"Kill TV," decision science, AI, and the Kill Web
BlogFebruary 25, 2021
Ray Alderman
VITA Standards Organization
WARFARE EVOLUTION BLOG. During the night of 7 October 2001, [Central Intelligence Agency] CIA-controlled Predator drone 3034 was flying over a mud-walled compound in Afghanistan, the suspected hideout of Taliban leader Mullah Omar. The infrared (IR) sensors picked-up heat signatures from three vehicles and a motorcycle leaving and heading toward Kandahar. The drone pilot, and the weapons officer controlling the two on-board Hellfire missiles, were sitting in a trailer on the grounds of CIA-headquarters (HQ) in Langley, Virginia. The video images from the Predator were being streamed, via satellite links, to the big flat-screen TVs at Langley, to the offices of military brass at the Pentagon, General Franks' office at central command (CENTCOM) in Tampa, Florida, to the offices of General Deptula in Qatar (who was controlling Air Force fighter planes and bombers over Afghanistan), and the office of General Jumper, the Chief of Staff of the Air Force. Ordinary soldiers call this video network "Kill TV," for reasons that will become obvious.
This tale comes from my reading of Andrew Cockburn’s book, “Kill Chain”, which covers this event in chapter 7. Also, Politico Magazine's excellent article (Sept 16, 2014) titled “How We Missed Mullah Omar” provides even greater detail. Now, back to the story.
A motorcycle and three vehicles traveling together, at that time of night, has the signature of a leadership convoy transporting an HVT (high value target). Omar was on the move. A military lawyer was standing beside Franks in Tampa, watching the screen. "Valid target," he said. That target was a Toyota Corolla in the convoy, the favored limousine of terrorist leaders.
The convoy meandered through the streets of Kandahar, stopped, and several people got out and entered a building. But the convoy moved on before the missiles on the Predator could be fired (at that time, Hellfires could not hit moving targets accurately). The convoy continued down the streets and stopped again. Several people got out and entered a mosque. "Not a valid target," the lawyer told General Franks. General Deptula already had fighter planes circling overhead, loaded with bombs, waiting for Franks' order to hit the building.
To read more Warfare Evolution Blogs by Ray Alderman, click here.
Moments later, all the brass watching their TVs saw one of the vehicles blow up. General Deptula, observing the fireball said, “Who the hell ordered that?” It was General Franks at CENTCOM, who gave the order to the CIA Predator operators at Langley to hit one of the vehicles outside the building with a missile, since he couldn’t use the fighter’s bombs to hit the mosque.
The remainder of the convoy vehicles scurried down the street, stopped at another building, and the passengers jumped out and ran inside (they are called “squirters”, those who survive an attack and run away). Could Franks flatten that building if civilians might be inside? He called up SECDEF [U.S. Secretary of Defense] Donald Rumsfeld for permission. Rumsfeld called-up President George W. Bush, and five minutes later, Rumsfeld gave Franks permission to destroy the building. But, a CIA officer at Langley told Franks in Florida that the building looked like another mosque.
Franks reviewed the video, decided it was not a mosque, and told Deptula to order the the F-18 pilots flying overhead to hit that building with bombs. The military lawyer standing beside him said, “You’re still good.” General Jumper, at the Pentagon, called-up Franks in Florida and told him that he was watching the video. He said he saw the HVT (Omar) escape from the building before the bombs hit. Franks called-up the Chairman of the Joint Chiefs of Staff and demanded that the flat-screen TVs in General Jumper’s office be removed. But, Jumper was right: Omar escaped unharmed at some point that night. He died in 2013 from tuberculosis.
My analysis
There’s a lot to analyze here, especially about the chain-of-command and the decision-making processes. So let’s see how the Kill Web concept will solve the problems. First, it’s clear that a lot of generals were involved, but most of them were not in the tactical chain-of-command. Franks was the tactical commander of the mission. It’s also clear that General Deptula, who was controlling the fighter planes over the mosque, was not in the communications loop between Franks and the weapons officer for the Predator at Langley. He was on a different communications link because he was surprised to see the vehicle blow up. He probably wondered if one of his pilots had dropped a bomb. And, it’s clear that some of the generals watching Kill TV, and not in the tactical chain-of-command, were getting close to micro-managing Franks during the mission. That’s why Franks demanded that Jumper’s screens be removed at the Pentagon.
Using artificial intelligence (AI), the Kill Web concept puts the right people in the right groups, and delegates authority and responsibility appropriately. Strategic people will allocate resources, set goals, and devise plans. Operational people will organize and task those resources. Tactical people will execute the plans. The tactical commanders will run the show, and the operational and strategic people will be observers.
