Ontologies of the Wayward Drone - Part II

A Salvage Operation:

Ontologies of the Drone

Amplifying Expertise

Ontologies of the Drone

Consider the rudder, excessively flailing on the luckless Global Hawk -- the renegade part that was the cause of the drone's undoing. As the hulking, ungainly vehicle roars through the sky, resembling a strange sea creature with no eyes on its disproportionately large head and no features on its vast, smooth stretch of luminous white skin, the rudder swerves back and forth at the ass end, lodged within the plane's fin. Its smooth, curved form is the material outcome of the need to harness the properties of moving air: to maximize the efficiency of the interactions between air and the solid bodies that move through it. The uncontrollable, persistent flapping of this tiny part -- whose weakness was unfortunately not detected before the fatal flight -- created, over time, a sufficient degree of destabilization to cripple the enormous plane and send it plummeting to earth.

Essentially, this part is rather dumb: a surface that awaits control. The control is provided by an actuator. The rudder is attached to its output hub and secured in place with hinges. An actuator is a device that converts a control signal to a physical action. In this case, it is a motor that drives the control surface of the rudder.

The actuator-rudder mechanism cannot function alone. Without the input of information or power it does nothing. Yet in certain cases it might be regarded as an autonomous entity. A human observer might isolate the mechanism, regard it in terms of its material and functional specificity, marvel at the contours of its design. The task that it must perform is relatively simple when understood at the level of each scale of its operation. At each of these levels, the action remains the same, though the task might be understood differently. At the most basic scale, its job is to move back and forth along a determined range of motion in accordance with received instruction. At a larger scale, its job is to change the shape of the tail fin's surface and subsequently vary the amount of force generated by it. At a still larger scale, its job is to control movement of the plane about its vertical axis -- to change the horizontal direction in which the nose is pointing.

These tasks cannot be accomplished alone. In order for the plane to maneuver in space, the rudder must work in conjunction with the plane's other directional control surfaces. It must cooperate with the elevators, the horizontal control surfaces on the tail section that control pitch. It must cooperate with the ailerons, the control surfaces attached to the wings that impart roll. The cooperation occurs across a number of scales. Actuators drive control platforms at their own local scale (such as at tail or wing), in ways that alter their aerodynamic features, and these movements, in turn, alter the aerodynamic characteristics of the larger-scale platform of the plane. Larger-scale actuators may also effect smaller-scale platforms: the propulsion system, including engines and propellers, is an actuator that effects the action of the fin. The overall cooperative job is to provide stability for the aircraft -- to keep it straight in flight.

In order for the aerodynamic characteristics of their platforms to be altered, the motors of the actuators must drive their control surfaces in accordance with received operational commands. Data must be sent to them, singly and collectively. In order for the correct information to be input, environmental conditions must be sensed. Working in cooperation with the actuator-platform affiliation, then, are affiliated actors whose functions are that of sensing, processing, and communicating the relevant information.

A sensor detects and measures a change in state, within a medium that might be mechanical, electrical, magnetic, or chemical. Its measurements might involve rate gyros, direction, speed, altitude, or distance to nearby objects. Sensors, too, have their platforms, which might take the form of pressure transducers, range finders, digital compasses, or electro-optical and infrared cameras.

Once these environmental conditions are sensed, the data is processed by the flight computers. The necessary information is exchanged via transmitters and receivers. The flight computers transmit relevant information to operating crews and other teams of actors who might be involved with launch and recovery elements, maintenance and logistical support systems, mission command and control, or image processing and dissemination, at various locations, however embedded in ground or air. The correct data, once assembled into coherent control signals, instructs the actuators to drive their respective control platforms. The plane is steered, its relative position, speed, and attitude adjusted in accordance with this instruction, and a cohesive flight is (hopefully) produced.

The plane's actuator-platform affiliations, then, are only able to perform accurately by connecting to affiliations that have the capacity to sense, process, and link. Affiliations, in general, are only able to perform their tasks by connecting to affiliations whose functions they must operate in accordance with. They have properties that remain stable, but they also have capacities that are only fully activated in cooperation. A property is an internal characteristic, whereas a capacity manifests an external contingency -- a characteristic that emerges in a shared field, and that relies upon the properties and capacities of other actors with which it works in conjunction. In this way an affiliation is dependent upon the kinds of couplings it seeks out or affords. Because of the rudder's properties -- its material quality, density, curvature, and texture -- it has the capacity to deflect and contour the air that flows into it. When coupled with a motor that has the capacity to move it, the rudder-actuator affiliation is endowed with the more complex property of back and forth motion. When it is coupled with an instructor capable of commanding it, the mechanism activates its potential to change the shape of the tail fin's surface. It now achieves its capacity to vary the amount of force it generates. When working in conjunction with the plane's other directional control mechanisms, with their various capacities to vary force levels (and that the properties of moving air can be counted upon to produce reliable pressure), it has the capacity to control the movement of the plane about its vertical axis.

