| ▲ | godelski 4 hours ago | |||||||
Often these rules are in place because they are statistically correct.What needs to be understood is that no rule can be so well written that there are no exceptions. Rules are guides. Understanding this we can understand why certain guidelines are created, because they are likely the right response 9/10 times. This is especially important when dealing with high stress and low information settings. BUT being statistically correct does not mean correct. For example, if the operator had information about the ETA of the ambulance (we don't know this!) then the correct answer would have been to tell them to not wait. But if the operator had no information, then the correct decision is to say to wait. The world is full of edge cases. This is a major contributor to Moravec's paradox and why bureaucracies often feel like they are doing idiotic things. Because you are likely working in a much more information rich environment than the robot was designed for or the bureaucratic rules were. The lesson here is to learn that our great advantage as humans is to be flexible. To trust in people. To train them properly but also empower them to make judgement calls. It won't work out all the time, but doing this tends to beat the statistical rate. The reason simply comes down to "boots on the ground" knowledge. You can't predict every situation and there's too many edge cases. So trust in the people you're already putting trust into and recognize that in the real world there's more information to formulate decisions. You can't rule from a spreadsheet no more than you can hike up a mountain with only a map. The map is important, but it isn't enough. | ||||||||
| ▲ | kelnos an hour ago | parent [-] | |||||||
This was exactly what I was thinking (though less eruditely) when I was reading the blog post. In this particular case, waiting for the ambulance led to a worse outcome, but I would not be surprised that, statistically, a you're better off waiting for the ambulance than trying to get to the hospital via other means. But unfortunately: > if the operator had information about the ETA of the ambulance (we don't know this!) then the correct answer would have been to tell them to not wait. But if the operator had no information, then the correct decision is to say to wait. I expect the operator just is not allowed to give advice like that, even if they did have information on ambulance ETA. There could be liability if someone is advised to drive to the hospital, and something bad happens. Even if that bad thing would have happened regardless. I think that's a bad reason to do the situation-dependent incorrect thing, but that's unfortunately how the world works sometimes. | ||||||||
| ||||||||