In What Should be Automated?, Matti Tedre discusses the fundamental flaw in the debate around automation. The question should not be "what should be automated?" but instead "How can one automate things efficiently and reliably?", which then shifted to "why things should be automated?":

"since the 1980s the focus in computing research has been gradually broadening from the machine and automation toward how and where computers are used, the actual activities of end users, and how end users collaborate and interact (...) Neither the theoretician's question "What can be efficiently automated?" nor the practitioner's question "How can processes be automated reliably and efficiently?" include, explicitly or implicitly, any questions about why processes should be automated at all, if it is desirable to automate things or to introduce new technologies, or who decides what will be automated."

Why do I blog this? preparing a presentation about failed futures, including some elements about the problems caused by "automation". The author of the paper argues that the shift from what to how/why lead computing researcher to a situation where they really have to pay attention to " the needs, wants, hopes, expectations, wishes, fears, concerns, and anxieties that people have regarding technology".