Jule Meyer, University College Maastricht
Since 2011 the US Air Force has been training more drone pilots than conventional pilots. Drone warfare has a long history of technological advancement, but its adoption is a rather modern phenomenon. The practice is used in conflict zones such as Afghanistan for targeted killing and surveillance purposes. The progression of this so-called surgical warfare becomes evident when looking at earlier models. Subsequently this has led to a particularly efficient practice of warfare. To understand how the US Air Force uses drone warfare one must understand its bureaucratized infrastructure. The practice of drone warfare only becomes clear, by looking at its strict division of labour and the co-agency between humans and technology. To an equally large extent, drone warfare is highly dependent on the interaction between all operators of the drone. Therefore, the entry is based on a transcript of an airstrike in Afghanistan on the 21st of February in 2010. It offers insight into the decision-making process within drone warfare and the use of strategies such as pattern prediction.
The most obvious artifact in drone warfare are the drones themselves. Military drones used by the US Airforce in Afghanistan have resulted from years of technological advancements. With every model the drones became a little bit more surgical and efficient. More surgical because attacks increasingly costed less casualties, material damage and collateral damage. More efficient because over time material costs and human labor significantly decreased while the operations generally had a higher success rate. The first models of these Unmanned Aerial Vehicles (UAV) were already developed during WWI. An early model is the Liberty Eagle which was created by Charles Kettering in 1918 (McKinnon, 2014, p.9). Also referred to as the Kettering Bug (see picture a) the pilotless plane flew on autopilot and included aerial torpedoes. This UAV functions using a timer that is set on the estimated time that it takes the vehicle to fly to its destination. Once landed, it shuts off the engine, releases its wings, and drops the UAV with the explosives on the ground to explode (ibid.). Thus, each attack comes at the cost of the total destruction of the vehicle. Despite Kettering’s ambitions to further improve the Bug, he never managed to solve the problem of self-destruction and the inaccurate timing under different weather conditions. Accordingly, Kettering’s Bug was never used in combat.
Technicians of the so-called operation Aphrodite took a different approach to building drones. They recycled old planes into explosive drones steered by a following plane allowing for a more exact timing of the bombs. Another great influence on modern-day military drones were the Radioplanes from 1935, controllable from the ground and fully recoverable. Drone technology further improved during the Cold War in the 1960s. This led to the development of the Lightning Bug, a drone used in Vietnam, China, and North Korea mainly for surveillance purposes. It was remotely controlled, auto piloted, took pictures, and could be recovered after an attack. However, the modern Predator and Reaper drones employed by the Air Force nowadays, resulted from the Albatross prototype developed in the 1970s. The feature of recording long-term surveillance footage is what ultimately made it successful. The prototype also overcame prior problems of weather resistance and could hold more weaponry than any other model. Equipped with such features, it successfully managed to sustain itself until today and has led to the modern Reaper and Predator drones that are employed to this day in Afghanistan.
The Predator and Reaper Drone
The Predator drone can operate for up to 24 hours at 25,000 feet (Cloud, 2011). It is a medium-altitude, propeller-driven drone, equipped with surveillance technologies and weaponry for attacks (see picture b.). It holds cameras and infra-red devices for night vision. It can carry up to two hellfire missiles. The Reaper drone can hold more weaponry and includes satellite-guided bombs and four missiles.
The historiography of drones is insofar important as it shows that this practice was developed to be as surgical and clean as possible (Gregory, 2011b). Only when total self-destruction could be prevented, drones were employed in warfare. Before, their usage was not deemed efficient. In its very nature, the practice of drone warfare exists to make general warfare more surgical and productive, costing fewer lives and little effort. In other words, while it might reduce the number of deaths in one camp it increases casualties in the other. The aim is to create an instrument allowing to destruct a specific target without the political risk of losing US soldiers. Subsequently, drone warfare has evolved into the silent and often hidden practice of warfare it is today. It projects power without vulnerability through sub-practices that will be explained in the following sections.