Understood in this way, the affiliation is a practice as much as a thing -- a system as much as an object-configuration. It is distributed yet consolidated. It is a heterogeneous entity, characterized by relations of both interiority and exteriority (properties and capacities), that is always in excess of its boundaries; yet at the same time, when stabilized and discerned at a shared scale of operation and embroiled in a mutual process of individuation, it can be regarded as uniform and discrete. To regard the affiliation in terms of its performative functions -- the roles that actors perform in affiliations at various scales of operation -- is to accommodate these seemingly contradictory understandings.

Understood in this way, the UAV -- or as it is increasingly coming to be known, the UAS (Unmanned Aircraft System) -- is an affiliation of entities and practices, components and systems that includes sensors, computers, communicators, actuators, controls, and platforms. Even though they operate at various scales and levels of complexity, these components and systems are somehow integrated into coherent, stable formations that can be replicated and relied upon, and the necessary expertise and skill emerges for operation. This occurs through the emergence and harnessing of commonalities -- common frameworks and parts. In order for a context to be created for the information, communication among components facilitated, and inferences from the data drawn, the overall system must function in terms of common standards. The tasks performed, whether at the small scale of control surfaces or the large scale of control infrastructures, are only accomplished by linking to other affiliations and functioning in accordance with them in the terms of common programs.

It is a matter of the modality of the linking. It is a process of bonding, synchronization, calibration, and agreement that, occurring across components and systems functioning at different speeds, scales, magnitudes, and levels of complexity, does not simply involve a conventional relational structure. The difficult question is not how actors relate to one another as separate entities, but how they gather together to stabilize in cohesive wholes that are more than the sum of their parts. It is a matter of how, once sufficiently stabilized, they replicate, become redundant and standardize, at various scales, across various platforms of endeavor.

The functions of sensing, processing, communicating, and actuating are distributed, shared, and consolidated across a number of ontological platforms. Many biological and machinic assemblages perform all of them. At the most basic level, all component actors are sensors and transmitters of energy. They transmit and absorb electro-chemical signals and electric or nervous impulses, and they emit and receive vibrations whose different frequencies they process in terms of language. These transmissions traverse the kinaesthetic and kinetic dimensions of experience: between movement that is sensed, experienced, and perceived internally and external environmental forces like gravity, momentum, speed, and phrasing. Transmitting agencies filter and calibrate flows, modifying one another at the level of affect, rhythm, and code, in ways that increase or diminish their ability to act, apprehend, relate, and materially exist. Within the exchanges and modulation of these flows, they acquire rhythm and articulation. The foundational structure of this relationality is not primarily based on difference. Actors may consolidate as discrete entities, yet they also vibrate in terms of constrained transmissions and modulated thresholds, however approached, attained, or crossed, at various frequencies or rhythms. Relationality involves the correspondence of elements, yet also involves the limitation of flows.

Again, in terms of the "recovery" operation, it is a matter of suspending recourse to conventional ontological categories and instead, regarding agencies in terms of their performative functions -- the roles they perform in affiliations at various scales of operation. Functions are always consolidated and embedded in the specificities of actors, which might be human, institutional, technological, spatial, or atmospheric in nature -- a fuselage, microprocessor, rudder, communication link, tag, engine, crew member, military base, controller, program. These actors achieve a level of discreteness -- functional and agential specificity -- in concert with other actors that rely on them. But the recovery operation is one of holding specificity and the distribution together -- placing consolidation and multiplicity, part and practice, component and system, together on the same analytical plane. The rudder's direction in manned aircraft was once manipulated with the movement of a pair of foot pedals by a pilot. While most of the Global Hawk's operations are now the result of programming and commanding the autopilot's computers -- a rudder command is sent encrypted via fiber optic overseas cable and satellite, and takes about three seconds to reach the plane -- this does not mean that the agency of the pilot has been fully replaced by a program or relocated in one human crew member at one site. (It takes a crew of around 60 people even to keep a small UAS like the Predator operating.) It is a matter of looking at the distribution and embeddedness of the piloting function: of understanding how its capacities have been redistributed in sensing, processing, and actuating affiliations at various scales and consolidated in new clusters of ontological significance.

