We’re rapidly entering a stage where ML-based techniques are becoming better performing at simple, easily defined problems than humans are (the challenge in HR analytics arguably is not the ML, but the fact that few problems are easily defined). We also know that individuals respond to incentives and behaviours in ways that aren’t always rational (in the economic sense). This understanding is what drives behavioural economics and the ‘nudge’ industry.
So should we be surprised that ML can potentially out-perform humans at identifying these non-rational behaviours? The research documented in the first article for me was expected, though how powerful it can be was a surprise.
Ethics as it touches data, ML and the workforce is a complex topic.