The first key aspect of drone warfare is its bureaucratization. The bureaucratic infrastructure of US drone warfare involves the Central Intelligence Agency (CIA), the Joint Special Operations Command (JSOC), and the Air Force. They cooperate with the president of the USA. Insights into the selection of targets by the executive branch remain limited due to the high level of confidentiality. Nevertheless, the CIA is responsible for the security intelligence and hence proposes and initiates the organization of an airstrike (Boyle, 2020). The process of targeted killing is officially referred to as Find, Fix and Finish (ibid.). The first step (I. Find) refers to identifying and locating a target. This happens through multiple agencies such as the CIA, JSOC, NSA, but also through sources such as media reports, or foreign governments. Based on the information from these sources the target environment is narrowed down, and the battlefield awareness grows (ibid). The operators of the predator or reaper drones then locate the target and forward the information to the CIA (II. fixed). The CIA then creates a so-called strike package summarizing the intelligence information and the location information. Afterwards, the president receives the complete strike package. It informs him about the target, the feasibility of the operation, and includes an estimation of potential civilian casualties. The third step of the process (III. finish) includes so-called killing boxes. They refer to the physical space in which the target is located. It reconstructs the civilians and objects around the target. Lawyers will then decide whether or not an operation can be granted. This decision is based on the level of evidence in favor of the operation. Operations are only granted if there are two intelligence documents pinpointing to the target without any contradictory evidence. Finally, the president approves the mission and authorizes the CIA to carry it out. The CIA then contacts the Air Force and their drone pilots to give them a time frame of sixty days for the completion of the mission.
The exact processes vary slightly between the different presidential administrations. Even though there is an established procedure for drone strikes, it leaves enough freedom for each presidential administration to decide on the exact one followed. The Obama administration, for example, used more lawyers to decide on the permissibility of an airstrike in comparison to Trump. Obama would also personally sign off all high-profile operations. Trump, instead, delegated most of the work and merely used lawyers for high-profile cases. Additionally, he personally only decided about whether to engage in a new country or not. He also abolished the requirement of a target being a continuing or imminent threat to Americans. Moreover, the required burden of proof was lower under Trump in comparison to Obama. He allowed the targeting of people with a ‘reasonable certainty’ of being a security threat, in contrast to a ‘near certainty’.
The bureaucratization of drone warfare is crucial, as it creates the impression that any operation is directly permitted by the executive branch (hence the government in power) even though the process is highly dispersed, involving multiple agencies. The idea that the president issues one centralized killing list would thus be an inaccurate description. Although such a killing list does exist, it is less simple than the term implies. Whether a person poses a security threat or not highly depends on the agency’s perception, hence creating a wide variety of documented names. To resolve this issue, so-called Terror Tuesday meetings are held. Under the Obama administration, the president had final authority over the names on the ‘killing list’. Everyone on the list would also be listed in a master database that includes names, biographical details, suspect locations, and operational details. Thus, this bureaucratization process is termed bureaucratized killing (Asaro, 2013). While the responsibility does lie with the US government, and the president themselves, it should also be understood as trickling down to the operational level. This phenomenon is further explained below.
Division of labour: crafting a view and created subjectivity
Another part of drone warfare is its strict division of labour. Some scholars term this the chain of killing, which incorporates the different positions that exist in operating a drone (Gregory, 2011a; Cockburn, 2016). Taking the US drone engagement in Afghanistan as an example, overall, around 185 people operate one drone. 59 of them work in Afghanistan, 43 in a US base in Nevada, and 83 are responsible for the analysis and signal intelligence in other US bases (Gregory, 2011a). More broadly they can be separated into five main bodies (Elish, 2017). First, there are the pilots who control the aircraft and the weaponry. Second, there are the sensor operators, controlling the sensors onboard of the drone. Third, there is the ground crew taking care of the maintenance of the aircraft, sensor, and all other equipment. Fourth, there are people responsible for the analysis of information from the sensors. Lastly, there is the mission intelligence officer taking care of the verification of information and assessing the overall mission planning. The commanding officer is responsible for the mission. The division of labour is reflected in a difference in status, skills, accessible data, and different physical locations. All five bodies are elaborated on in the following sections.