Secondly, everyone involved here was looking at grainy blurry videos on old low-resolution flat-screen TVs. Franks was making decisions under uncertainty with imperfect intelligence. Our IR sensors and flat-screen TVs have come a long way since 2001. The images tactical commanders see today are higher-resolution and clearer. Therefore, better resolution will eliminate the CIA officer, who said that the third building looked like a mosque. Franks would not need to review the video before ordering Deptula to order his fighters to drop their bombs. That delay possibly gave Omar time to escape before the building was hit. The Air Force operates armed drones today, distancing CIA further from the tactical picture. The Kill Web will continue to integrate better technologies and remove elements of confusion and uncertainty.
Finally, let’s look at the decision-making process. Franks had to integrate the video with observations from the Predator operators, comments from the generals outside the tactical chain-of-command watching Kill TV, Deptula and his fighter planes, the SECDEF, the President, CIA officers, and his lawyer. AI will integrate all the sources of information into a composite picture. That will eliminate "information overload" and give clarity to all the tactical commanders involved. Therefore, the Kill Web will speed-up the 5F tactical model: find (identify), fix (track), fire, finish, and feedback. In the movements of the convoy noted above, the Kill Web could find opportunities to hit Omar before he ever got to the first building. In other words, the concept of command and control (C2) will change. The Kill Web will put humans in command, but the intelligent machines carrying the sensors and weapons will be in control.
At this point, you should be aware that there is a discipline called decision science. You can get degrees in this subject from several universities. Decision science is based on decision theory, which is a combination of cherry-picked elements from psychology, statistics, philosophy, and mathematics, that explains how people decide to eat nutritious vegetables (decisions made in their best interest). On the other hand, the economists have branched-out into behavioral economics, which is another decision theory that combines specific pieces of classical economics with cherry-picked elements from psychology, statistics, neuroscience, and mathematics, that explains how adolescents decide to take the challenge and eat laundry detergent packets (decisions not made in their best interest). Basically, they are rational-choice and irrational-choice decision models, respectively. The more you study these two theories, the more they make astrology look like it’s based on rigid scientific principles.
Thankfully, the military doesn’t follow the academics in this area. The Kill Web has five basic decision models, used by both the humans and the machines: emergent coordination, the greedy shooter (the weapon closest to the target takes the shot), hierarchal coordination, centralized coordination, and consensus coordination. Each of these models can be used dynamically, in different places in the Kill Web at different times and under varying conditions. Military decisions in war are trade-offs between accomplishing objectives and the loss of lives and adjacent destruction. It’s about proportionality: how much loss of life and collateral damage can be justified by the military advantage gained? The AI decision-making algorithms in the Kill Web will find the best solutions to the tactical problems, with respect to proportionality. And they can do it in seconds.
I know what you’re thinking: what about the military lawyer shadowing General Franks? Will the Kill Web have lawyer algorithms? The law is not based on logic. It's based on ethics, which is challenging to code into algorithms. Imaging algorithms can stop a fighter plane from dropping a bomb on a mosque in the picture, just like the lawyer stopped General Franks. They can see a clear field of fire with no restrictions too, but they are not perfect. Tactical commanders could declare the combat status of a battlefield to all the weapons in the Kill Web, from low to high, based on conditions. Fire-only-when-fired-upon is the lowest level. Declaring a free-fire zone (area of hostile activity) is the highest level, where everything is a valid target. But that’s problematic in a constantly changing environment. In my Army days, I met soldiers who were combat medics and combat engineers. But, I never met a combat lawyer. The person with General Franks, declaring targets as valid or invalid, looks suspiciously like a combat lawyer to me. Things have changed a lot since the Vietnam era.
I’m getting contemptuous and starting to ramble here, a clear sign that I need to wrap this up. Just understand that the AI programmers are trying to code appropriate decision models, with some level of ethics, into the algorithms running on the intelligence-gathering systems and weapons platforms. Until those are tested and validated, I’m afraid that tactical commanders will be stuck with combat lawyers by their side.
I have been reading a recent book on the Kill Chain and the Kill Web (released in 2020). It contains different and interesting perspectives on many of the topics we've discussed in this Warfare Evolution series. Next time, we’ll look at mass (quantity) versus the capability (quality) of the intelligence and weapons systems going into the Kill Web. Any new topic would be more interesting than decision science or behavioral economics.
