So,
apologies for the gap in posts. I’ve been quite busy with one things and another. I’ve been working on ideas of automatic, automated, and autonomous bombing recently, and the legal frameworks behind them. I’ve been thinking about the temporal and moral distance between legal decisions about bombing operations, and the actual carrying out of the operation itself.
I’ve been tracing bombing back from targeted killing operations currently happening, to the Cold War when Communist targets were decided upon in planning cells well in advance of any actual fighting, and also thinking forward to legal frameworks for deploying automated/autonomous weapon systems, and computerised target selection processes also.
This week
Conveniently, two articles have been written about automated strike decisions this week. The first by Ars Technica details how a data scientist, Patrick Ball from Human Rights Data Analysis Group states that leaked documents released by The Intercept show that NSA AI algorithms for finding persons of interest via mobile phone metadata, resulting in thousands of innocent people in Pakistan being killed by drones.
The authors of the piece, Christian Grothoff and J.M. Porup, also refer to US Special Operations Forces involved in Targeted Killing/Capture operations as ‘Death Squads’, which is most unfair. Death squads, and assassinations are always illegal. The fact that some of these operations take place using legal logic that is not accepted by all legal scholars does not equate US SOFs with the Nazi SS, or South American death squads.
However, by Martin Robbins at The Guardian details that Ball makes a fundamental mistake. The leaked NSA slides state that they are looking for couriers for terrorist networks, not terrorists to strike. By passing messages between terrorists, couriers have access to many terrorists. If intelligence agencies can map out their contacts, they can find a lot of terrorists. Much like how US Intelligence found Osama Bin Laden through his courier Al-Kuwati. Thus, neither the couriers, nor the thousands of people found using this analytical method, are going to be subject to strikes, but by communicating with lots of terrorists will lead to the terrorists who might then be targeted.
Anyway, on to Automated Bombing…
Bombing/Basic Encyclopaedia
I’ve been reading about the Basic/Bombing Encyclopaedia of the World. This is essentially a database of US bombing targets with World Aeronautical Chart grid references (see feature image at the top), and information about the targets, that it’s a bridge, or airfield, etc. The most thorough explanations of it I’ve found online is here, on Derek Gregory’s Geographical Imaginations blog.
I’m fascinated by this, as it is a clear example of pre-determined decisions. The pilots would drop their bombs on their specified targets, potentially without regard to their own thoughts on the target. In this sense, bombing of this type could be seen as automatic. It would be a process operating without human control. Although humans were in control of the plane, and of the actual bombing apparatus, the bomber crew was not in control of the whole bombing process once it had begun. If high-command initiated the relevant bombing mission, the specified targets were going to be struck, most likely without question from the bomber crews, and so in a sense, the mission would be operating without human control.
I’m really interested here about the legal decision-making. Along with the mission planning, and target selection, legal decisions might have been made many weeks, or months before an operation was actually carried out. Although it is highly likely they would have been regularly updated, if operations were to be launched, some legal decisions would have been made well in advance. It’s this legal temporal distance that I’m quite interested in.
Project Brass Ring
Project Brass Ring is also an interesting case study in pre-determined legal decision-making. This was the plan to deliver nuclear bombs by drone versions of b-47 Stratojet bombers. When nuclear bomb blasts became so great that planes would not be able to get out of the blast radius quick enough to ensure the safety of the pilots, and the missile technology wasn’t there yet, this was tested as a potential solution. Derek Gregory has also written some really interesting things on this too.
Again, I’m really interested in the temporal distance between decision-making, and the operations. The drone element also adds a bit of spice, particularly in how the current discourse is around the destructive power of Reapers with 500lb missiles, but not realising that these drones would have been equipped with nuclear bombs. The automated element here is interesting to me, but possibly slightly less important. With such a politically and militarily significant strike, decision making would probably have been re-evaluated close to the operation, even if the targets were originally selected long before. The temporal distance is perhaps only really relevant from the moment the drone aircraft would have been let loose.
Nuclear Submarines
Another aspect of automated nuclear bombing that I’m interested in is that from nuclear submarines. It’s a story I’ve heard a few times that should a British nuclear submarine commander be unable to listen to BBC Radio 4’s the Today programme for three days in a row, this could be taken as evidence that London had been destroyed, and that they should open their secret targeting instructions for their nuclear warheads and fire them.
These elements of pre-determination and automation in the decision-making, and performance of the orders is fascinating for my research on automated legal decisions in armed conflict, because it has the added nuclear element. The discussion of using nuclear weapons couldn’t even be properly settled by the International Court of Justice, so it gives me plenty of things to work with.
I’ve not got round to reading it yet, but apparently the Radio 4 aspect detail in Peter Hennessy’s book The Secret State: Whitehall and the Cold War, 1945–1970. I’ll have to get a copy.
Joint Targeting
The most interesting document I’ve been looking at the week is the US Joint Chiefs of Staff Joint Publication 3-60 on Joint Targeting. It even has a section (appendix B) called Targeting Automation. It is about as thorough as it could be when it’s trying to hide crucial information from both terrorists and foreign militaries. It basically says that computers do a lot of the boring work for targeting, but humans always have the last word.
Interestingly, it mentions the Modernized Integrated Database, as a ‘authoritative, all-source repository of worldwide general military and targeting data.‘. I’m speculating that perhaps, the Basic Encyclopaedia has been moved onto this database, or at least something similar. However, the document is also very clear that individual people can also be the subject to the Targeting Cycle. It seems to me that, potentially, automated bombing may have moved from being aimed at Communists, to being aimed at terrorists. In doing so, instead of striking at the heart of Soviet and Chinese abilities to make war, US automated bombing is instead, now aimed at the ability of terrorist groups to produce successfully destructive plots.
Either way, all this automated bombing I’ve found generates lots of data for my research. Hopefully I’ll be able to find some more information on these, and ideally on the legal decision-making involved in authorising targets in advance of them being struck. If anyone has any further information on these, I’d love to read it. Please get it touch, details in the ‘Contact Me’ tab.
Until next time!
The Law of Killer Robots 