A central part of the piloting happens via satellite communication at Ground Control Stations (GCS) in the US. The GCS taking care of the high-profile missions is the US primary Creech Air Force Base in Nevada (Cole, 2012). The GCS is far away from where the drones are actually flying. At the GCS there are different cockpits for each drone, as can be seen in picture c. The cockpit holds two seats, one for the pilot and one for the sensor operator. Both work next to each other but focus on different aspects of the drone. The pilot and the sensor operator sit in front of several screens, keyboards, menus, and toolbars that show the different information that the drone is gathering through its visual and non-visual sensors. Their work starts with configuring the instructed information into the Reaper or Predator drone as coordinates (Cullen, 2011). The pilot is an officer in command of the crew and the aircraft. He is responsible for steering the vehicle to the correct location, focusing on air traffic maps, cartographic information, warning systems, and communication channels to the other entities involved in the operation of the drone. Based on the available information the pilot steers the drone and controls the weapons. Additionally, he takes care of the communication with the commanding officer granting the final authorization for airstrikes. The information on their targets stems from the soldiers on the ground. Thus, the pilot is a key point of communication between the aircraft, the aircrew, and the ground forces. Although the drone flies partially automated, the drone pilot still needs to manually fly the vehicle in cases of bad weather conditions, special strategic requirements, or during system failures.
The sensor operator is responsible for the sensor system which includes an image amplifier, a daylight camera, a laser marker, and a laser illuminator (Franz and Queisner, 2011). Most importantly, the sensor operator takes care of adjusting the sensor system correctly and directing the camera at the specified target. The camera is held in the sensor ball attached to the drone. One sensor operator oversees up to 12 ground location video feeds and 30 different video streams from the drone (Asaro, 2013). They are also required to effectively report and identify targets or threats to the aircrew. This highly demanding job is executed by image analysts having received special training in interpreting images. Nevertheless, they are not officers with experience in flying drones (Cullen, 2011).
A third key member of the GCS is the mission intelligence coordinator, mediating between the intelligence analysts and their entities. The coordinator communicates information to the other members and helps interpreting and understanding the gathered information (Asaro, 2013). The intelligence coordinator can be understood as the interface between archived computer databases and human intelligence analysts (ibid.). He holds the main responsibility for the communication between drone pilot, sensor operator, ground troops, and analysts. Not part of the GCS are the dozens of analysts that are responsible for the verification, analysis, and interpretation of the video material. They are spread out over the US among different Air Reserves in Arizona, Arkansas, California, Florida, Missouri, Nevada, New Mexico, New York, North Dakota, Ohio, Pennsylvania, Tennessee, Texas, and Virginia (Elish, 2017).
The described division of labour leads to the following aspects: bureaucratized killing, crafting a view and subjectivity. Just like on the organisational level the killing process is also bureaucratized in its operational parts. The killing process is distributed amongst several different entities all with responsibilities and expertise that the others do not share and hence cannot oversee, nor control. One can thus speak of a bureaucratization since the working of the entities are co-dependent and hence restrained by the work of another. It is not just the drone pilot who is responsible for the deaths caused by an airstrike, but it is the responsibility of all entities. Generally, this allows for a flexible transferal of moral responsibility between them. While the pilot oversees route and weaponry, the sensor operator monitors the image amplifier, camera, laser marker, and laser illuminators. The intelligence commander oversees particular intelligence information passed on to them from the analysts, ground forces, or higher officers. Finally, the analyst oversees real-time video footage. The ground forces maintain the drone. The professionals from the different entities are experts in the data they monitor but know little about the fields of their colleagues. Additionally, in preparation for a mission, all entities have to go through a considerable amount of administrative work, doing phone calls, having meetings, proposing strategies, and joining information briefings.
Because drone warfare is based on a locally dispersed network, scholars speak of the process of crafting a view. To see what the drone is seeing is not as simple as one might assume. Some might argue that the 185 people operating a drone all only see a fragment of what the drone sees due to their different professional backgrounds and the different types of data they access. Although all the cameras and sensors relevant to the practice are centrally attached to the sensor ball, seeing happens in a highly decentralized way. Scholars such as Franz and Queisner (2011, p.126) speak of crafting a view when all the different bodies come together and aggregate their data. Some might argue that none of the involved bodies can see what the drone sees, they only gain a fragment of the picture. Only together they can combine their individual pieces into the big picture. Even the intelligence coordinator who is closest to overseeing the entire picture will never be able to gain insight into everything due to the sheer amount of data that the drone is gathering. But most importantly, because even he is merely one part of the bureaucratized practice.
