One of the more remarkable aspect of the U.S. Air Force’s investigation into the disastrous Aug. 29 strike in Kabul is just how circumscribed its findings are. In last week’s press conference, the lead investigator, Lt. Gen. Sami D. Said, advises multiple times not to extrapolate from this “very unique” strike – conducted in a compressed time frame in a high-threat environment – to the drone program at large. But in hearing Said’s own account, read alongside excellent reporting from The New York Times and Washington Post, it’s hard to avoid some profound questions about the fundamental methodologies that undergird the program. Indeed, this may well be a watershed moment for the drone program, one that forces the administration and the military to do a lot of work if U.S. drone strikes are ever to be seen as sufficiently credible again or if reforms touted by policymakers are to be seen as meaningful. Or it could be seen as the strike that broke confidence in the program and leads policymakers and the public to conclude that fewer strikes and greater reliance on defenses should guide counterterrorism operations going forward.

The tragedy for the Afghans killed is searingly obvious. Yet when it comes to evaluating U.S. actions on Aug. 29, at least one way to view the strike is with empathy for the personnel involved. This is where Said ends up. Coming off of a tragic bombing at Kabul airport that killed 13 American servicemembers and almost 200 Afghans, U.S. forces were at a heightened state of alert. It’s the kind of situation that our personnel train for but is still immensely stressful in the moment. By Said’s account, U.S. military personnel were doing their best to keep the airport – and thousands of civilians amassed around it – safe while upholding high operational standards. Further, the targeting timeline was compressed (to about eight hours) due to the assessed imminence of the threat, and while U.S. Central Command head Gen. Kenneth F. McKenzie previously said that “this was not a rushed strike,” Said concluded that the timeline made it difficult to do the careful intelligence analysis required to catch errors like the misidentification of an innocent man, Zemari Ahmadi, as a terrorist operative.

Yet the results of the strike were unspeakably tragic – an innocent man targeted, ten innocents killed, including seven children – and Said offered up fairly modest solutions, what one journalist at the press conference told him felt like, “tweaks around the edges.” His recommendations were carefully bound only to self-defense strikes against imminent threats to forces, and he spoke in vague terms about checking confirmation bias, improving communications and situational awareness, and reviewing how the military assesses the risk of civilian casualties prior to the strike.

The problem is that it’s hard to constrain judgment to just this kind of strike or to chalk the errors up as simply honest mistakes made in the moment. The situation here is not a checkpoint gunner mistakenly firing on a speeding vehicle. Drone strikes are complex analytic and operational undertakings that rely on techniques and procedures developed and refined over many years. Sure, the strike was taken during a tense time and the rules of engagement reportedly required only “reasonable certainty” that they had the right target and that no civilians would be harmed, as compared to the higher “near certainty” requirement that governs many other strikes. But the details made public suggest some fundamental flaws in the underlying targeting methods that should raise deeper questions as to how the U.S. military identifies targets, seeks to prevent and investigate civilian casualties, and communicates about its operations around the world.

Confirmation Bias

Let’s start with one of Said’s most notable findings – the prevalence of confirmation bias throughout the targeting process. We don’t know much about what the intelligence showed regarding the threat, or how granular that information was, though Said alluded several times to classified information that would substantiate his account. Yet, more recent New York Times reporting suggests that the military still doesn’t know the true location of the ISIS-associated structure that was supposedly at the center of the plot. Regardless, once the military had erroneously landed on Ahmadi as a possible ISIS terrorist, they proceeded to massively misread a number of totally innocent and commonplace facts and events.

They were reportedly on the lookout for a white Toyota Corolla at the (misidentified) ISIS compound. Toyota Corollas are the most common vehicles in Afghanistan, and anybody who has spent any time in Kabul can tell you the model is frequently featured in threat reporting. Further, the targeting cell reportedly saw Ahmadi receive a laptop bag from his boss, which matched intelligence on how ISIS carried out its Aug. 26 attack on the airport. Finally, Ahmadi was observed on several occasions filling large bottles with water to take to his home, where he had erratic water service, yet the targeting cell interpreted the containers as explosive materials.

