I think you raise good points here. We have unclear statistics on the issue and that leads to unclear ideas on what to do about it.
I want to offer one possible confounder when taking the raw numbers of 5 out of 30 officers as a random sample: like-minded coworkers are more likely to be working together than those that aren't. So if Bad Apple #1 was present, it's likely to find Bad Apples #2-5 present too. That's an optimistic reasoning for how many were there, but I think it's worth considering.
"Of course, this approach is riddled with limitations. It assumes that the 5 officers present were a random reflection of the SCORPION unit in general. "
In hindsight, the section should have led with that, not finished with it.
I think some would say that people who would stand aside and not intervene in current conditions can be conditioned to intervene - that this is not some overriding principle that they never intervene but some thin balance of incentives that can easily be shifted in most. (This is obviously related to the debate about how many Russians support Putin's politics and how many of them would stop doing this if they weren't literally programmed to by television and similar sources.)
To expand on Lumberheart’s point: I disagree that the example of Scorpion and other bad police departments somehow invalidates the “few bad apples” argument. What it invalidates is the “few bad apples being randomly distributed among police departments” argument. Or rather, it shows that the latter argument is insufficient to account for all police brutality cases.
We have to take into account confounders like clustering effects. It seems that police culture is pretty important to behavior, and so it would be reasonable to believe that one department with a bad policing culture might be full of cops who would mostly all stand by or participate in brutality, whilst another department might be mostly full of officers who would do the opposite.
An additional problem is sampling bias, since we don’t really hear stories (or at least they don’t get the same traction) of good police interactions. For all we know, only 5 out of 100 police departments are rotten, and the actions of their officers attract outsized attention. Additionally, the “bad culture” theory isn’t mutually exclusive to the “few bad apples that are randomly distributed” theory. It could be a mixture of both effects. Perhaps some departments possess a bad culture which instills bad behavior, and perhaps there are also a small minority of sadist cops who are part of both ‘good’ and ‘bad’ departments, and who would behave atrociously regardless of police culture.
As for sampling bias- keep in mind that there's a good chance we would hear about dramatic stories like a group of cops trying to beat someone up, and getting stopped in the middle of it by another group of cops.
Hmm I’m not so convinced. A few things we need to take into account:
1. Sure that would definitely make the news, but would it make as a big a splash as police brutality cases? The Tyre Nichols case engendered a huge media presence; almost everyone is aware of it and has commented on it. It’s hard to believe a group of cops intervening to prevent police brutality would achieve quite that level of media coverage.
2. It’s not like bad cops are stupid, most of them likely won’t instigate brutality in front of fellow cops if they didn’t already believe they could mostly get away with it. So that would create a deliberate self-selection effect, which would lower the rate of the incidents you describe.
3. When cops are successful in preventing brutality, by definition that will likely mean that little to no brutality has occurred, in effect making the case seem more “mild”. For example, here’s a video of fellow police officers preventing another cop from hitting an arrested woman (https://m.youtube.com/watch?v=vOKhqabg-Ys).
It’s plausible to think that in different circumstances, that women might have undergone a severe beating. But because the incident was prevented relatively quickly, it comes off as a case of police officers intervening in a relatively “mild” case, and therefore doesn’t attract the same media attention.
Some hold we can make the police better, some hold that.... the underlying system that the police are being asked to police has simply become too fractured, polarized, chaotic, hopeless, etc etc etc to be policed by people who are trained the way they are. Police officers are essentially being set up for failure, and no one has even considered that possibility, because similar to how police officers aren't very good at policing, intellectuals are not very intellectual, *on an absolute scale*.
I would challenge the formulation of the question - I have a view of human nature that's either more pessimistic or more optimistic, depending on how you look at things.
I think very few (<5%) of the population is truly sadistic, in the sense of taking pleasure in the undeserved suffering of others. Those are the kind of people I think are bastards. There may be a disproportionate number of them in the police force, but not enough to really explain outcomes like this.
