Great work by you and your girlfriend! It takes courage to intervene in a situation like that, and skill to actually defuse it. Well done.
I don’t agree about what you’re calling the first error. Her job is to take in statements like yours, and output decisions. She could output “send police to ask questions”, or “send a SWAT team now”, or “do nothing”. She chose a decision you don’t agree with, but she had to choose some decision. It’s not like she could update the database with “update your prior to be a little more suspicious of Alexes in hatchbacks”.
I also don’t think it’s correct to call it arbitrary in the same way that the p < 0.05 threshold is arbitrary. I don’t really know how to say this clearly, but it’s like… the p < 0.05 rule is a rule for suspending human thought. Things you want to consider when publishing include: “what’s the false negative cost here? false positive cost? How bad would it be to spread this knowledge even if I’m not yet certain the studied effect is real?”. The rule “p < 0.05 yes or no” is bad because it throws all those questions away. It is arbitrary, like you say. But it doesn’t follow that any questionable decision was made by an arbitrary decision rule. If she thought about the things you said, and decided they didn’t merit sending anyone out to follow up, that isn’t arbitrary! All it takes to not be arbitrary is some thinking and some weighing of the probabilities and costs (and this process can be quick). You did that and came to one decision. She did that and came to another. That difference… seems to me… is a difference of opinion.
I don’t know the actual conversation you had with her, and it sounds like she didn’t do a very good job of justifying her decision to you, and possibly said obviously incorrect things, like “you have literally 0 evidence of any sort”. But I don’t think the step from “I think she was wrong” to “I think her decision rule is arbitrary” is justified. Reading this didn’t cause me to make any negative update on police department bureaucracy. (the security company is a different story, if indeed someone was there just watching!)
Great work by you and your girlfriend! It takes courage to intervene in a situation like that, and skill to actually defuse it. Well done.
Thank you :)
But I don’t think the step from “I think she was wrong” to “I think her decision rule is arbitrary” is justified.
To be clear, I’m not making that step because I think her output was wrong. It’s the way she went about it that made me think that her decision rule was problematic.
She treated things like guns and knives as “fact” and everything else as “opinion”. It sounded to me like because she categorized the car thing as an opinion, she was very quick to dismiss it, rather than pausing to consider how dangerous it actually is, and perhaps ask some follow up questions, perhaps regarding how fast and reckless the driving really was. On the other hand, if she had the mindset that it all counts as Bayesian evidence and her job is to try to judge how strong the evidence is, I think she would have spent more effort thinking about it.
I guess that’s the crux of it to me: giving things appropriate thought and not dismissing them. Thinking in terms of Bayesian evidence rather than fact vs opinion is a means to that end, but I think there are other means to that end. I even think fact vs opinion could be a means to that end, if you actually think carefully about whether the thing should be considered fact or opinion. But here, I don’t think that was happening. It sounded like a snap judgement.
It is possible that hard cutoffs like these are actually for the best. That if you give officers more freedom to make judgements, bad things will happen. However, I’d bet pretty strongly against it. I don’t have any experience in the field so I can’t be too, too confident, but I’d imagine that a 911 operator would be presented with highly, highly varied situations and would need to use a good amount of judgement to do the job. If you give that operator a list of things that “count” as fact/evidence, there are surely going to be things that you forget to include, if only because they are too obscure to imagine, such the situation here with trying to fling someone off the roof of your car like a bull.
Reading this didn’t cause me to make any negative update on police department bureaucracy.
I’m curious how you feel about the more punitive aspect of it. Ie. one reason to go after the guy is if you think he is a danger to society, but another reason is because he did something wrong and wrong things are supposed to be punished, either for punitive reasons or preventative ones. Personally, I updated pretty strongly towards thinking that people often get away with things like this. Previously I would have imagined that they’d come after you for something like this, but now it seems like that usually doesn’t happen, because of a) poor judgement and b) lack of resources.
Great work by you and your girlfriend! It takes courage to intervene in a situation like that, and skill to actually defuse it. Well done.
I don’t agree about what you’re calling the first error. Her job is to take in statements like yours, and output decisions. She could output “send police to ask questions”, or “send a SWAT team now”, or “do nothing”. She chose a decision you don’t agree with, but she had to choose some decision. It’s not like she could update the database with “update your prior to be a little more suspicious of Alexes in hatchbacks”.
I also don’t think it’s correct to call it arbitrary in the same way that the p < 0.05 threshold is arbitrary. I don’t really know how to say this clearly, but it’s like… the p < 0.05 rule is a rule for suspending human thought. Things you want to consider when publishing include: “what’s the false negative cost here? false positive cost? How bad would it be to spread this knowledge even if I’m not yet certain the studied effect is real?”. The rule “p < 0.05 yes or no” is bad because it throws all those questions away. It is arbitrary, like you say. But it doesn’t follow that any questionable decision was made by an arbitrary decision rule. If she thought about the things you said, and decided they didn’t merit sending anyone out to follow up, that isn’t arbitrary! All it takes to not be arbitrary is some thinking and some weighing of the probabilities and costs (and this process can be quick). You did that and came to one decision. She did that and came to another. That difference… seems to me… is a difference of opinion.
I don’t know the actual conversation you had with her, and it sounds like she didn’t do a very good job of justifying her decision to you, and possibly said obviously incorrect things, like “you have literally 0 evidence of any sort”. But I don’t think the step from “I think she was wrong” to “I think her decision rule is arbitrary” is justified. Reading this didn’t cause me to make any negative update on police department bureaucracy. (the security company is a different story, if indeed someone was there just watching!)
Thank you :)
To be clear, I’m not making that step because I think her output was wrong. It’s the way she went about it that made me think that her decision rule was problematic.
She treated things like guns and knives as “fact” and everything else as “opinion”. It sounded to me like because she categorized the car thing as an opinion, she was very quick to dismiss it, rather than pausing to consider how dangerous it actually is, and perhaps ask some follow up questions, perhaps regarding how fast and reckless the driving really was. On the other hand, if she had the mindset that it all counts as Bayesian evidence and her job is to try to judge how strong the evidence is, I think she would have spent more effort thinking about it.
I guess that’s the crux of it to me: giving things appropriate thought and not dismissing them. Thinking in terms of Bayesian evidence rather than fact vs opinion is a means to that end, but I think there are other means to that end. I even think fact vs opinion could be a means to that end, if you actually think carefully about whether the thing should be considered fact or opinion. But here, I don’t think that was happening. It sounded like a snap judgement.
It is possible that hard cutoffs like these are actually for the best. That if you give officers more freedom to make judgements, bad things will happen. However, I’d bet pretty strongly against it. I don’t have any experience in the field so I can’t be too, too confident, but I’d imagine that a 911 operator would be presented with highly, highly varied situations and would need to use a good amount of judgement to do the job. If you give that operator a list of things that “count” as fact/evidence, there are surely going to be things that you forget to include, if only because they are too obscure to imagine, such the situation here with trying to fling someone off the roof of your car like a bull.
I’m curious how you feel about the more punitive aspect of it. Ie. one reason to go after the guy is if you think he is a danger to society, but another reason is because he did something wrong and wrong things are supposed to be punished, either for punitive reasons or preventative ones. Personally, I updated pretty strongly towards thinking that people often get away with things like this. Previously I would have imagined that they’d come after you for something like this, but now it seems like that usually doesn’t happen, because of a) poor judgement and b) lack of resources.