Furthermore, scholars identify that due to the division of labour, subjectivity is another key part of the practice. Everyone is reliant on each other’s subjective information making the work of the intelligence coordinator vital. The required coordination effort is high since all of the different bodies referred to above have their specific expertise. The pilot is the only one with a background in flying UAVs, while the sensor operator is very skilled at the technological aspects of the sensors. The intelligence commander has an exclusive position as they have insight into all the communication and can access archived information that does not stem from the observed situation. Lastly, only the video analysts see the whole of the video material gathered by the drone. Consequently, the operation of the drone depends on the coordination of all of the different bodies operating the drone. While the spherical distance of these entities is bridged through constant communication channels, poor communication skills bear a large risk to the success of the missions. Thus, each of the entities sees something completely different making their viewpoint highly subjective. Only the accumulation of all individual subjective perceptions in the process, result in a complete drone operation. This accumulation, however, not only depends on human interaction but also on the interaction between humans and technology as discussed next.
Another part of drone warfare is the close interconnection between human and technologies. Any of the entities cannot do their work without their technological counterparts. Furthermore, in many ways they just respond to information which is predetermined by the system. As a result, the agency and decision-making ability of the human pilot is bound to the automated systems they are working with. Scholars, therefore, speak of co-agency as the pilot is always just reacting to prior decisions made by automated systems (Franz and Queisner, 2011). Essentially drone warfare works through the interplay between humans (operators), machines (aircraft and missiles), and the sensors (camera and video). The flying of the drone does not happen completely automated but as said earlier requires human and automated labour, this contrasts the term ‘unmanned aerial vehicle’.
For pilots and sensor operators, this means working in highly technologized cockpits. These include many different technologies that help the pilot to steer the vehicle and the sensor operator to oversee the information gathered by the sensors. Picture g. shows the anatomy of their working station and all of the technological equipment they are reliant upon. The pilot and drone operators thus work with and alongside the technology. Their decisions depend on the information that the systems show them. The pilot will fly to the coordinates in the system and the sensor operator reacts to trouble warnings of the sensor systems. Their main function is to be able to read, understand, and react to the information that is provided by the system. Furthermore, in the case of the video analysts, the interplay between system and human is highly relevant. In their work algorithms filter out information that is particularly relevant to them (Weber and Suchman, 2016). In case the algorithm fails to identify important video material the analysts will not see it either. Thus, the agency of the analyst is highly interlinked and dependent on the algorithm at (Franz and Queisner, 2011).
The key part of drone warfare is the interaction between the different entities involved on operating a drone. As a matter of fact, to understand how drone operators decide to pursue an air strike it is insufficient to look only at the video stream since all the different operators have access to different streams and their own niche information. Thus, it follows that only in their interaction the full picture from the drone is visible. This was earlier described as crafting a view. Representative of this process is for example a transcription of their interaction. The following part will thus deal with a transcript of the interaction between ground troops, drone operators (Nevada) and video analysers (Florida) during an airstrike in Afghanistan from 2010. The interaction serves as an example of how the drone-based information leads to strike decisions.
The description is based on an interaction occurring during a drone strike in Uruzgan, Afghanistan on the 21st of February in 2010. On that day, the US Air Force kills between 15 and 23 civilians based on false premises. During the operation the drone pilot identifies worrisome objects, potential weapons, in the battlespace. The team assumes them to be held by males within the Afghan military age range, who are driving three vans. The operators survey the targets for three hours before they decide to strike. Later, analysts confirm that the strike in fact, targeted civilians including children and women (Boon and Weaver, 2010).