As with all affiliations, the further one zooms in, the more complex matters become. Often the necessity of the zooming is only revealed in the advent of the failure, the crash providing the impetus for the probe.


At the onset of the Global Hawk catastrophe, the investigation was set into motion. It located the rudder-actuator: its failure, its uncontrolled, excessive flapping, was pinpointed as the agent of the plane's demise. The fault could have been located deep within the actuator itself -- in the input shaft/crank mechanism, electromagnetic brake, stainless steel gearing, or servo valve -- or in a defect of its lacquered, controlled surface, flailing in the sky. It could have been located in the canals that regulate the coursing of fluid or the flow of wind. As it is not just a matter of the functional makeup of the affiliation itself, but also that of the affiliations within which it is embedded and with which it functions in accordance, the fault could have been located in the agency that commanded the actuator.

These agencies, as with all component agencies of the drone, can be material formations consolidated in space or scalar zones distributed in time. They can exist as configurative instantiations or transmissive fluctuations: elements related or flows constrained. The UAV is a rigid flying platform, yet it is a dynamical system that is also defined by the atmospheric, technological, and institutional systems that it moves through -- air, informational transfers, protocols. The faulty agency of the actuator's command, then, could have been located in the signal that was sent to it, in the instructions themselves, in the program through which these instructions were mobilized, or in the agency that programmed them, however located in the information flows within the vehicle or between the vehicle and its larger command networks.

The functioning of the actuator's output hub is monitored by a sensor connected directly to it. The sensor provides a position feedback signal. Because the loosening of the rudder-actuator complex was not detected, the fault could lay in the sensor's performance. There could have been a loss of accuracy through a small bias in measurements, or a slow-drifting of them. Or, the sensor could have become stuck on one particular value. These faulty measurements alter the measurements required by controllers, and, depending on their severity, may scale up to degrade the UAV feedback loop. The UAS, like all complex affiliations, is built of a collection of components functioning at different levels of physical scale and organization. The output of each level provides units of assembly for the next level up the scale, in ways that can also rebound back to affect lower levels. The sensor's data could have been transmitted and processed correctly at one scale but processed incorrectly at another -- even at the level of the crew unit (another type of information processing machine). Communications are subject to environmental intrusions, both atmospheric or manufactured, that interfere with the signals and block their paths, introducing echoes, noise, and jamming. The actuator could have failed completely and become unable to respond to any command. Or, if its output hub had indeed become loose, the attached rudder could have responded only partially, or not at all, to the instruction sent to it, no matter how correct the command. One contingent fault may lead to another, cascading upward through the levels of the system to affect its overall performance. Small-level inaccuracies may have the larger-scale effects of destabilizing the overall flight path. The small-scale fault can lead to the large-scale failure -- the malfunction of a system component or function to its complete breakdown.

Actors relate as discrete entities, yet they also modulate and constrain flows at various thresholds of experience. The threshold might be that of scale, magnitude, or frequency. It is determined by affiliative complexity: the extent of bonding, synchronization, and agreement across populations of actors. It is determined by the nature and influence of an affiliation's organizing principles: the programs through which actors gather together to stabilize in cohesive wholes that are more than the sum of their parts, and through which they can replicate and standardize. A failure is a fault that, having accumulated a higher level of affiliative complexity, has crossed a critical threshold. As there are no hard-and-fast boundaries between actors, or between affiliations, there are no hard-and-fast boundaries between fault and failure. But there is a transition point. It is a matter of developing a control system equipped with a sufficient degree of robustness to fault. It is a matter of the efficient, enduring management or maintenance of sufficient stability against instability. The overall "health management" of the UAV -- one dimension of its control system -- is designed to absorb faults. If the control system is not equipped with some form of fault tolerance, or if the fault-tolerant control system is not capable of providing sufficient recovery to the fault, the component or function may lose stability and exhibit an unpredictable pattern. Loss of stability at one scale can lead to a loss of continuity and cohesiveness at another -- perhaps to catastrophic ends.

