Tuesday, November 05, 2019

Police Shootings: Mortal Threats vs Tragic Mistakes

Police sometimes face mortal threats.  They also face innocent people whom they mistakenly judge to pose a mortal threat.  If too slow to react to the former, the police risk being killed.  If too quick to react to the latter, they risk killing innocent civilians.  What is the right way to balance these risks?  Three options present themselves:
(1) Give some extra weight to protecting innocent civilians, as per the duty to "protect and serve".
(2) Give some extra weight to their own life, as per the standard prerogatives of self-interest. OR
(3) Count both risks equally, and so seek to minimize innocent deaths (whether self or other) overall.

Though the choice between these options might make a slight difference at the margins, the amount of "extra" weight that could be justified in either of the first two options is presumably limited in size.  So I take it that police should give close to equal weight to either type of risk.  They should not, for example, kill someone who is very likely to be innocent.

In 'How many Police Shootings are Tragic Mistakes? How many can we Tolerate?' Christian Coons notes that (i) approximately 30 police are killed each year in the line of duty, whereas (ii) approx 600 people are killed by on-duty police each year.  How many of the latter constituted genuine mortal threats, and how many were "tragic mistakes"?  Coons notes that if assailants and police were equally likely to get in the first lethal shot, we could conclude that police collectively face approx 60 mortal threats per year. Adjusting for higher police firearm skill and body-armor (increasing police survival rate in the face of genuine threats) yields an estimate still below 150 mortal threats per year.

While an extremely rough estimate, this analysis suggests that the vast majority of police shootings are "tragic mistakes" rather than objectively justified responses to genuine mortal threats.  If police are making approx 450 "tragic mistakes" annually, while losing only 30 of their own number, this would seem to suggest a terrible miscalibration: the burden of risk is falling overwhelmingly upon the civilian populace, instead of being fairly shared between civilians and police alike.

The solution? Coons suggests that officers "should shoot only when a threat seems probable. On that view, if officers' judgments are fairly reliable then we should expect no more than half tragic mistakes. On the other hand, if officers' judgments about who is a threat are not at all reliable, then they shouldn’t be using lethal force at all. [...] Compliance with a probable threat standard minimizes net killings while valuing police and civilian lives equally."

I think that last sentence is mistaken, though. To see why, suppose that police were practically invincible: that attempts on their lives would almost invariably fail. In that case, shooting someone 49% likely to be innocent would clearly not value police and civilian lives equally. They could wait to be more certain of the threat, thereby reducing the risk to innocent civilians, without facing any significant increase in risk themselves. Of course, in reality the police are not so invincible. But Coons' model does assume that they are much more likely to emerge victorious from a gun fight. (Otherwise the estimated number of mortal threats would be much lower, and the number of "tragic mistakes" correspondingly higher.) So a moderated version of the same principle would seem to apply: police could decrease the risk to innocent civilians by more than the increase in risk to themselves, by adopting a standard that is more stringent than merely requiring a "probable" threat.

Presumably the way to count innocent lives equally, and minimize net killings, is instead to adopt whatever standard equalizes the number of police deaths and "tragic mistakes". In our current context, that might well be something closer to "beyond a reasonable doubt" rather than just a standard of "probable" threat.

One complication that isn't addressed here: The division between "mortal threats" and "innocent civilians" is a false dichotomy.  Police may often face aggressive criminals who aren't truly innocent, but neither do they pose a genuine mortal threat to armoured police officers.  While mere criminality doesn't mean that you deserve to die(!), one might nonetheless judge it reasonable for this population to bear a (significantly) greater share of the risk burden.  Anyone inclined towards that view would likely want to reserve the term "tragic mistake" for truly innocent victims, and use a more neutral label such as "unnecessary killings" for the broader category that Coons is talking about.  And we'd then need to know how many of these unnecessary killings had innocent vs criminal victims.  If overwhelmingly the latter, some might on that basis decide that the status quo was not so obviously unjust after all.  It'd take different arguments from those surveyed here to argue against such a position.


  1. Hi Richard,

    Regarding your scenario in which police are practically invincible, it seems to me that if the officers reckon (properly) that there is a 51% chance that a person is a threat, then what they reckon is that there is a 51% chance that that person has the means and willingness to harm them (perhaps lethally, though I do not think lethal threats are the only ones that should count, as the use of lethal force can be justified against nonlethal serious threats if lesser force is not available). In particular, the suspect has the means to pierce their almost invincible armor and harm them (maybe kill them, if you're only counting lethal threats). If that is not what the police reckon, then what is it to be an actual threat to the police officers, as opposed to not a threat?