The official executive summary of the incident states that three vehicles were surveyed due the assumed weapons and suspicious tactical manoeuvring (McChrystal, 2010, pp.2-3). Tactical manoeuvring of vehicles is among others, one aspect under which a vehicle can become a possible target (ibid., p.2). Additionally, the vehicles were assumed to be flanking and people close to the vehicle were observed to be engaged in a scuffle which could have been shielding something (ibid.). Lastly the video analysers identify children in the vehicle but later correct their identification into adolescents based on the opinion of the pilot. Yet, the correction happens without any defined age range. Due to an assumed danger of the weapons for ground forces and their tactical behaviour, the drone operators decide to pursue three air strikes (ibid., p.3). The first missile hits the first vehicle, the second one strikes the third one. When zooming-in pilots of surrounding helicopters confirm that the occupants looked like they were holding weapons that were ready to fire. The third missile targets the second vehicle (ibid., p.4). Furthermore, rockets were fired onto the people running from the scene. The pilots later identify them as females wearing burqas (ibid.). Once identified they immediately cease their engagement (ibid). The report states that at least three females and one child were near the vehicles (ibid.). Officially the task for this operation was to disrupt the enemy, deny safe haven and freedom of movement while avoiding collateral damage (ibid., p.6). “The vehicles were engaged under self-defense rules” (ibid., p.7). As commanders always have the right to self-defence in case of a hostile act or demonstrated hostile intent the strikes were in accordance with the official rules of engagement (ROE). A hostile intent is per definition the threat of imminent use of force against the US (ibid., p.8). No warnings were given before the engagement which is also consistent with the ROE (ibid. p.18).
When answering the question if the behaviour posed a direct threat to the US the executive summary states that the assessment in the given situation was reasonable but a re-examination of the information should have happened (ibid., p.17). The movement that was identified as a potential flanking manoeuvre, the low number of weapons and the presence of children should have led to a re-examination of the situation. Responsibility is directed to the leaders of the operations. There is no evidence that any of the individuals killed or wounded are associated with anti-Afghan forces (ibid., p.24). Condolence payments were made by SFC to those who were injured, killed or had their equipment damaged. Family members of those that died were identified and they received 5,000 per death (ibid.). For the 23 deaths a total sum of 115,000 dollars was paid (ibid.). 3,000 were paid to those injured (ibid.). The total condolence fund paid 175,000 dollars (ibid.). This particular case offers insights into how a target comes to be killed by drones of the Air Force. The aspect of pattern prediction and the decision-making process will now be further explained.
Objectification and pattern predicitions
To understand the description and reading of the official transcript the concept of necropolitics and biopolitics is useful (Allinson, 2015). Per definition necropolitics refers to an apparatus of racial distinction that presents humans as objects, or in this case as targets (ibid., p.117). The concept of necropolitics aims to explain the power relations underlying the social construction of race (ibid.). It links to Foucault’s (ibid., p.118) concept of biopolitics which refers to a kind of politics in which the sovereign power defines who within population gets to live. A distinction is drawn between those worthy to live and those who must die, death is seen as increasing the health of the remaining population (Su Rasmussen, 2011).
The transcript provides inside into biopolitics of the practice of drone warfare. The observed individuals are deemed as justifiable targets once they fit into a prior defined pattern determined by the drone operators. In this case it is the assumption that they are holding weapons, in combination with the tactical movement and being males of military age that creates a pattern. Subsequently the fulfilment of the pattern leads them to believe that the targets intent to threaten the US. Henceforth, this gives them the necessary ground to decide to pursue the airstrike and kill them. What terms like child or adolescent specifically refer to, or if they are actually real is of second importance to the operators. The pattern identification is focused on potentiality. The potentiality of the individuals to become a threat to ground troops. The operators thus quickly dismiss the potentiality that the individuals are civilians for the potentiality of them being a threat. In this instance they take a group of people gathering around a vehicle as tactical movement, and the children as adolescents. Their objective is to find pre-defined categories and patterns in the environment they observe. This comes at the expense of a sensitivity for alternative explanations. The following excerpt shows that the sensor operator was fast to dismiss a possible un-threatening character of the surveyed individuals that was indicated by the ground commander before.
“When the Screeners first identified children, the Predator sensor responds ‘bullsh*t, where?’ The Predator pilot follows with ‘at least one child… Really? assisting the MAM, uh, that means he’s guilty/yeah review that (expletive deleted) … why didn’t he say possible child, why are they so quick to call (expletive deleted) kids but not to call (expletive deleted) a rifle.’…The Predator sensor says on internal comms, ‘I really doubt that children call, man I really (expletive deleted) hate that” (Allinson, 2015, p.15).
