The Biden Administration Faces a Reckoning Decades in the Making Over the United States’ Use of Air Power and Civilian Harm

​​This past fall, a series of reports, principally from the New York Times, revealed that the United States’ air war against the Islamic State may have been more indiscriminate than U.S. officials have previously acknowledged. The Times reported in November 2021 that a March 18, 2018, single strike indiscriminately killed dozens of civilians during the Islamic State’s last stand in Baghuz, Syria. And in late January, the Times reported that another strike in 2017 on the Tabqa Dam—a massive structure on the Euphrates River in Eastern Syria—caused the dam to fail so catastrophically that it provoked a momentary truce among the Syrian government, the Islamic State, and U.S.-backed forces so that engineers could prevent the loss of tens of thousands of civilian lives due to flooding. 

After the Times reported on the Baghuz strike, U.S. Secretary of Defense Lloyd Austin conceded to reporters that the Pentagon needed to do more to prevent “tragic” civilian deaths and to fix casualty reporting. Then, at the end of December 2021, two more stories from the Times suggested that deeper organizational issues were leading U.S. Central Command (CENTCOM), the U.S. military headquarters that oversees operations in the Middle East, to both tolerate and underreport civilian casualties. After that story’s publication, Austin issued a memorandum on Jan. 27 entitled “Improving Civilian Harm Mitigation and Response.” This memorandum directed the under secretary of defense for policy to provide a “Civilian Harm Mitigation and Response Action Plan” by April 27 of this year.

In light of all of this, it’s worth digging into some relevant background on targeting. Over the past 30 years, the U.S. military has spent considerable effort developing a targeting policy—generally called the “joint targeting process”—that is disciplined, that requires the routine participation of operational lawyers to ensure compliance with international law, and that scrupulously seeks to minimize the risk of civilian harm. One of the most disturbing aspects of the recent Times reporting is not that this disciplined targeting process fails in practice, but that it has been routinely circumvented. 

Advocates have made the case that the entire joint targeting process needs significant reform to prevent the recurrence of past civilian harm—Luke Hartig discussed the arguments for broader changes in a recent Just Security article. But as the Department of Defense considers targeting reform, it can take one significant practical step to deter circumvention of the targeting process altogether. The department can incorporate checks and balances—specifically poststrike audits—into the targeting process to deter circumvention.

Targeting Theory and the Prevention of Civilian Harm

The United States has long wrestled with reconciling its traditional superiority in—and hence heavy reliance on—air power with the law of war and its humanitarian values. Relative to other modes of combat, air power is notoriously underregulated in conventional international law. Whereas armies are regulated right down to what bullets they may fire, and several conventions and manuals regulate navies down to when, where, and what may be attacked, there has been only one major effort in conventional international law to similarly regulate air power, and it failed.

As a consequence, questions of proportionality and steps taken to prevent civilian harm were treated largely as matters of policy, not law, throughout most of the 20th century. Despite the profligate use of area bombing during World War II, no Axis leaders were prosecuted as war criminals for the indiscriminate use of air power, despite many being hanged by the neck for indiscriminately devastating cities with ground forces. Only Japan codified any laws that specifically treated indiscriminate bombing as a war crime (and then only for dubious reasons). And the Cold War emphasis on strategic nuclear conflict, where casualty projections were quantified in “megadeaths,” made any pretense of proportionality regulations facetious.

That changed starting in the early 1990s, when what is now called “operational law” put judge advocates—and, consequently, considerations of international law—into the routine conduct of targeting operations. This change corresponded not just with the end of the Cold War but also with the development of increasingly precise munitions that made it feasible for operational commanders to consider nearly every bomb dropped. Since then, the U.S. military has built a robust, layered process for choosing whether and how to bomb enemy persons, places or objects that incorporates the law of armed conflict at nearly every turn. 

Targeting decisions a century ago relied on large maps labeled to ​​reflect whole cities’ population density, strategic value, and inflammability, and lawyers were nowhere to be found. The modern military instead relies on what is generally called the joint targeting process. Its elements are specified in a detailed manual (the most recent public version of which runs almost 150 pages), which specifies a six-step joint targeting cycle.

