Below, I publish a written Q&A between myself and the Pentagon about rules for targeting and battle damage assessments, including how civilians who provide support to a military force, and are killed in strikes aimed at that force, are counted.
Although I was addressing my questions to military lawyers, the exchange took place through Lt. Col. Valerie Henderson, a spokeswoman for the department, and she asked that its answers be attributed to her.
I was pursuing a theory about non-combatant deaths from airstrikes outside of conventional war zones that appears to be incorrect, so I do not see a general-readership news article here. But the exchange may nevertheless be of interest to scholars and specialists. In particular, note that the department is signaling that it may revise certain disputed sections of the Law of War Manual involving “assumption of risk” by civilians and proportionality analysis.
I am sincerely grateful to the Defense Department for taking the time to engage with me in such a detailed and serious way.
BUT FIRST, SOME BACKGROUND
When the Obama administration revealed the government’s official understanding of the scope of non-combatant deaths from airstrikes outside of conventional war zones — that is, drone strikes and other missile attacks in places like tribal Pakistan, Yemen, Somalia, and Libya — there was a broad reaction of puzzlement. The administration asserted that between 64 and 116 civilian bystanders had been killed in 473 such strikes between January 2009 and December 2015. But that seemed too low, based on what we think we know about such strikes. In particular, the entire low end of the range would seem to be eaten up by our understanding of civilian deaths caused by just two strikes: a December 2009 JSOC missile strike at an AQAP camp in Yemen that is believed to have killed 41 Bedouins from a tribe that turned out to be camped next door and was selling food to the terrorists, and a March 2011 CIA drone strike in Pakistan that killed between 26 and 48 people, most of whom are thought to have been villagers meeting to discuss a chromite mining dispute. [See Power Wars, Chapter Six (“Targeted Killings”), parts one (“Two Airstrikes in Yemen”) and ten (“Signature Strikes”).]
At the same time, I had been tracking a dispute over some of the targeting rules in the Pentagon’s new Law of War Manual, including complaints from various legal scholars that it appeared to improperly exclude various categories of civilians from being weighed in proportionality analysis by commanders considering a strike at a military target where civilians could be killed too, like civilians supporting a military force by working in a munitions plant. See this article about a recent revision to the manual addressing a controversial section about journalists in combat zones, and how it did not address the proportionality issues scholars were most concerned about.
These two storylines raised in my mind a possible explanation for the low end of the range of non-combatant deaths: perhaps, I speculated, Central Command had excluded the adult members of that Bedouin tribe killed in December 2009 from its count of non-combatant deaths outside war zones because the tribespeople had been providing support services to the enemy force and thus were being treated more like military contractors than innocent civilians? That approach would be controversial but at least it would make the low-end of the range understandable. That prompted me to ask some questions to the Defense Department, which the Office of the General Counsel preferred to answer in writing. I asked a follow-up round, and they replied again. Here is the exchange:
ROUND ONE FROM ME TO DOD:
I have some questions about how DOD defines civilian casualties during battle damage assessments in Centcom’s area of responsibility, which obviously may include JSOC in certain contexts. Would it be all right if I spoke to your top JAGs about them?
Q: In terms of ex ante targeting decisions, does this mean that my example above (the food services providers) would not count for proportionality analysis?
A: No, the food service providers in your example would be considered in a proportionality analysis and other analyses related to collateral damage. As an initial matter, in your example, the Bedouin camp is “next to” a terrorist camp as opposed to “in or on” the terrorist camp, and it is not clear that the Bedouins know that the terrorist camp is susceptible to military attack as part of an armed conflict. More importantly, however, part of the principle of “proportionality” under the law of war is the obligation by the attacking force to take feasible precautions to reduce the risk of harm to civilians in conducting attacks (e.g., warnings, adjusting the timing of the attack), and section 5.11, which describes this requirement, would apply to the Bedouins, even if their camp was located in or on the terrorist camp. Further, as a matter of policy and practice, DoD applies procedures and standards (often much more restrictive than what would be legally required under the law of war) for reducing the risk of non-combatant casualties and other collateral damage, and individuals like the ones in your example generally would be “counted” in such analysis.
Q: In terms of post facto battle damage assessments, would DOD classify my example as “combatants” or “non-combatants?”
A: Non-combatants. Persons who are not part of an armed group and who do not take a direct part in hostilities, but who are providing support services to a non-State armed group, would be deemed to be non-combatants.
Q: Does anything change in the answers if a strike is inside an area of active hostilities versus outside such an area?
A: With regard to the issues presented in your first two questions, the answers we have provided apply as a matter of law whether a strike is conducted inside an area of active hostilities or outside such an area. However, as a matter of policy, outside areas of active hostilities the United States applies a standard in its counterterrorism operations that goes far beyond the requirements of proportionality; it applies a standard of “near certainty” that non-combatants will not be injured or killed. Therefore, outside an area of active hostilities, the attack would not be authorized unless there was near certainty that none of the Bedouins in your example would be killed or injured, whether they were next to or in the terrorist camp.