Therefore, part of this pattern predictions is also intention prediction. If individuals fit into a specified pattern, they become targets who are to be killed and assumed to be intentionally threatening US soldiers. Some scholars (Wilcox, 2016; Allison, 2015; Malaviya, 2020) have therefore termed this target construction leading to a production of killable bodies. This aspect also shows that drone warfare creates virtual battlespaces on the screens of their workstations. From their remote-control stations, the operators define and create the battle ground for example by categorizing individuals not as children but as adolescents and thus to them as threatening. The battlespace is detached from the country it is placed in and its surrounding. It primarily exists through the lenses of the drone, everyone outside of it is irrelevant. Cultural aspects such as social habits or clothing styles are disregarded, which leads in this example to the description of traditional afghan clothing as a “man dress” (Allinson, 2015, p.17).
“does it look like he is ho’n something across his chest. It’s what they’ve been doing here lately, they wrap their *expletive* in their man dresses so you can’t PID it.’ Then on the radio to [redacted] he says ‘looks like the dismounted pax on the hilux pickup on the east side is carrying something, but we cannot PID what it is at this time but he is carrying something.’ After the Predator crew prompted them twice in mIRC, the screeners call out a possible weapon and then ask the crew to go white hot to get a better look. The response from the sensor operator is ‘white hot is not going to give us anything better, that truck would make a beautiful target” (ibid.).
Furthermore, the objectification in this case comes along side an idealization of the target as beautiful. It thus associates the trucks to be beautiful as targets, which raises the question what the exact criteria of a beautiful target are. In any case, this description shows the kind of objectifying framework that operators base their opinions on. According to the ROE targets should primary be identified as somewhat threatening but not to be beautiful. The destructive character of air strikes is completely lost in this description. Scholars (Parks and Kaplan, 2017, p.134) speak of the alienation from the consequences of the pilot’s actions as the vertical distance (or verticality) which first, exist between the drone and the individuals on the ground. Secondly, it also exists between the operators sitting in Nevada and the drones in Afghanistan. Lastly, the operators and the individuals on the ground are also vertically distant. The verticality is psychological, but also cultural as described beforehand.
Diffusion of labour
In stark contrast to the explanation of the strict division of labour from beforehand, stands the following excerpt of the transcript:
“additionally, on several occasions the Predator crew identified weapons on their own, independent of the screener’s assessment. At 0511D the Predator makes a radio call to [redacted]… They prompted the screeners in mIRC [the online chat relay system used by the US military] to let them know if they could PID any weapons, but at 0518D, the screeners reported that they could not confirm any weapons.” (Allinson, 2015, p.16)
This extract shows how the false assumption that the individuals had weapons came about. The drone operator claimed to have identified weapons independent of the analysing crew (screeners). Although this falls outside of the field of responsibility of the pilot, they claim to be right despite contradictory evidence from the screeners. The strict division of labour is diffused by the fast pace, real time work environment and a false sense of the pilot to be able to have a full picture of the situation. The video footage grants him a clear vision of what is happening, but he is insensitive to the subjectivity of his own work and knowledge.
Part of drone warfare therefore is its surgical manner, bureaucratized killing processes, inherent subjectivity, crafting a view, co-agency, the production of killable bodies, pattern prediction and in contrast to what was mentioned earlier a diffusion of labour. Nevertheless, drone warfare includes many more aspects that are outside of the scope of this entry which only focuses on organizational and operational aspects. Other parts of drone warfare are for example the psychological effects of flying drones in comparison to planes. But also, the experience of drones from the people living under them. The here described processes cover the structural basis of drone warfare insights into the psychological effects can be gathered from documentaries such as National Bird by Sonia Kennebeck (2016).
Jule Meyer studies liberal arts and sciences (BA) at University College Maastricht, focusing on international relations. She wrote her entry during an Applied Research Internship at the Institute for International Law of Peace and Armed Conflict at the Ruhr-Universität Bochum (March 2021)
Allinson, J., 2015. The Necropolitics of Drones. International Political Sociology, 9(2), pp.113–127.