The joint targeting process requires commanders to carefully consider weaponry, civilian infrastructure and civilian casualties at each decision-point before ordering a strike. And it is explicitly structured to implement law of war requirements into operational decisions, nearly always with the real-time advice of military lawyers. Commanders are required to conclude that the anticipated military advantage from a strike would outweigh expected civilian casualties. The targeting manual incorporates language from the Additional Protocol I to the Geneva Convention, including precautionary measures and criteria for defining a military target. And the distinction requirement—that combatants distinguish combatants from noncombatants and target only the former—is assessed in three parts: positive identification, combat identification and target identification. 

The joint targeting manual presents targeting as a humanitarian math problem: Through a calculation called collateral damage estimation (CDE), different types of anticipated civilian harm are assigned varying values based on computer modeling, historical data, field tests and more. As targets accumulate more points—based on the types of collateral damage attacking them might bring—they “level” up and require more stringent precautionary measures and, eventually, approval from higher-ranking commanders in the chain of command. The specific approval criteria for varying levels of harm remain classified, but declassified documents and a 2011 study by Gregory S. McNeal provide enough flavor. This whole system is designed to keep the risk of collateral damage below 10 percent, but when the risk is to critical infrastructure that threshold drops to 1 percent.

The joint targeting manual is candid that the goal of the joint targeting process is not perfection, but to provide commanders “a rational and iterative process that methodically analyzes, prioritizes, and assigns assets against targets systemically.” The “fog of war” makes this kind of a structured, layered, and predictable approach to strikes all the more important and accountable. Commanders face life-and-death decisions about their own troops, enemy combatants and civilians when they make strike decisions. The joint targeting cycle helps commanders think clearly. The joint targeting process incorporates clear and deliberate steps, rather than relying on ad hoc intuitions about whether a target meets the proportionality requirement. This disciplined approach aims to reduce civilian harm in the aggregate.

The modern U.S. military conducts three types of strikes, two of which rely on the rigors of the joint targeting process. The first, and default targeting mode, are “deliberate” strikes. The manual assumes up to 10 days of lead time for target selection and vetting, but typically requires a minimum of 24 hours. Such strikes entail the full rigor of the joint targeting cycle.

The second type of strike, where there is no time for a deliberate strike, are “dynamic” strikes that permit a more abbreviated use of the joint targeting cycle. A number of circumstances might push a targeting decision into the dynamic category. For example, an enemy asset might present itself unexpectedly as a “target of opportunity,” or a “time sensitive target” might enter the battlefield and threaten coalition forces. Even the dynamic targeting process, though, contains detailed requirements that are designed explicitly to ensure commanders comply with the law of armed conflict. Circumstances sometimes mean that judge advocates are unable to participate directly in specific dynamic targeting decisions, but the joint targeting manual requires their general input on the system that commanders employ for approving dynamic strikes.

The third type of strike is the “self-defense” strike. These strikes are not explicitly incorporated into the joint targeting manual. Instead, when a U.S. unit faces a “hostile act or demonstrated hostile intent,” the standing rules of engagement give U.S. forces the inherent right to self-defense. This right contemplates, for example, situations where U.S. ground forces are in such imminent need of air support that there is no time for even the abbreviated procedural requirements of dynamic targeting. But in the fight against the Islamic State, U.S. forces have invoked the right to self-defense in situations where the threat they faced was not so clear. 

Civilian Harm in Practice

Over the past two decades, there has been no shortage of public reporting on the civilian harms caused by U.S. airstrikes. Most reports have focused on the fatal consequences of faulty intelligence. After a suicide bomber killed nearly 200 people outside the Kabul airport in August 2021, the United States bombed a car in the mistaken belief that it was carrying another bomb, killing an aid worker and seven children instead. The recent Times reporting, however, suggests a far more pervasive problem that may be a predicted consequence of the Trump administration’s policy of delegating significant war-making decisions further down the chain of command. 

A May 2021 report from the Defense Department Office of Inspector General on CENTCOM’s targeting and civilian casualty reporting practices concluded that CENTCOM “followed … [p]rocedures for kinetic strikes” and casualty reporting. But in December, the Times published a scathing review of the adequacy of 1,300 newly released CENTCOM civilian casualty reports going back to 2014. The Times reporting was corroborated in January by an independent assessment conducted by the Rand Corp., pursuant to a requirement included in the National Defense Authorization Act for Fiscal Year 2020. The Rand study had a damning set of conclusions: The Pentagon had failed to accurately capture civilian casualty data, is not sufficiently well organized to mitigate and respond to civilian harm, and is not prepared to address the issue in future combat operations. 