Confirmation bias is a risk in any analytic endeavor, and perhaps there were other factors that analysts saw as confirming their assessment, but it’s hard to fathom how such unremarkable things – driving a ubiquitous car, handing off a bag, and filling water bottles in a place with water shortages – could be interpreted as definitively confirming threat reporting. And it’s even harder to imagine how the military did not develop systems to correct for this risk – even basic things like red-teaming, which Said proposed – long before now.

Misidentification of Civilians

Then there is the misidentification of civilians. The first accounts of the strike suggest that the military saw a man standing in the courtyard where the strike took place and assessed him to be a combatant because of his association with Mr. Ahmadi. It’s unclear if the military did any additional due diligence on the man, who turned out to be Mr. Ahmadi’s cousin and a former U.S. military contractor. But it points to another pitfall of air strikes: bad initial intel tends to compound.

It also raises serious questions about who the Department of Defense considers to be targetable as a combatant and the permitted factual inferences to meet those standards. If determinations of lawful targets are made with less than the utmost rigor, then further assessments of associated combatants will be, by definition, built on shaky foundations.

Presence of Children

As to the children killed in the strike, Said confirmed that kids were seen in the courtyard two minutes before the Hellfire missile was launched. But he insisted that this fact was “100 percent not obvious. You have to be no kidding looking for it.” Said then told The New York Times that four adults and two children were visible nine seconds before the launch of the missile. Yet Said concluded that it would have also been easy to miss the children in viewing surveillance footage. This is a remarkably generous interpretation of a major operational misstep. We have been conducting targeted airstrikes for two decades, and one of the most frequent critiques has been that the strikes kill children. How, then, did the targeting cell not have trained personnel present to really look for the presence of children, even when their appearance was not obvious?

Situational Awareness and Information Sharing

CNN and Times reporting suggests that CIA analysts actually did have indications that children were present prior to the strike, but the analysts were not co-located with the targeting cell and a communications breakdown prevented this information from getting to the operators in time. Said appears to confirm this reporting (though he’s vague due to classification concerns), which fueled his recommendation for enhancing situational awareness and better information sharing.

This was not the only failure to gain and maintain full situational awareness. The targeting cell apparently also had no idea that one of the places Ahmadi visited in the hours before the strike was the offices of the U.S. non-governmental organization (NGO) where he worked. This is a remarkable failing. After twenty years of being in Kabul, the military still did not seem to have a full accounting of certifiably non-combatant locations. Or if it did, it had not fed those locations into a do-not-strike system. This failure is particularly troubling considering that, six years ago in Afghanistan, the U.S. Air Force similarly failed to ensure that an aircrew had possession and properly used a no-strike list, which included an internationally recognized NGO-run hospital, resulting in the misidentification of the hospital as a Taliban-controlled building and a strike that killed more than forty doctors and patients.

Secondary Explosions

Finally, there’s the military’s erroneous claim that there were secondary explosions, likely indicating that Mr. Ahmadi’s vehicle was carrying explosive material to be used in an attack. In Said’s account, the operators saw a larger than expected explosion and only some time later assessed that the missile likely detonated some sort of gas tank. It’s incomprehensible that an organization that looks at explosions day in and day out would make such a mistake. Within a couple of days, the Washington Post and New York Times both spoke with civilian explosives experts who pointed toward clear signs that the only significant explosion came from the missile – things like blast patterns, that nearby trees had not had all their leaves blown off, that surrounding structures remained intact.

This is surely not the first time that a strike, especially one conducted in a closed courtyard in an urban environment, set off nearby flammable materials. How, then, could the military fail to distinguish between a small gas tank and explosives, either in the initial blast or in post-strike surveillance of the surrounding area? And does this mean that there have been other cases where exploding gas tanks were taken as confirmation that the military had hit a bombmaker or the like?

***

In short, Said’s public discussion of his classified report evinces empathy not just for the operators serving under trying circumstances – which may well be warranted – but for a set of procedures and methodologies that were presumably refined over many years but that simultaneously and catastrophically collapsed in eight hours.