It is also a matter of catching faults in advance through routine system checks. However, the checking itself could be faulty, the blame located in the quality and frequency of the particular system test in question, whether it might involve position accuracy, torque, speed, stiffness, or frequency and step response. The ability of a test to accurately detect fault is due to the quality and enduring relevance of its assumptions and procedures. It is due to its effectivity within changing conditions, including anomalous behaviors and abrupt environmental shifts. It is due to the agencies of its application. A test, like a text, is nothing outside of the modality of its usage, its constitutive agential positionings and enacted routines. Like any actor, it is a matter of its functional role in the system. It manifests by way of its action and maintenance: through the ways it comes to perform, at various scales, magnitudes, speeds, directions, and degrees of complexity. The action of the test is neither internally nor externally decided: it courses through its attendant actors, as these actors perform within the dynamics of the situation. It is a question of how the situation matters -- its shared priorities that come into play, as they are sustained in practices: recurrent composites of action, stance, and form.

The drone's components and systems take shape in degrees of coalescence and disruption, at various frequencies, rhythms, magnitudes, and scales of endeavor. They are subject to external forces, to the environmental stress placed upon them. How much can a part take before it fails, decouples from its job, spins out of synch? Forces of temperature, mass, and vibration conspire against it. The pressure is also discursive. The plane flies as an affiliation of maintained states, and adequate performance is a matter of maintaining sufficient stability at numerous scales of practice, whether these might involve software, hardware, or institutional and public dialogue. The drone works as a platform because the agents that it helps assemble, however organic or inorganic, material or discursive, "agree" that it works.

It is a matter of how these "agreements" come to exist, how they bond and calibrate, how they endure over time. In order for actors to move and endure, they must find their way from one moment to the next by drawing on available resources. They must modulate flows and facilitate correspondences, that are affective, rhythmic, or linguistic in nature, as well as cultivate their own availability for the modulations, calibrations, and correspondences of external agents -- increasing or diminishing their potentials to materialize and endure. They must cultivate their modulation in affiliations that can maintain them, offering consistency and coherence, and the gaining of influence, relevance, and intimacy. In so doing, they must negotiate adherence to the demands for movement and attendance that these affiliates maintain. It is a matter of the setting of the terms -- the common organizing principles, or programs, through which sustained affiliation is achieved.


On a brisk winter evening in Southern Arizona, a Shadow UAV plowed into the earth near Fort Huachuca, about fifteen miles north of the U.S./Mexico border. A volatile mix of colliding metal, electronics, engine fuel, and dry air, the drone violently destabilized, its energies cascading outward and out of control. The violent expansion sparked a flame. Gathering momentum, it began to ravage the countryside. Awash in combustible greenery, the fire burned across two acres of land. As with any incident that ignites a large blaze in the arid, forested landscape of this part of the country, where wildfires can easily burn out of control, the catastrophe was met with no small degree of alarm. However in this particular military base, home to the largest UAV training center in the world, catastrophes of this scale are generally no big deal. Drones slam into the earth here on a regular basis. Their resulting explosions and fires, blazing fiercely across the terrain, are simulated with ease. A cavalcade of drone wrecks can be called up at will, their burning remains smoldering in storage.

The burgeoning global UAS market relies heavily on companies and organizations that provide service, logistics, and training for the unmanned vehicles that proliferate across the skies. Facilities like those at Fort Huachuca must accelerate their output of skilled operators to meet the growing demand for drones. Since access to national airspace is largely restricted, much of the training is done on simulations. The interfaces of these simulations are familiar to any aficionado of games, roleplaying environments, and high-tech adventure films. Like the control panels of actual flight crews, sitting in their Nevada desert trailers, they bear the traces of the commercial game formats from which they are derived. One can detect the influence of Xbox controls, used by the Army, and the engines of games like Halo, upon which Raytheon's UAV control system is based. Even simulations like the Marine Corps's Virtual Battlespace 2 are based on commercial game engines that are boldly reflected in their titles.