This increase in civilian casualties as a result of U.S. targeting might surprise those who have paid attention to the development of modern targeting doctrine. With military lawyers at nearly every step in the targeting cycle, what else could be done to mitigate and respond to civilian harm? By historical standards, and the standards of most of its peers, the modern U.S. military is more attuned to operationalizing the law of armed conflict into its use of air power than any major armed force has ever been. 

If the Times reporting is taken at face value, the major problem is a loophole that allows U.S. forces to skip the targeting process altogether. According to the Times, the strikes in Baghuz and against the Tabqa Dam were ordered at the exclusive direction of special operations forces on behalf of coalition partners. In the case of the Baghuz strike, Syrian Democratic Forces (SDF) fighters reported being under attack and that there were no civilians in the area. A U.S. special operations task force ground commander radioed to a fighter jet and had it drop three 500-pound bombs on what were claimed to be Islamic State fighters. In the strike against the Tabqa Dam, special forces operators coordinated directly with a bomber jet overhead, which dropped three 2,000-pound bombs on the 18-story structure to deny the Islamic State its use as a staging area. Both strikes came as a total surprise to military operators in the Combined Air Operations Center (CAOC) in Qatar, which was ostensibly tasked with directing the air war against the Islamic State. 

Consistent with the joint targeting process, the Tabqa Dam was reportedly on a “do not strike” list. (At the same time, the military was devoting enormous resources to protecting the fragile Mosul Dam in neighboring Iraq.) The Times detailed that the strike in Baghuz stunned CAOC operators, who were watching real-time drone footage of the area and initially were unsure who had been dropping bombs. One CAOC operator was reportedly so troubled by the sight of a crowd of women and children being annihilated, that he submitted a formal report claiming to have witnessed possible war crimes. A Defense Department analyst from the Office of the Inspector General could identify from the video evidence only two people with weapons in the crowd.

Because special operators reportedly invoked self-defense, the joint targeting process was not followed in either case. The Times’s reports and the government’s own disclosures show that U.S. strikes in the fight against the Islamic State have departed from the carefully assembled joint targeting cycle by making dynamic strikes and self-defense strikes the rule, rather than the exception. One Air Force officer told the Times that self-defense strikes represented about 80 percent of U.S. sorties by late 2018. 

Dynamic targeting and self-defense strikes are obviously necessary in some circumstances. But because they do not leave time for careful consideration, they also carry increased risk of civilian harm. As the Rand investigation recently described, dynamic strikes and self-defense strikes are often missing “intelligence package[s], planned weaponeering, and CDE.” And precisely because self-defense strikes can be directed from the tactical level—beneath the layers of command structure that the joint targeting process requires—they offer a tempting way to circumvent the oversight and accountability safeguards the process is designed to preserve.

Circumventing the joint targeting process increases the risk of civilian harm. The process requires a close analysis of how the means and methods used might mitigate the possibility of civilian harm. For example, reporting suggests that whether and how to target Islamic State fighters near the Tabqa Dam had been subject to extensive deliberations, with analysts concluding that any strike ought to use only low-yield munitions to avert the possibility of a humanitarian catastrophe. It is unclear whether the strike was technically an incident of dynamic targeting or self-defense. But the unit that ordered the strike ignored all of the joint targeting process’s embedded safeguards, including a weaponeering assessment. Likewise, CENTCOM responded to the Times’s reporting on the Baghuz strike by asserting that the 500-pound bombs carried by the F-15s overhead were the only munitions available. But as Ryan Goodman noted in Just Security, proportionality under the law of armed conflict is not excused simply because more targeted ordnance is not available at the time of a strike. The result was the deaths of more than 60 women and children.

These risks are exacerbated by the apparent willingness of special operators on the ground to order hasty strikes at the behest of partner forces like the Iraqi Army or SDF. These strikes—in the name of “collective self-defense”—have been indisputably important to protecting partners who have done the bulk of the fighting in America’s conflicts, particularly against the Islamic State. But the joint targeting manual has historically treated the need to respond to a “threat to friendly forces or allies” as a type of dynamic targeting that is still governed by the joint targeting cycle. Yet, the Baghuz and Tabqa Dam strikes were reportedly undertaken to answer self-defense calls from coalition partners, not because U.S. special operations forces were in danger. 

