We should definitely expand on this in our own post, there's a lot of nuance behind what to automate and what the confidence levels must be on order to make it work. Specifically there's questions about fragility of systems that come up where the non-automated system has more slack to recover from mistakes or errors than the automated systems. One particularly bad trap is rolling out an insufficiently reliable automation but is reliable enough to remove the human oversight, here's a really good example of such a situation https://idiallo.com/blog/when-a-machine-fired-me
shuttle1987
commented
4 years ago
We should definitely expand on this in our own post, there's a lot of nuance behind what to automate and what the confidence levels must be on order to make it work. Specifically there's questions about fragility of systems that come up where the non-automated system has more slack to recover from mistakes or errors than the automated systems. One particularly bad trap is rolling out an insufficiently reliable automation but is reliable enough to remove the human oversight, here's a really good example of such a situation https://idiallo.com/blog/when-a-machine-fired-me
_Originally posted by @shuttle1987 in https://github.com/shuttle1987/CPS_content/pull/105_