Like the actual drones of which they are a component, the coherency and discreteness of these interfaces and applications dissolve upon scrutiny, scattering into arrays of component actors that are shared by other affiliations. These component actors -- visual and rhythmic motifs, behavioral conventions, perspectival formats, software codes, tags, controllers, users, corporate procedures, game architectures, rules -- circulate and bond across multiple domains of experience, traversing the divides between corporation and government, combat and entertainment, simulation and reality. The particular applications in which they accumulate, largely developed by the commercial game industry and influenced by commercial formats of cognitive and affective engagement, are made to excite the gamer, with characters that run faster and jump higher than is humanly possible, and explosions and flames that burn more fiercely than normal -- much brighter and more intensely than that which actually occurred in the wake of the crashed drone outside the Fort Huachuca facility, which surely would have disappointed the seasoned player. The problem is not that people, environments, and behaviors are un-lifelike, but that they are more than lifelike and must be downscaled, along with the expectations of their human accomplices, to calibrate with the velocities, magnitudes, and textures of the real world.

The component actors of these gaming, control, and simulation affiliations relate as discrete entities, yet they also modulate and constrain flows at various scales of experience. They are relatively stabilized, consolidated platforms but also dynamical systems defined by the atmospheric, technological, and institutional systems that they move through. As they configure and fluctuate in their relations and modulations, they require continuous adjustments across the various scales, rhythms, and magnitudes at which they are active. The agency of this adjustment traverses the bounds of the interface, neither fully internal nor external to any ontological divide. In the world of the simulation game, action emerges in a shared landscape of participation: one wants to get into the flow, carried forth by the activity, responding to situations in direct ways that bypass the mechanisms of thought. Differentials, commonalities, and alignments are negotiated, discrete changes or shifts in scale responded to, but in ways that do not involve hard and fast separations between user and action: one endeavors to get fully absorbed, to cease being aware of oneself as separate from the actions one is performing. The activity into which an actor is "swept up" is initiated neither fully internally nor externally, but courses through all of the actors in attendance, as these actors perform within the dynamics of the situation. It emerges in a shared field of endeavor. It is a matter of being attuned to the situation property, so as to be alerted to its priorities: the movement, stance, and positionality that it summons as most opportune.

Agency manifests by way of its action and maintenance: through the ways it comes to perform, at various scales, magnitudes, speeds, and degrees of complexity, and the extent to which this performance is recognized, valued, and maintained. An actor endeavors to be an adequate player of the game. It is a matter of what is deemed adequate performance: the shared priorities that come into play, as they are sustained in recurrent composites of practice. It is a matter of maintaining sufficient stability at numerous scales of practice, to the extent that shared formats, agreements, and standards can come to exist -- programs that can be allied with, offering propagation and endurance over time. The adjustments and calibrations required are entraining: actors are acclimated to one another within the terms of these specific formats, programs, and practices. Agential form is a matter of adjustment and compromise, across the various frequencies, rhythms, magnitudes, and scales of experience in which affiliation occurs.

Simulations often require nothing more than a joystick and personal computer -- a laptop can run all of the vehicle dynamics, including the sensors. Rendered portable, the same high-end environments that are found in stationary systems can be brought home for practice or taken directly into the field. Further narrowing the gap between rehearsal and mission, some simulations are plugged directly into the ground control stations that are used to manipulate real UAVs, allowing for training and operation to be done together, with operators toggling between simulation and actuality within a functional crew station.

The integration between gaming, simulation, and mission happens not only at the scale of the crew station but at the level of the command structure. Simulations like Virtual Battlespace 2 allow data that is gathered by UAV sensors within the gaming environment to be fed directly to command and control systems for a commander's strategic planning. It is said to provide a more comprehensive view of the battlefield, with real participants vying with simulated ones for evaluation, engagement, and participative hands-on training, in networks that amplify access to knowledge, situational awareness, and collaborative endeavor. It also provides analysts with simulated back-end processing of the raw data collected by the sensors. Ground base operations, inter-service and multi-national training events, and game based training situations together become essential precursors to deployment, increasingly integrated into command and control systems and actual operations in realtime.

Ground control stations, training simulations, and video games occupy a common economic, affective, and cognitive terrain: sites of data rendered actionable. Together they constitute an interlocking, visual, rhythmic, and orienting complex, harnessing the imaginary, that conditions orientation in the world. Like the material realities and infrastructures of the bases and training facilities within which it unfolds, however virtually, geographically, or institutionally embedded, the enacted routines of this complex play a large material and materializing role. Their transmitted signals, electrical and vibratory, are modulated and rendered discrete as coded meaning, in concert with the software and personnel that channel them. The relations and modulated transmissions are mutually generative: they configure agency, and are configured by agency, through their limitations and correspondences in the enacted routines of practice. They course through their attendant actors, as these actors perform within the dynamics of the various situations that arise, in various degrees of attunement to the shared priorities that they may reveal: priorities acted upon and inhabited, in various degrees of frequency, scale, and magnitude, in stance and position.

