Charting a Fair Path Forward on Big Data & Algorithms in the Workplace

Last week, CDT filed comments with the Equal Employment Opportunity Commission (EEOC) on the future of algorithms and big data in the workplace. Automated decision-making systems used in hiring often advertise their ability to improve productivity and make better judgments about potential employees than humans. But some of these systems may have unintended harmful consequences, such as perpetuating existing societal biases. To explore these issues, the Commission convened a meeting on the subject and called for comments on the effects of big data on equal employment opportunity.

Algorithms are designed to look for patterns in data, and those tasked with predictive hiring analysis may seek to replicate successful past hiring decisions in identifying job and promotion candidates. This bias toward past success likely favors historically advantaged groups who have had more opportunities to succeed in the workplace. Moreover, the trend toward collecting and analyzing large amounts of predictive data on existing employees threatens to erode important barriers between personal life and work life.

This bias toward past success likely favors historically advantaged groups who have had more opportunities to succeed in the workplace.

However, the use of algorithms for employment-related decisions may also potentially promote more diverse hiring and inclusive workplaces. For example, responsible algorithms can alert companies to unintended bias in their hiring processes or help them attract more diverse candidates.

CDT’s comments offer principles to help guide the Commission’s actions, both in support of the positive uses of algorithms in the workplace context and to mitigate the negative impacts of big data in the workplace. The principles are as follows:

• Algorithms must be consistent with employment and anti-discrimination law;
• Hiring models should be based on rigorous and wide-ranging research and logic;
• Big-data models should proceed with caution when attempting to statistically predict subjective attributes such as personality, emotional state, or overall value as an employee;
• Employers should not engage in any unnecessary data collection on, or tracking of, employees;
• Employee manuals should clearly state the employer’s policy with respect to the data it collects and algorithmically analyzes;
• Data analytics should be directed at institutions; and
• Good intentions aren’t enough when using data to diversify hiring.

We believe the Commission should incorporate these principles into its guidance for companies on the uses of big data in the workplace, as well as into their audits for companies on the algorithms they use for hiring, managing, evaluating, and promoting decisions.