An Air Force Inspector General review of the Aug. 29 drone strike that killed 10 civilians in Kabul is not expected to delay the issuance of a final DoD Instruction (DoDI) on civilian harm by the end of this year, the Pentagon said on Sept. 22.
“The DoD Instruction on Civilian Harm Mitigation and Response remains in active coordination across the department,” Michael Howard, a spokesman, wrote in an email. “While we do not anticipate that the August 29, 2021 strike in Kabul will in any way delay publication of the DoD Instruction, we are always actively working to identify lessons and recommendations from operations to advance the department’s approach to civilian harm mitigation and response.”
In that strike, a General Atomics MQ-9 Reaper shot a Lockheed Martin [LMT] AGM-114 Hellfire missile that killed 10 Afghan civilians, including seven children and Zamarai Ahmadi, a 43-year-old aid worker with the U.S.-based Nutrition and Education International nonprofit who was bringing water containers to his family. U.S. Central Command Marine Gen. Kenneth McKenzie has said that six Reapers had tracked a white Toyota Corolla for eight hours and that analysts had feared that the containers were laden with ISIS-K explosives.
On Sept. 21, Air Force Secretary Frank Kendall tapped Air Force Inspector General Sami Said, who has had extensive experience flying F-15s and F-16s, to lead a 45-day review of what happened and why.
“There must be accountability,” Sen. Chris Murphy (D-Conn.), chair of the Senate Foreign Relations Subcommittee on Near East, South Asia, Central Asia, and Counterterrorism, said on Twitter on Sept. 17. “If there are no consequences for a strike this disastrous, it signals to the entire drone program chain of command that killing kids and civilians will be tolerated.”
Larry Lewis, the director of the Center for Autonomy and Artificial Intelligence at the non-profit research group CNA and an author of a 2010 civilian casualty study for the Joint Staff, said on Twitter that the Aug. 29 strike fit a typical pattern of civilian casualties in U.S. air strikes–misidentification and transient civilians.
“The U.S. has committed thousands of civilian harm incidents, and they all have things in common, patterns that emerge,” Lewis said. “The best way to improve is to identify the patterns and find ways to mitigate them. See the forest for the trees.”
Lewis said that in 2011 he compared manned aircraft to drones “and found that drones were 10 times more likely to cause civilian casualties than manned aircraft.”
“Drones have a number of features that increase the risk to civilians,” Lewis said. “One of those features is a distributed process for processing, exploitation, and dissemination. I found that there was a pattern. In a set of cases, the fact that civilians were present in the target area was known by someone in the group of operators and analysts, but that information didn’t reach the commander making the decision to fire.”
Pentagon officials and advocates for civilian harm mitigation have discussed the future of NCV, which the Pentagon removed from its doctrine in 2018 in a move that has not gained wide attention. Numerical NCVs gained currency in Operation Inherent Resolve (OIR) in Iraq and Syria during the Obama administration. The targeting process ostensibly took into account urban density, the value of a target and surrounding structures, such as residences, government buildings, apartments, mosques and bridges, to provide insights to commanders on whether to strike or hold off.
If a strike was likely to kill more persons than specified under the NCV, the strike commander would have to appeal to higher authority, whether the combined air operations center (CAOC) commander, the combatant commander, the defense secretary, or even the president.
DoD asserts that its collateral damage estimation (CDE) process is rigorous, that DoD routinely updates its CDE methodology with weapons and strike analysis, and that no need exists for operational effectiveness statistics on CDE aids, including the former Digital Precision Strike Suite Collateral Damage Estimation (DCiDE) algorithm and the current Digital Imagery Exploitation Engine (DIEE) tool, which incorporated DCiDE and resulted in a more streamlined and automated CDE process, according to DoD.
But Lewis has said that the Pentagon needs to have such statistics to calibrate its CDE tools.
“I strongly disagree that it’s not practical to develop statistics on a CDE tool’s effectiveness,” Lewis wrote in an email. “Not only is it practical, it is needed in light of differences we observe between predicted and actual collateral damage outcomes.”
The apparent lack of U.S. verification of pre-strike estimates of civilian casualties with the actual numbers of post-strike casualties reduced confidence in the former NCV measure.
Combatant commands use the guidance in the classified Chairman of the Joint Chiefs of Staff Instruction (CJCSI) 3160.01C—No Strike and the Collateral Damage Estimation Methodology, a process modified since its adoption more than a decade ago.
CENTCOM used the guidance of CJCSI 3160.01C during OIR against ISIS targets in Iraq and Syria. Analysts perform civilian pattern of life analyses and use algorithms to inform the targeting process.
Under CDE, targeting planners examine the possible effects of various weapons to propose recommendations to lessen collateral damage to comply with rules of engagement (ROE) and the law of armed conflict (LOAC).
Asked what changes to targeting practices the Pentagon has made because of OIR, the Pentagon has said that “among the many lessons reflected in CJCSI 3160.01C, No-Strike and the Collateral Damage Estimation Methodology, is the elimination of the use of Non-Combatant Cutoff Values in the CDE methodology.”
Chris Woods, the director of the United Kingdom-based Airwars, a non-profit that investigates civilian harm, has suggested that no data is available that would indicate that CDE is properly calibrated to give an accurate measurement of the likely harm to civilians.