how the ADS was used in their case.
- ADS for productivity monitoring:
Before using an ADS to monitor employee productivity, employers must also notify the department of labor standards with details about the nature and use of the algorithm and the number of employees affected. These systems must be reviewed by the department before they can be implemented.
Employers who fail to comply with the requirements related to ADS usage will be subject to fines of $2,500 to $20,000 per violation, depending on the specific violation.
This bill has not been passed yet. Once passed it will go into effect immediately.