The situation matters in stances and positions. It matters in the evaluative alignments and postures that are taken in communicative encounters, however expressive, referential, or material, as they traffic between routined physical activity and larger social structures. It matters in the dynamic agential roles that are instantiated in communicative encounters, which help to give directional form to experience. The mattering might accumulate in values, tastes, desires and dispositions. It might involve body alignment, pace, expression, stylistic action, voice, shift of footing, poise, manner, affinity, or mood. In order to matter, it must be sustained in practices -- recurrent composites of action, stance, and form. Actors are integrated with, and instantiated through, these composites, registered and enacted at the level of one or more platforms of affiliation, primed in various patterns and integrated into coordinated response systems. They emerge from the dictates of a developing structure, but are neither fully internally or externally decided. Flows are absorbed and released, constraints enacted, forms negotiated, correspondences staged. Data from the environment is absorbed, contexts created for that data, internal and external communication facilitated, inferences drawn, and appropriate physical response output. Action is catalyzed, guided both from within and without the staged confines of the actor.

As the material agency of trained crew members coalesces an affiliation of maintained states, in alliance with a multiplicity of actors -- human, mechanical, informational, environmental, institutional -- so, too, does the material agency of the drone that is flown. Through practices, they are maintained in continuities, cooperating and competing for endurance, in whatever degree of simulation or actuality. They do not always conform or affect one another in linearly causal terms. The affiliations of the pilot and plane are connected, in resemblance and limitation, to a degree and scale that they can be stabilized and sustained: they exist in the world, inform and influence one another, with some degree of reliability, relevance, and intimacy.

Some routine practices stabilize into organizing principles, or programs. While programs are dynamic sites of social negotiation and organization, some achieve a higher degree of endurance, influence, and intimacy. Some work more efficiently or better than others, some have more allies, some more relevance. They perpetuate their standards such that other actors come to move in accordance with their terms. Actors necessarily adjust themselves in accordance with the programs of the gatherings with which they affiliate, acquiescing to their terms of negotiation. They entrain, and are entrained, in accordance with the programs through which sustained affiliation is achieved. Things fall into place.

Across these dynamic, entraining affiliations, functional organizations of knowledge and skill -- capacities enacted and roles played in the organization of the system -- are redistributed and re-constrained, along with positions, categories, and divisions of labor. As agencies circulate and bond across multiple domains of experience, traversing the divides between corporation and government, combat and entertainment, research and commerce, affiliations composed of unlikely bedfellows are brought together through economic need. If the "unmanning" of systems moves soldiers off the battlefields, it brings technology companies directly into them, in search of groundlevel feedback for updating existing products and developing new ones, in an increasingly competitive global industry. The redistribution of manpower in the "unmanning" -- the shift from soliders in battlefields and fighter planes to those in double-wide desert trailers and high-tech command centers -- challenges the stances, positions, and qualifications that have defined previous generations. The values, tastes, and dispositions of unmanned warfare do not always align with the gendered roles, imaginaries, and concepts of adequacy that were present in the noble, heroic ideals of the past. The Air Force now trains more drone pilots than fighter and bomber pilots combined. The "top gun" archetype is on the wane. Yet, as past ideals of heroic masculinity are threatened, new ones are created, embroiled in new forms of agility, knowledge, and prowess display. They might be manifest in the subtle alignments of the body, its pacing, expression, stylistic action, inclination, or mood. Warrior archetypes migrate into alternative geometries of privilege, however gendered -- myths of male identity wrestled with in the reinventing, rather than resuscitating, of a fading ideal.

With these redistributions comes a retooling of notions of skill and expertise. As intelligence migrates into unlikely, shared sources, even those spatial and atmospheric, and agency is understood to be distributed and embodied in all manner of organic and inorganic actors, a concept of skill emerges whose source is in negotiation rather than domination: an alliance with material actors rather than an assertion of command over them. Here an actor works with a material rather than against it, cultivating an existing, emergent meaning rather than externally imposing one -- a "knowingness" that is not simply categorical but affective and rhythmic. It transforms objects into situations, their contours not determined in advance but arising within the terms of the encounter. When an object becomes a situation, one cannot assert one's authority over it so easily. The benefit comes not from taming so much as listening: it becomes an occurrence to be learned from, patiently. This is about proximity rather than power, an "excessive closeness" to that which cannot be contained or possessed, and over which the impulse to control diminishes. It requires a break of routine, a more flexible notion of practice that can accommodate that which is revealed in the negotiation, often unexpectedly and outside of preoccupation. An affiliation's enduring relevance is not just due to competition but cooperation, often in subtle, sensory ways: it is not just influence, but intimacy. In the face of these concepts, analytical notions of power and desire diminish in their relevance.

