This is a classic case of the “substitution myth” of automation – the idea that you can just replace humans with automation, one-for-one, without impacting the rest of the system. The author is right insofar as that focusing solely on operator effectiveness measures is not always the right strategy, and that many tasks could legitimately benefit from automation. But what he misses is that any reasonably complex automated system will itself have to include measures to support operator effectiveness. It’s not an either-or proposition. 30 years of research on automated systems have shown that if people have to monitor, maintain, and handle exceptions for automated systems (which they always do), then you need to design for effective coordination (e.g., displays to show what the automation is doing and why, and methods for intervening and re-directing the automation in exceptional situations, and training to allow operators to use them). The greater the complexity and scope of the automation (e.g., the “lights-out” variety), the more critical this need becomes. Automation doesn’t remove the need for human-centered design, it just shifts the target.