Ekelhof (2019) approaches MHC by predicating it on military operation practice that both supports and constrains targets in areas of operations. Her approach relates to MHC in that it is a function of the role of designers [as with Santoni di Sio et al. (2018, 2019)] and of technical targeting procedures (as with Leveringhaus (2016)]. But Ekelhof’s approach differs in its level of abstraction by focusing on higher level of organization and operational control of the military as a supraindividual agent. This addresses the fact that the ‘autonomy’ of AWS (and of any human agent in the military, such as soldiers) is necessarily constrained by such operations. The result of these constraints is that ‘full’ autonomy, which is often construed in discussions on AWS, is not ‘full’ in the sense that is often implied (e.g., self-determining agents). Instead, it is restricted to various operational decisions and planning a priori to deployment and operations.
Ekelhof uses a case of conventional air operations to frame human involvement in operations through a dynamic targeting process. By framing the role of human agent decision-making within distributed systems, she outlines ways policymakers and theorists can determine how military planning and operations actually function. AWS can then be deployed within the context of use of these practices. Characterizing the human role in military decision-making, she outlines a six-part preoperational briefing package followed by a six-step landscape for mission execution. I briefly summarize them below.
Before the mission is undertaken, the air component receives a briefing with information on mission execution. Such briefings are often highly detailed with information such as “target location, times, and munitions”; however, they are less detailed when we consider dynamic targeting in situ (Ekelhof, 2019, p. 345). Such information is distributed to various domains of operations to specialists, who then vet and use it in more detailed planning. The executers of the mission, in this case fighter pilots, are then brought in for briefing on the mission details. The pilots take the time to study the information provided while also taking care of any last-minute preparations for execution.
The following six components can be included in the briefing package:
Target (a military compound) description consisting of all available knowledge;
A collateral damage estimation (CDE) to give operators an estimate (not certainty) of expected collateral damage (NATO 2016). In this example, the risk of collateral damage is low as long as predetermined mitigating techniques are applied;
Recommendations for the quantity, type, and mix of lethal and nonlethal weapons needed to achieve the desired effects (i.e., weaponeering solution) (USAF, 2017). In our example, these are GPS-guided munitions;
The joint desired impact, which is used as a standard to identify aim points; and
A weather forecast that, in this case, describes a night with overcast condition (clouds cover most or all of the sky) and heavy rainfall (Ekelhof, 2019, p. 345).
Coupled with other information such as the rules of engagement, the operator can then leave to execute the mission.
In situ operations
Step 1: Find
To find the target for operations, intelligence and data are required. Such targets are pre-programmed in the navigation systems of both the fighter jet and the payload. Whereas a dynamic target requires in situ data collection, the task here involves arriving at the preprogrammed “weapon’s envelop (i.e., the area within which the weapon is capable of effectively reaching the target)”. This process is displayed on the operations heads-up display (Ekelhof, 2019, p. 345).
Step 2: Fix
Once the operator arrives within the weapon’s envelope, onboard systems aim to positively identify the target confirmed during operational planning. This ensures payload delivery complies with relevant military and legal norms (e.g., NATO, 2016). In this case, targets were preplanned and confirmed. For positive target identification, the operator usually does not engage in visual confirmation; instead, they refer to onboard systems and the validation that took place during operational planning to ensure lawful engagement of the identified target. Even in this fixed case of pre-planning, the human pilot does not need to attend to anything else during this phase other than arriving within the weapon’s envelope (Ekelhof, 2019, pp. 345–346).
Step 3: Track
The operator tracks the target within the weapon’s envelope to ensure the continuity of positive identification. This also provides concurrent updates regarding the position and status of the target. In the case of a static target (e.g., a military compound), tracking is relatively straightforward and involves simply entering the weapon’s envelope as in the fix phase (Ekelhof, 2019, p. 346).
Step 4: Target
During this phase, the relevant rules of engagement (RoE), laws of armed conflict (LoAC) and other targeting rules are invoked to ensure lawful targeting and deployment. These also address other considerations, such as issues related to collateral damage and risk factors that may result to one’s own forces. In this predetermined and validated target case, the legal and military experts who vetted the target permit the pilot to simply input relevant data into the vehicle and weapons payload delivery systems to ensure proper execution. Given the visually impairing weather conditions, any further collateral damage estimates cannot be attained. Planning at pre-mission stages validated that collateral damage estimates were low and were conducted according to the norms that govern them. The human pilot thus does not actively participate or intervene beyond piloting the vehicle into the weapon’s envelope (Ekelhof, 2019, p. 346).
Step 5: Engage
Once the operator enters the designated weapon’s envelope, the onboard computer suggests to the pilot the most opportune time for releasing the payload to ensure effectiveness. This suggestion is based on its knowledge of the capabilities of the equipped weapons system. Given that the payload system itself is GPS guided, there is no need for any other forms of targeting based on visual identification. Once the pilot authorizes the release of the weapon, the munitions guide themselves to the target (Ekelhof, 2019, p. 346).
Step 6: Assess
At this point, the results from the previous stage are assessed to determine the effects of the strike. Of course, a visual assessment from the pilot can be impaired by a number of factors (weather conditions, in this case). Similarly, visual assessments of collateral damage from the vantage point of a pilot may fail to accurately reflect the efficacy of the strike and its consequences. In the case of aerial engagements such as this, ground support forces may be required for a more accurate assessment of engagement (Ekelhof, 2019, p. 346).
In considering MHC then, it appears that most (if not all) of the performance latent to each step is beyond the pilot’s control. It could be argued that this is emblematic of contemporary aerial operations more generally. While the pilot can be seen as in direct operational control of some of the operation, piloting the craft to the weapon’s envelope and engaging in weapons release, this type of control is not sufficiently meaningful. This is because the pilot lacks full “cognitive clarity and awareness” of the situation within which they are participating (Article36, 2015). The privation begs the underlying question of whether the pilot actually possesses levels of clarity and awareness sufficient enough to be deemed substantial in a meaningful way.
Discussions at the pilot level could provide some future insight both for operations employing AWS as well as modern aerial crafts. But these would converge on the operator, which is the wrong vector. Alternatively, such discussions should emphasize how the military as a supraindividual agent (i.e., an organization) can have MHC over targeting operations. Because of this, the ongoing international debate on AWS focuses overly much on the deployment stage of AWS and their relations to individual operators. In doing so, the debate attempts to locate the vector for MHC between those two agents (AWS-human). But it ignore the broader covariance of the distribution of labour between agents within a military complex that determines decision-making practices. The steps outlined above, particularly the pre-mission briefing stage with its collateral damage and proportionality assessments, are largely sidelined in these discussions.
This approach shows the need for a distributed notion of MHC to accurately account for numerous decision and measures undertaken by different agents in the broader decision-making mechanism before deployment. Different agents have different levels of control over any given vector in the process. Any sufficient conception of MHC must therefore reflect this. Of course, this does not negate the role that human operators play. Rather, it positions the role within the larger distributed network of decision-making. Here, ‘full autonomy’ is not full in the sense that is commonly intuited. It is constrained by the larger apparatus within which it forms a part.Footnote 2