    That aside, and regarding the weight of the burden, consider the following scenario: guns are restricted, and people who threaten the police with guns only carry real guns in about 1/3 of the time. Now, a police officer points a gun at a suspect and tells him to get on the ground. The suspect responds by pulling a gun himself, pointing it at the officer, and telling him to give him his gun, or else he'll shoot the officer in the head. Without any further relevant information, the police officer reckons there is a 1/3 chance that the gun is real. Handing over his own gun would guarantee the suspect would be a threat. Failing to do so without shooting result in at least a 1/2 chance (he reckons, let's say properly) of getting killed if the weapon is real. On the other hand, he's already pointing the gun at the suspect. If he pulls the trigger, chances are the suspect won't be able to shoot before getting hit - so he'll very likely miss the officer, who has a powerful enough gun. So, the police officer reckons there is a 1/6 chance of getting killed if he does not shoot, and a much lower chance if he does. Do you think the officer has an obligation not to shoot first?

    Granted, the scenario is not realistic (though neither is the "invincible" scenario), but it's meant to test the proposed general principles.

  2. Hi Angra, I was supposing that anyone attempting to kill the police, who had any non-zero chance of success, counts as an "actual threat" (even if their chance of success is very low). Now, if there's a 51% chance that a person is trying to kill the police (but with a less than 1% chance of success), and a 49% chance that they're completely innocent, then it would -- I claim -- not be moral or reasonable for the police to shoot first.

    Your case involving threats with fake guns raises interestingly different issues. There we suppose that the officer knows for certain that the agent is not "innocent" -- they are explicitly threatening to kill the officer, but the officer just isn't sure whether they really have the means to do so or not. In that case, I'm inclined to think that the criminal, by deliberately threatening the officer in this way, has forfeited the usual right to protection against being shot. So the officer is within their rights to shoot first.

    This example doesn't really speak to the principles that are addressed to cases involving potentially innocent persons, however.

  3. Hi Richard,

    Regarding what counts as an actual threat, I wonder then whether there is a misunderstanding between you and Coons. Are you sure he means that any nonzero chance of success in killing the police officer or a third party counts?

    In re: the other example, I agree it does not work in the case of innocents, but the example was not meant to be about innocent people, because the proposed solution by Coons is that police "should shoot only when a threat seems probable". The example is meant to show that that is not true. And it was also meant to challenge the 'beyond a reasonable doubt' alternative for the same reason. The problem, I think, is people who, while not posing an actual threat, aren't innocent civilians either (it's the "false dichotomy" problem you mention).

    Now, if your proposal (or his) is meant to operate only in the case that the person is otherwise innocent, then the counterexample does not work. However, in that case, an objection would be that Coons provides no data about the number of innocent civilians who get killed.

    Maybe a solution would be to argue that police should not fire without evidence beyond a reasonable doubt that the person they are firing at is not an innocent civilian, but without arguing that they have to have evidence that the person is probably an actual threat in order to fire (e.g., if someone claims to have a bomb vest, the police don't need over 50% probability that it's real to shoot him: as long that there is an immediate risk to other people conditioned to the bomb being real and there is a non-negligible chance it's real, I think it's okay to kill him in order to stop him).

    1. Well, even one who committed some prior crime may nonetheless still be innocent of threatening anyone in the present context. So I think the correct principle is rather that the police should not fire without sufficient evidence (beyond a reasonable doubt) that their target is actually threatening anyone. The relevant sense of "threat" here includes bluffing threats, but excludes someone who earlier committed a crime, now looks kinda scary to the cops, but isn't actually dangerous or claiming to be dangerous.

    2. Yeah, I understood "innocent civilians" as innocent at the time, even if they had a criminal record, and was going for a necessary condition, not a sufficient one. But the principle you propose now seems plausible to me, though the matter can be complicated when it comes to suspects who flee when told to stop and who appear to be a significant threat to the public given previous actions. The problem is that they may well not be threatening anyone during the chase, but if allowed to escape, they seem to be a significant threat. In those cases, maybe it's okay for the police to shoot at the legs or something like that, if they have no tasers or other means to stop them, but it's not okay to shoot to kill.

    3. There is still the question of when lethal force is justified when the person they target is threatening them or someone else. For example, it would not be justified if the police officers have evidence beyond a reasonable doubt that their target is bluffing and poses no actual threat, but there might be other cases.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.