I work in enterprise data & analytics, and my company recently did a watch party & discussion of Coded Bias, which really brought the issues around this to the attention of folks who may not have considered it before.
There’s a bit in it describing an issue at IBM - after it was pointed out in one of their ML projects, they actually did some work to address it, including some tests to actually verify results to confirm the particular bias(es) in question were no longer skewing results. That was very cool to see!
Of course, there are tons of possible things to look at, so it needs to be an ongoing process.