However, I think the majority of the population (perhaps even as much as 95%,) derives some satisfaction from the suffering of others, if they think they deserve it. Whether or not these people (which likely includes all of us, unless you're on the extreme end of the pacifism spectrum) are "bastards" mostly depends on who they've been conditioned to think "has it coming".
So I feel like it really does depend on who they're beating to death - if the cop is attacking another cop, there's no way his buddies wouldn't intervene. If it's some "upstanding citizen", (white, middle class, educated) I'm confident they'd quickly intervene, even if the guy's been accused of a horrific crime - that's just not how we do things!
But if it's the kind of person we think of as the typical victim of police brutality (non-white, poor, uneducated), I think there's a decent chance they wouldn't intervene - there's all sorts of rationalisations to explain why this person deserves it, why this is a proportionate use of force, why this is the only way to deal with "those sort of people", ect.
I'm not sure how this really affects the conclusion though. On the reform side, you could argue that you just need to add training and accountability to the system - unless we think there's something inherently different between the US and UK populations, or the US and Scandinavian populations for that matter, there's no reason to expect that with the right institutions and culture in place within the police force, we can make disproportionate use of force less acceptable, and therefore less likely to happen.
On the abolition side, you can argue that the way the system is set up will corrupt the perspective of even officers with the best of intentions, and the whole thing needs to be got rid of and replaced with something else. I guess I take the approach that I haven't encountered any proposals that seem like they could effectively deal with violent crime, I think the ideas with the most potential in the US context (short of some hard to achieve cultural change) are just removing the police from enforcing drug and traffic crimes. Unfortunately, I think that the desire for retribution is pretty deep-rooted in the human psyche, and I'm not sure if an alternative organisation wouldn't face the same issues as the existing system - there are lots of alternatives to police throughout history, and few of them seem like an improvement on the status quo in the USA.
I think your idea of a statistical model is interesting! Let’s explore it a little. Things we need to define:
-What counts as an observation (data point)? Any police-civilian interaction? Any interaction that makes the news?
-What counts as bad apple behavior? Brutality? Harassment?
-What is our model of human behavior? Will a bad apple always refrain from in bad apple behavior when a good apple is around, and always engage in bad apple behavior in the absence of good apples?
The probability exercises in the post are about just one observation, where that observation is an instance of brutality. Testing the Hypotheses that X% of apples are bad is asking the question “If 5 police officers were pulled at random from a pool where X% are bad apples, what is the probability that all 5 police officers are bad apples?” This leads us to reject hypotheses of low % of bad apples with high confidence! However, to Alex’s point, this ignores all of the cases where police brutality could’ve happened but didn’t.
On the other end of the spectrum, if we took as our sample A, the set of all police-civilian interactions where 5 police were involved, and defined bad apple behavior B as Instances of police-civilian murders, testing the hypothesis would be asking the question “If 5 police officers were pulled at random A times from a pool with X% bad apples, what’s the probability that B groups contained no bad apples.” I would guess this test would reject all but the most optimistic hypotheses, since there are way more interactions with 5 officers than there are police-civilian murders. However, maybe a model that assumes bad apples are people that would murder civilians every time a good apple isn’t present is setting the bar a little low.
All of this is to say, statistics is complicated, and the devil is in the details. If you are interested in learning more about using statistics to model police behavior, I recommend checking out the work of Roland Fryer. Here’s a great paper!
I think a lot of the disagreement between the left and right comes down to different beliefs about what the actual numerators (instances of police brutality) and denominators (police interactions) are. Both sides make mistakes. I think the left tends to overestimate the instances of truly heinous police brutality-- the belief that police are systematically murdering black men is hard to square with annual national police killings of unarmed black men in the teens. On the flip side, I think the right underestimates the pervasiveness of “low level discrimination,” in traffic stops and general leniency, which is still massively disruptive. This is consistent with Fryer’s findings.
I think you raise good points here. We have unclear statistics on the issue and that leads to unclear ideas on what to do about it.
