I was told by three managers at IBM that the field sales layoffs beginning in October of 2016 continuing into March of 2017 were generated by WATSON (the AI) . Those impacted weren't able to Open Door their termination nor were their first or second line managers able to intercede. The question people should be asking is, "How was Watson trained?" Were classes protected? Were age, s-x, race, start date, etc. used for training? If not specifically used, were these fields masked from the AI? How was the AI prevented from being biased? What test did they use on the resulting data to determine if there was bias? Those that understand machine learning will recognize the potential for bias based on the training dataset. What safeguards did IBM apply to prevent unintentional bias, specifically age discrimination? S-x? Race?
Reposted from @JtAUPMv-64eoh for asking the right questions. I'd like to know the answers as well.