AI kool-aid consumption: naivety and hubris?


The ability of AI technology and automation to discriminate at scale is well documented. Just this week we have seen both IBM and Amazon back-peddle on face recognition technologies. The dismal record of corporate recruitment sites for accessibility has been my hobby horse for some time. We read time and again of cases where HR technology discriminates against age, gender, race, disability, religion, or class background. We see software built for US regulations applied without much thought on workforces in other jurisdictions.

I read the recent Mercer survey report, and I was surprised to see:

67% of HR leaders are confident that they can ensure that AI and automation are institutionalizing bias.

Mercer report 2020

I’d say that the other 33% are right. We simply can’t be confident today that AI and automation aren’t institutionalizing bias. If the last two weeks have taught us anything, we have layers and layers of institutional bias that require serious attention. If we are not very careful AI and automation will make the problems worse, not better. HR has clearly not solved for analogue bias in the workplace, so they need to start there. When I last looked, assessing algorithmic fairness wasn’t in the skills profile of your typical HR leader, but again then most HR organizations don’t have accurate skills profiles for the workforce….