Performance ratings and fairness
1. Performance ratings
Most workplaces try to incentivize employees to perform better by providing additional rewards to the highest-performing employees. To do that, the organization needs to evaluate the employees, and sometimes compare them against one another. That seems straightforward, but most employees do not produce output that can be easily quantified, so the performance appraisal process is inherently subjective.2. Fairness
This means that the rating that comes out of the process, such as "meets expectations", is prone to bias. Conscious bias is relatively rare. Most managers and leaders are not deliberately giving better performance ratings to one group over another in a biased or unfair way. On the other hand, everybody is subject to some measure of unconscious bias. Unconscious bias is a result of our brain's amazing ability to process information and make decisions without you having to stop and think about each one. Your brain uses rules of thumb, or heuristics, to assess choices, situations, and even people without your consciously thinking about it. When those automatic assessments prefer one type of person over another, that is bias.3. Bias in the workplace
Choosing whom to hire, whom to promote, and how to rate employee performance are especially prone to bias. Studies in the US using various versions of resumes found that resumes with references to the candidate's race were far less likely to be called for an interview. Women's performance reviews receive 40% more critical subjective feedback, as opposed to positive or objective feedback. In this chapter's case study, you will be analyzing performance ratings and checking whether, in this sample dataset, women are significantly less likely to be labeled a high performer than men. The "high performer" label matters more than the variations in performance because those that are considered high performers are more likely to receive promotions, bonuses, and raises.4. The analyst's role in HR analytics
This dataset was constructed for educational purposes and is meant to illustrate the ways you could check for various types of bias in a workplace, including bias based on age, race, or other demographic traits. As people data scientists, we assume that any two demographic groups would perform similarly when accounting for understandable differences, such as different levels of on-the-job training. When a dataset shows one demographic group with lower performance or higher turnover than another demographic group, you must not conclude that the difference is because of their demographic traits. Rather, you must determine what additional factors - including the possibility of bias - could be influencing the results you see. When doing these types of analyses with real data, I strongly recommend meeting with internal legal counsel first. They will help ensure you handle the analysis and the results appropriately.5. Let's practice!
Now it's up to you to analyze the performance data for any signs of bias.Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.