Asaro, P.M., 2013. The labor of surveillance and bureaucratized killing: new subjectivities of military drone operators. Social Semiotics, 23(2), pp.196–224.
Boon, J. and Weaver, M., 2010. Afghan ministers voice anger as civilians killed in Nato air strike. The Guardian, [online] 22 February. Available at: https://www.theguardian.com/world/2010/feb/22/nato-airstrike-deaths [Accessed 01 February 2021].
Boyle, M.J., 2020. The drone age: how drone technology will change war and peace. New York, NY: Oxford University Press.
Cloud, D., 2011. Anatomy of an Afghan war tragedy. Los Angeles Times, [online] 10 April. Available at: https://www.latimes.com/archives/la-xpm-2011-apr-10-la-fg-afghanistan-drone-20110410-story.html [Accessed 01 February 2021].
Cockburn, A., 2016. Kill chain: the rise of the high-tech assassins. New York: Picador.
Cole, C., 2012. Drone wars briefing. Drone wars UK: Oxford.
Cullen, T., 2011. The MQ-9 Reaper Remotely Piloted Aircraft: Humans and Machines in Action. Massachusetts Institute of Technology. Available at: https://dspace.mit.edu/handle/1721.1/80249
[Accessed 26 February 2021].
Elish, M.C., 2017. Remote split: A history of US drone operations and the distributed labor of war. Science, Technology, & Human Values [e-journal], 42(6), pp.1100–1131, 10.1177/0162243917731523.
Franz, N. and Queisner, M., 2018. The Actors Are Leaving the Control Station. The Crisis of Cooperation in Image-guided Drone Warfare, In: L. Feiersinger, K. Friedrich & M. Queisner eds. 2018, Image action space: Situating the screen in visual practice. Berlin/Boston: Walter de Gruyter GmbH. pp. 115-132.
Gregory, D., 2011a. From a view to a kill: Drones and late modern war. Theory, culture & society, [e-journal] 28, pp.188-215. 10.1177/0263276411423027.
Gregory, D., 2011b. The everywhere war. The Geographical Journal, [e-journal] 177(3), pp.238–250. 10.1111/j.1475-4959.2011.00426.x.
[Kettering Aerial Torpedo “Bug.”], 2015, [image online] National Museum of the United States Air ForceTM. Available at: https://www.nationalmuseum.af.mil/Visit/Museum-Exhibits/Fact-Sheets/Display/Article/198095/kettering-aerial-torpedo-bug/ [Accessed 20 January 2021].
Malaviya, S., 2020. Digitising the Virtual: Movement and Relations in Drone Warfare. Millennium: Journal of International Studies, [e-journal] 49(1), pp.80–104, 10.1177/0305829820971694.
McChrystal, G., 2010. AR 15-6 Investigation, 21. February 2010 U.S. air-to-ground engagement in the vicinity of Shahidi Hassas, Uruzgan District. American Civil Liberties Union [online]. Available at: https://www.aclu.org/drone-foia-department-defense-uruzgan-investigation-documents [Accessed 20 January 2021].
McKinnon, G., 2014. The Birth of a Drone Nation: American Unmanned Aerial Vehicles Since 1917. Louisiana State University. Available at: https://digitalcommons.lsu.edu/gradschool_theses/403 [Accessed 26 February 2021].
Parks, L. and Kaplan, C. eds., 2017. Life in the age of drone warfare. Durham: Duke University Press.
[Remotely Piloted Aircraft (RPA) Sensor Operator.], 2017 [image online] Available at: https://www.airforce.com/careers/detail/remotely-piloted-aircraft-rpa-sensor-operator [Accessed 6 January 2021].
Su Rasmussen, K., 2011. Foucault’s Genealogy of Racism. Theory, Culture and Society, 28(5), pp.34–51.
Weber, J. and Suchman, L., 2016. Autonomous weapons systems: Law, ethics, policy. In N. Bhuta, S. Beck, R. Geis, H.-Y Liu, and C. Kreß, eds. 2016. Autonomous Weapons Systems. Cambridge: Cambridge University Press, pp.75-102.
Wilcox, L., 2016. Embodying algorithmic war: Gender, race, and the posthuman in drone warfare. Security Dialogue, 48(1), pp.11–28.