(Since my questions this round were longer, I interwove the answers directly)
Q: The answer to question one says the hypothetical Bedouin tribe would be weighed for pre-targeting proportionality analysis in part because the tribe is not camped out “in or on” the terrorist camp, but rather next to it. But I had read 220.127.116.11’s carve-out from proportionality analysis includes civilian support-providers who are merely in close proximity to a military target without actually being in or on it. The section says: “Moreover,the law of war accepts that the defender may employ these persons to support military operations near or within military objectives. If these persons could have the effect of prohibiting attacks by the attacking force, then the defending force that used such persons in proximity to its forces or military objectives would be unlawfully using the presence of such persons to shield its operations or its military objectives from attack.” What am I missing?
A: Section 18.104.22.168 addresses only harm to certain persons employed “in or on” military objectives; it does not address harm to persons “merely in close proximity to a military target without actually being in or on it.”
By its terms, section 22.214.171.124 applies only to “[h]arm to certain persons who may be employed in or on military objectives.” (See, for example, the title of section 126.96.36.199, the first sentence of 188.8.131.52, and the third bullet of 184.108.40.206.) We used the words “in or on” in order to keep the scope of this provision narrow. Being “merely in close proximity to a military target” therefore would not suffice to make this provision applicable. For example, this provision could apply to a civilian worker in a munitions factory, but this provision would not apply when that worker was at home or merely near the factory. When section 220.127.116.11 mentions defenders placing certain persons “near” or “in proximity to its forces or military objectives,” section 18.104.22.168 refers to the obligations applicable to a defending force. But the rule set out in section 5.12 applies to an attacking force. The reference to the defender’s obligations was intended to help the reader understand the rationale for the rule in section 22.214.171.124, not to define the scope of that rule.
In addition, section 126.96.36.199 does not state that these persons are in a “carve-out from proportionality analysis.” (By contrast, section 188.8.131.52 provides that harm to persons who are military objectives (e.g., enemy combatants) “would not need to be taken into account.”) The persons covered by section 184.108.40.206 would be assessed to be civilians or non-combatants. As noted in our prior response, the proportionality analysis and procedures that are applied before strikes would include taking feasible precautions with regard to such persons (e.g., the factory might be struck at night when workers were not present). Moreover, these persons would be considered by a commander in determining whether he or she should refrain from a strike because it was expected to result in excessive harm in relation to the military advantage expected to be achieved. We acknowledge that the statement that such persons “assume the risk” could be understood to mean that harm to such persons is not taken into account. This is not what we meant. This, and other language concerning proportionality, is being reviewed and will be adjusted as appropriate to avoid any such confusion. There is significance to “assumption of the risk,” but that significance does not extend to any sort of “carve-out” from being subject to a proportionality analysis. We note that, in practice, DoD generally applies standards that are far more restrictive and protective of civilians and noncombatants than the legal requirements.
Q: The answer to question one also suggests that if the Bedouin tribe did not know that the terrorist camp was susceptible to military attack as part of an armed conflict, that would also require counting them in pre-targeting proportionality analysis–even if they were actually in or on the target itself. (Am I interpreting you correctly?) Here I am confused about two different categories in 220.127.116.11 that seem like they create a loophole: there are “civilian workers who place themselves in or on a military objective, knowing that it is susceptible to attack, such as workers in munitions factories” but there are also “persons authorized to accompany an armed force,” who do not get such a knowledge requirement. Under 4.15, authorized accompanyers include non-military persons who are serving that armed force, such as contractors and civilian provisioners for an army. Could the Bedouin tribe be counted as persons authorized to accompany an enemy force,and therefore exempted from any knowledge requirement?
A: No. In your example, members of the Bedouin tribe are not “persons authorized to accompany the armed forces.” As explained in section 4.15 of the manual, the phrase “persons authorized to accompany the armed forces” is a term of art under the law of war that specifically refers to a category of persons entitled to prisoner of war status under the 1949 Geneva Conventions. In your hypothetical, the Bedouin tribe has not received authorization from a State’s armed forces, but instead is supporting a non-State armed group.
Q: The answer to question two does not include the two conditions mentioned in the answer to question one. To clarify, are you saying that for purposes of post-strike battle damage assessments, non-members of an armed force who were providing support services to that force and killed during a strike at that force are unequivocally counted as non-combatants, even if they were “in or on” the military target and even if they knew the target was susceptible to military attack as part of an armed conflict?
A: Yes, for the purposes of post-strike battle damage assessments and more generally, persons who are not part of an armed group who were providing support services to that force that did not amount to direct participation in hostilities would be counted as non-combatants, even if they were “in or on” the military target and even if they knew the target was susceptible to military attack as part of an armed conflict. (In your example, it does not seem as though the Bedouin tribe is functionally part of the armed group, see section 18.104.22.168, nor does it seem as though the Bedouin tribe was taking a direct part in hostilities, see section 5.9.)
It’s worth noting that the rule in section 5.12 requires refraining from certain attacks that are “expected” to cause excessive harm. Thus, a person applying 5.12 would rely on information available before the strike, rather than information after the strike, such as a battle damage assessment of that strike.
Lastly, we are aware of comments that have been made on section 5.12 of the manual. We are continuing to consider these comments, as well as all other comments on the manual, with a view toward updating and improving the manual.