Putting the full might of U.S. air power at the disposal of coalition partners leads to deficiencies in both target selection and after-the-fact reporting. The Rand study, for example, found that the Operation Inherent Resolve (OIR) Civilian Casualty (CIVCAS) cell, responsible for keeping track of civilian harm, received significantly less video footage of self-defense strikes than of preplanned targeting strikes. And poststrike reporting from self-defense strikes done at the behest of coalition partners was even worse. Rand reports that, in those circumstances, the OIR CIVCAS cell was left relying on unreliable civilian casualty assessments produced by the Iraqi military to conduct poststrike assessments.

As former CENTCOM commander Gen. Joseph Votel (Ret.) warned, the United States has increasingly limited its involvement in foreign conflicts to its special operations forces and air power, which means that partner forces increasingly shape strike decisions. Intelligence about the nature of targets and proportionality analysis from partner forces, however, is likely to reflect a substantially different view of what the law of armed conflict requires. And this may help to explain why, as Larry Lewis, a former State Department and Defense Department adviser told the Times, civilian casualty rates were 10 times higher in Syria than in Afghanistan. U.S forces were more heavily involved on the ground in Afghanistan than in Syria—the military did not rely as heavily on reporting from partner forces to make strike decisions.

Preventing Civilian Harm

The Jan. 27 Austin memo directs the under secretary of defense for policy to provide a “Civilian Harm Mitigation and Response Action Plan” (CHMRAP) by April 27. The CHMRAP, which will eventually lead to a Defense Department instruction, is intended to emphasize the duty of “all leaders” to protect civilians and establish a robust, long-term framework for mitigating civilian casualties. Certain recommendations in the Austin memo come directly from the Rand report: (a) the formation of a Civilian Protection Center of Excellence, (b) standardized data collection and management on civilian harm, (c) a review of the Defense Department’s condolence and apology policies, and (d) the incorporation of civilian harm mitigation and response guidance into doctrinal and operational plans. The contents of the Rand report likely preview much of what might emerge from the CHMRAP.

The Austin memo is conspicuously silent on the need for reforms in the United States’ ongoing operations in Syria. It notes that the conclusion of U.S. involvement in Afghanistan and the transition to an advisory role in Iraq offers an “opportunity” to “institutionalize improvements.” This long-term outlook is important. But U.S. special operations forces are still fighting the Islamic State in Syria. Nothing in the Austin memo addresses near-term combat situations in which there might be a significant risk of civilian harm. Because the Islamic State fights as an insurgency, with its members (and leaders) hiding among the civilian population, the risk of civilian harm from airstrikes remains extremely high. If the focus of the CHMRAP is on the mitigation of civilian harm in hypothetical future large-scale combat operations—the Rand report’s focus—then significant gaps remain in present-day combat operations. It is reassuring that Austin reportedly put Christopher P. Maier, the assistant secretary of defense for special operations and low intensity conflict, in charge of the CHMRAP plan. In choosing the senior-most Pentagon official responsible for special operations forces, Austin demonstrated that he recognizes that special operations forces have instigated many of the strikes that cause the most civilian harm. But it remains to be seen whether the end product will address contemporary combat operations. 

The concrete reforms outlined in the Austin memo also focus heavily on the aftermath of strikes, rather than on preventing civilian harms. Condolence payments and poststrike data collection are important, and information about past strikes can help inform future strike decisions. None of the memo’s concrete directives address the prestrike decision process. The memo says that the new Civilian Protection Center of Excellence will “expedite and institutionalize” the development of strategies for the prevention of civilian casualties. But the Defense Department should also consider including some of Rand’s concrete recommendations, such as “[i]ncorporat[ing] civilian harm into pre-operation intelligence estimates and postoperation assessments of the cumulative effect of targeting decisions” and “[e]stablish[ing] guidance on the responsibilities of U.S. military forces in monitoring partners’ conduct and offer[ing] assistance to partners in building their own assessment capabilities if needed.”

The Austin memo also fails to mention one of the more practically meaningful steps that could be taken to minimize civilian harm in the near term: closing the gaps in the joint targeting process. Closing those gaps, if done thoughtfully, could contribute significantly to reducing civilian harm and increasing accountability while still providing tactical operators flexibility on the battlefield.

