Let’s say it’s normal to keep someone on pain meds for 4 to 8 days after surgery. Each day, you assess the patient and check a number of factors to determine when to stop pain meds, like: how much pain do they say they’re in? How much do they wince when they walk? How comfortable do they seem? Do they seem distracted when talking to you? Etc. Each of those assessments is subjective, and therefore can be influenced by biases you don’t even realize you have. Over a year, maybe that means you stop pain meds on the 5th day, on average, for Black patients, and on the 6th day for white patients. You’re not really withholding pain meds from any one patient. Each patient probably doesn’t really notice the difference. But over time, that slight difference compounds and adds up to poorer quality of care for one group.

This is why it’s so important to measure things like this subjectively, and look for and fix the reasons they’re happening. It’s very hard, probably impossible, to fix these issues by just assuming that well-meaning people will be able to be completely unaffected by bias. And sometimes people overcorrect: managers in tech are less likely to give Black employees critical feedback, for example, because they don’t want to be racist, and that behavior harms Black employees by not giving them opportunities to correct behavior that’s holding them back from advancement. Again, tiny behaviors that compound at scale.