This may well be a watershed moment for the drone program

Since leaving government, I have written extensively and critically of the drone program and the policies that govern it, but I have been steady in defending the professionalism of the men and women who carry out these operations. I have seen countless hours of drone footage, gotten to know the people conducting these missions, and came away amazed at the actions our operators take to address terrorist threats without harming civilians. I would bet that the personnel involved in the Kabul strike were similarly adept and focused on doing their best in a tough situation. But as Loren DeJonge Schulman and Pauline Shanks Kaurin have noted, military professionalism is about more than being skilled and trying to do no harm. It is about incorporating the profession’s ethics into everything it does and embracing the characteristics of a learning organization. As Larry Lewis wrote yesterday, this is only the latest in a steady stream of operational mistakes that produced tragic results. With Said’s conclusion that there was no negligence, nothing approaching criminality, the message to operators may well be: keep doing your job and try to be better. If so, we will have missed an opportunity for deeper learning. Professionalism means introspection and embracing deep operational reforms. Lives are on the line and they deserve not just our best efforts and intentions but a set of systems and procedures to match.

Our credibility as a nation is also at stake. In the days after the strike, while the military’s initial review was ongoing, the military low-balled the potential civilian casualty assessment, even though al-Jazeera and others were already reporting ten civilian casualties. In an internationally broadcast press briefing, the Chairman of the Joint Chiefs of Staff referred to the strike as “righteous,” even as early information suggested the results were not clean and an investigation was ongoing. The military would revise its story several times more in the days and weeks ahead. As New York Times reporter Eric Schmitt put it, “Nearly everything senior defense officials asserted in the hours, then days and weeks, after the drone strike has turned out to be false.”

In reviewing this coverage, many of us have wondered what would have happened if not for the extraordinary Times and Post coverage of the strike. The Times’ video investigation of the strike was a remarkable almost minute-by-minute account of Mr. Ahmadi’s movement on Aug. 29, complete with security camera coverage of his comings and goings across the city, all made possible by years of reporting from Kabul. This type of coverage and insight has been lacking in the countless strikes that have occurred in places like the remote reaches of Nuristan, Afghanistan; Marib, Yemen; or Lower Shabelle, Somalia.

Indeed, we have many such examples, where local journalists and international human rights groups have alleged civilian casualties. In years past, the military often didn’t even address the allegations. Now the military will usually address the situation, but quite often that means disputing the account without saying why. Certainly there are times when journalists or NGOs get it wrong, but in some cases, it’s reasonable to assume that there were civilian casualties but a military relying on flawed procedures pronounced the strike righteous anyway. The Kabul strike will only deepen this credibility gap, and additional steps toward transparency may not be enough to bridge it.

Yet it’s more important than ever that the Biden administration does begin to bridge this gap. As a first step, U.S. Special Operations Command and Central Command should release their own findings on the strike and address whether they have taken any of the administrative or non-criminal disciplinary actions that Said indicated are available. It will be important to show not just overdue accountability for this strike but serious reforms intended to ingrain lessons learned from this tragedy.

Then there’s the question of new drone policies. In the coming months, the administration is set to release new guidance for drone strikes and commando raids. It will likely move the program toward still higher standards when it comes to constraining the use of force, protecting civilians, and improving transparency. But the Kabul strike will undermine the credibility of that document before it’s even completed and amplify the questions that my colleagues at Just Security have raised for years – e.g., about how combatant vs. non-combatant assessments are made, what reasonable certainty vs. near certainty mean, how imminence is defined for external vs. self-defense threats, and when a threat reaches a threshold that allows for a strike.

All of this suggests that we need more than a policy review of direct action. We also need a fundamental review of the operations and procedures that underpin the drone program and how those have evolved in the wake of previous mistakes. It should get into the weedy details of how operations are conducted and whether they meet what policymakers expect of them, and it should include a strong civilian voice, which is often underrepresented in such operational reviews. The administration needs to consider far more significant steps towards transparency on such operations if is to regain the public’s trust. And it may need to consider more dramatic measures that have been floated previously but foundered in the bureaucracy, such as establishing an outside or independent oversight board. Finally, the Kabul strike may reinforce the growing sense that no matter how much we try, drone operations go bad for a range of errors, both human and technological. At some point, the solution set must be more than technical fixes but rather a recognition that maybe we should conduct fewer strikes, that we shouldn’t fumble our way around countries we don’t understand very well, and that we should rely more on our defenses to keep our country safe.

There are certainly unique circumstances to the Kabul strike, but if we miss the bigger lessons, we only invite further tragedy.