To be sure, the rigor of the joint targeting process can feel like cumbersome and even life-threatening red tape when real-time battlefield conditions require air support. And self-defense strikes are certainly needed sometimes. In 2018, for instance, a special operations unit relied on self-defense strikes to stop an oncoming horde of hostile Russian mercenaries and pro-Assad troops in Deir al-Zour. Just 40 American special operators defeated 500 attackers overnight without a single U.S. casualty, in no small part due to a barrage of air strikes from coalition fighter jets, bombers and attack helicopters that killed 200-300 pro-Assad forces. 

But several factors distinguish the Deir al-Zour strikes from those at Baghuz and Tabqa Dam. According to public reports and a CENTCOM statement, the Deir al-Zour strikes’ targets posed an immediate, deadly threat to U.S. forces that U.S. forces had themselves identified, not coalition partners. And the orders for the strikes also appear to have gone through the CAOC, the very channel that a U.S. military task force reportedly circumvented in the Tabqa Dam strike. 

The most dangerous gap in the existing joint targeting process is not that it can be circumvented when the needs of self-defense require. It is that it lacks any disincentive against circumventing its safeguards entirely and routinely by declaring every strike an act of self-defense. 

One of the simplest ways of providing that disincentive is with mandatory post hoc review of strikes that circumvent standard targeting process. Institutionalizing such automatic poststrike audits would not be unprecedented. In 2005, the Israeli Supreme Court issued a landmark opinion on the Israel Defense Forces practice of targeted killing. The court emphasized four requirements that dictated whether the risks of harm were sufficiently imminent to warrant the use of force instead of arrest in cases where suspected terrorists could be targeted with lethal force. Three of those requirements echo factors that drive the joint targeting process: a proportionality assessment, quality intelligence respecting the target, and the least dangerous means to counter the threat. A fourth requirement, not found in the joint targeting manual, was that force could be used only if there was a guarantee of an independent and “thorough investigation regarding the precision of the identification of the target and the circumstances of the attack” after the fact. 

Controversially, the Israeli Supreme Court drew this fourth requirement from human rights law principles. But its logic as a means of operationalizing the law of armed conflict around targeting is compelling: People will take more care to abide by rules when they are certain someone will check their work. 

As things stand, the only significant check against improperly circumventing the joint targeting process is the prospect that a strike turns out to be so egregious that it triggers a DoD-internal war crimes investigation, an unlikely prospect. That is hardly a compelling deterrent, particularly for those who believe that calling in air power will save lives. A special operator who knows that calling in a self-defense strike will trigger an automatic and independent audit, however, will be more apt to see the situation clearly and, when possible, seek out the second-looks, deliberation, and shared accountability the joint targeting process aims to foster. And to ensure the perfect not be set against the good, the rigor of such post hoc audits could be calibrated to the degree of prestrike compliance with the joint targeting process to encourage the greatest compliance feasible. 

The routine character of such audits would also go a long way toward building up reliable data on strike effectiveness and collateral damage to better inform the joint targeting process in the future. One of Rand’s recommendations was to monitor coalition partners’ conduct and compliance with the law of armed conflict. Rigorous and consistent poststrike audits will help to quickly provide metrics for which coalition partners are reliable and which partners’ information requires more careful vetting. 

Finally, the Biden administration’s review of targeting policy should also include a reassessment of the extent to which significant decision-making authority was both subdelegated and fragmented during the Trump administration. One of the more disturbing aspects of the Times reporting, if true, was that the CAOC was routinely cut out of the loop in the conduct of major air operations of significant humanitarian as well as political sensitivity. This circumvention led not only to confusion within CENTCOM but also to false public statements that, when subsequently disproved, undermined the United States’ credibility. There have been other indications that the Trump administration’s efforts to loosen the rules of engagement for airstrikes led to a rise in civilian casualties. 

The exercise of initiative and discretion by those closest to the ground is obviously valuable. But diffuse command and control weakens accountability, invites lawlessness, and increases civilian harms in ways that not only are tragic but also undermine the nation’s moral authority. The United States is positioned to project its values as a source of diplomatic as well as military strength, but only so long as it is willing to live up to them.