I want to offer one possible confounder when taking the raw numbers of 5 out of 30 officers as a random sample: like-minded coworkers are more likely to be working together than those that aren't. So if Bad Apple #1 was present, it's likely to find Bad Apples #2-5 present too. That's an optimistic reasoning for how many were there, but I think it's worth considering.
Yeah absolutely, this is why I say:
"Of course, this approach is riddled with limitations. It assumes that the 5 officers present were a random reflection of the SCORPION unit in general. "
In hindsight, the section should have led with that, not finished with it.
I think some would say that people who would stand aside and not intervene in current conditions can be conditioned to intervene - that this is not some overriding principle that they never intervene but some thin balance of incentives that can easily be shifted in most. (This is obviously related to the debate about how many Russians support Putin's politics and how many of them would stop doing this if they weren't literally programmed to by television and similar sources.)
To expand on Lumberheart’s point: I disagree that the example of Scorpion and other bad police departments somehow invalidates the “few bad apples” argument. What it invalidates is the “few bad apples being randomly distributed among police departments” argument. Or rather, it shows that the latter argument is insufficient to account for all police brutality cases.
We have to take into account confounders like clustering effects. It seems that police culture is pretty important to behavior, and so it would be reasonable to believe that one department with a bad policing culture might be full of cops who would mostly all stand by or participate in brutality, whilst another department might be mostly full of officers who would do the opposite.
An additional problem is sampling bias, since we don’t really hear stories (or at least they don’t get the same traction) of good police interactions. For all we know, only 5 out of 100 police departments are rotten, and the actions of their officers attract outsized attention. Additionally, the “bad culture” theory isn’t mutually exclusive to the “few bad apples that are randomly distributed” theory. It could be a mixture of both effects. Perhaps some departments possess a bad culture which instills bad behavior, and perhaps there are also a small minority of sadist cops who are part of both ‘good’ and ‘bad’ departments, and who would behave atrociously regardless of police culture.
Yeah, absolutely- see my response to Lumberheart.
As for sampling bias- keep in mind that there's a good chance we would hear about dramatic stories like a group of cops trying to beat someone up, and getting stopped in the middle of it by another group of cops.
Hi Philosophy bear,
Hmm I’m not so convinced. A few things we need to take into account:
1. Sure that would definitely make the news, but would it make as a big a splash as police brutality cases? The Tyre Nichols case engendered a huge media presence; almost everyone is aware of it and has commented on it. It’s hard to believe a group of cops intervening to prevent police brutality would achieve quite that level of media coverage.
2. It’s not like bad cops are stupid, most of them likely won’t instigate brutality in front of fellow cops if they didn’t already believe they could mostly get away with it. So that would create a deliberate self-selection effect, which would lower the rate of the incidents you describe.
3. When cops are successful in preventing brutality, by definition that will likely mean that little to no brutality has occurred, in effect making the case seem more “mild”. For example, here’s a video of fellow police officers preventing another cop from hitting an arrested woman (https://m.youtube.com/watch?v=vOKhqabg-Ys).
It’s plausible to think that in different circumstances, that women might have undergone a severe beating. But because the incident was prevented relatively quickly, it comes off as a case of police officers intervening in a relatively “mild” case, and therefore doesn’t attract the same media attention.
Just some things to keep in mind.
Some hold we can make the police better, some hold that.... the underlying system that the police are being asked to police has simply become too fractured, polarized, chaotic, hopeless, etc etc etc to be policed by people who are trained the way they are. Police officers are essentially being set up for failure, and no one has even considered that possibility, because similar to how police officers aren't very good at policing, intellectuals are not very intellectual, *on an absolute scale*.
I would challenge the formulation of the question - I have a view of human nature that's either more pessimistic or more optimistic, depending on how you look at things.
I think very few (<5%) of the population is truly sadistic, in the sense of taking pleasure in the undeserved suffering of others. Those are the kind of people I think are bastards. There may be a disproportionate number of them in the police force, but not enough to really explain outcomes like this.
However, I think the majority of the population (perhaps even as much as 95%,) derives some satisfaction from the suffering of others, if they think they deserve it. Whether or not these people (which likely includes all of us, unless you're on the extreme end of the pacifism spectrum) are "bastards" mostly depends on who they've been conditioned to think "has it coming".
So I feel like it really does depend on who they're beating to death - if the cop is attacking another cop, there's no way his buddies wouldn't intervene. If it's some "upstanding citizen", (white, middle class, educated) I'm confident they'd quickly intervene, even if the guy's been accused of a horrific crime - that's just not how we do things!
But if it's the kind of person we think of as the typical victim of police brutality (non-white, poor, uneducated), I think there's a decent chance they wouldn't intervene - there's all sorts of rationalisations to explain why this person deserves it, why this is a proportionate use of force, why this is the only way to deal with "those sort of people", ect.
I'm not sure how this really affects the conclusion though. On the reform side, you could argue that you just need to add training and accountability to the system - unless we think there's something inherently different between the US and UK populations, or the US and Scandinavian populations for that matter, there's no reason to expect that with the right institutions and culture in place within the police force, we can make disproportionate use of force less acceptable, and therefore less likely to happen.
On the abolition side, you can argue that the way the system is set up will corrupt the perspective of even officers with the best of intentions, and the whole thing needs to be got rid of and replaced with something else. I guess I take the approach that I haven't encountered any proposals that seem like they could effectively deal with violent crime, I think the ideas with the most potential in the US context (short of some hard to achieve cultural change) are just removing the police from enforcing drug and traffic crimes. Unfortunately, I think that the desire for retribution is pretty deep-rooted in the human psyche, and I'm not sure if an alternative organisation wouldn't face the same issues as the existing system - there are lots of alternatives to police throughout history, and few of them seem like an improvement on the status quo in the USA.
I think your idea of a statistical model is interesting! Let’s explore it a little. Things we need to define:
-What counts as an observation (data point)? Any police-civilian interaction? Any interaction that makes the news?
-What counts as bad apple behavior? Brutality? Harassment?
-What is our model of human behavior? Will a bad apple always refrain from in bad apple behavior when a good apple is around, and always engage in bad apple behavior in the absence of good apples?
The probability exercises in the post are about just one observation, where that observation is an instance of brutality. Testing the Hypotheses that X% of apples are bad is asking the question “If 5 police officers were pulled at random from a pool where X% are bad apples, what is the probability that all 5 police officers are bad apples?” This leads us to reject hypotheses of low % of bad apples with high confidence! However, to Alex’s point, this ignores all of the cases where police brutality could’ve happened but didn’t.
On the other end of the spectrum, if we took as our sample A, the set of all police-civilian interactions where 5 police were involved, and defined bad apple behavior B as Instances of police-civilian murders, testing the hypothesis would be asking the question “If 5 police officers were pulled at random A times from a pool with X% bad apples, what’s the probability that B groups contained no bad apples.” I would guess this test would reject all but the most optimistic hypotheses, since there are way more interactions with 5 officers than there are police-civilian murders. However, maybe a model that assumes bad apples are people that would murder civilians every time a good apple isn’t present is setting the bar a little low.
All of this is to say, statistics is complicated, and the devil is in the details. If you are interested in learning more about using statistics to model police behavior, I recommend checking out the work of Roland Fryer. Here’s a great paper!
https://scholar.harvard.edu/files/fryer/files/empirical_analysis_tables_figures.pdf
I think a lot of the disagreement between the left and right comes down to different beliefs about what the actual numerators (instances of police brutality) and denominators (police interactions) are. Both sides make mistakes. I think the left tends to overestimate the instances of truly heinous police brutality-- the belief that police are systematically murdering black men is hard to square with annual national police killings of unarmed black men in the teens. On the flip side, I think the right underestimates the pervasiveness of “low level discrimination,” in traffic stops and general leniency, which is still massively disruptive. This is consistent with Fryer’s findings.