The situational event of the crash reverberates across a dynamic agential field. It is a matter of being attuned, to a sufficient degree, to what matters there. Occasioned by the dynamics of the crash, drone ontology bleeds into epistemology.

Amplifying Expertise

It is a crisp October day in Afghanistan, in the midst of Operation Enduring Freedom. A U.S. Predator has just taxied and departed from Kandahar Air Field for a routine reconnaissance mission. The plane is assigned to an Expeditionary Wing at Creech Air Force Base in Nevada and operated by crew members at March Air Reserve Base in California. Suddenly, during the flight, the crew receives a direct task order from the Combined Forces Air Component Commander: they are to provide immediate air support to U.S. and Afghan ground forces that are under siege. The enemy fighters, numbering about 300, appear to be carrying out a large, coordinated attack. Given the intensity of the battle on the ground below, the circumstances of the attack, and the immediate and critical need for support -- U.S. soldiers were being killed -- the Predator crew is consumed with a high-degree of urgency. Their attention fully focused the battle, their awareness of the bigger picture diminishes. The pilot's distraction leads to a fatal mistake: he fails to see that the UAV is headed toward a looming, 17,000-foot mountain. The drone smashes into it, abstracted into a cloud of black smoke, its parts scattering into the desolate terrain below.

Human attention can be too tightly focused along one zone of experience, to the exclusion of wider expanse of contextual information -- what the military calls "situational awareness." Or it can be too scattered: not focused enough on anything. In order for it to be effective, a dynamic between stabilization and destabilization must be actively maintained. Yet however vigilant it might be, human attention is faulty and undependable -- ill-equipped to keep up with the demands placed upon it. As drones gain the ability to "dwell and stare" -- recording activities on the ground over much longer timeframes -- the vast amounts of data they absorb can easily outrun the capacities of personnel. (On a single day the Air Force must process nearly 1,500 hours of full motion video and another 1,500 still images.) Cameras and sensors become ever more sophisticated, yet they are of limited value unless they can be accompanied by improved human intelligence and skill. The task of interpreting what the UAV is seeing falls partly into the hands of the flight crew, who always has access to the aircraft's live video feed; they may also be joined by an image expert trained precisely for this purpose. The video is also sent to image analysts at other bases -- analysis and dissemination sites like the Joint Base Langley-Eustis in Virginia, inside of whose cavernous rooms image analysts sit, filtering vast streams of data arrayed on constellations of monitors. They, too, are hard-pressed. Staring for hours on end, nearly inert at their chairs, they try to ferret out the singular target -- the single, telling deviance in the normalized flow. Armed with the skill of extracting relevant data from image flows and information arrays, they attempt to organize that data into patterns of affiliation from which further extrapolations can be made.

The UAS, as an affiliation of components and systems, relies on analysis and dissemination sites like these. They are vital platforms of the drone in its shared perceptual and analytical capacities, its sensing, processing, communicating, and actuating functions -- nodes through which its data is streamed, formatted, tagged, and rendered searchable across networks of datasets. The platforms and nodes of these affiliations are many, from personnel bases in Nevada to storage facilities in Iowa -- repurposed shipping containers within which arrays of servers, tasked solely to house the video data generated by Air Force drones, quietly hum, as unremarkable as the rooms and trailers in which their operating crews sit. As the image data is organized and stored, it becomes the primary site through which correlations can be made and inferences drawn. Databases, activated through search algorithms, become the primary repository of knowledge. A backlog of replayable events is generated, seen from above: a searchable, historical record of a region's activity, as viewed from the UAV.

This movement that is detected, however geographically understood, is not necessarily causal and continuous: a history inscribed upon ground or air. It is rather a result of calculations calibrated across datasets: correlations among relatively stabilized and standardized elements that do not move across space so much as flicker or fluctuate within it. Movement is less a continuous transfer -- over ground, land, or spatial volume -- than a configurative interpolation. It is a trajectory assembled in retrospect, piecemeal as a correspondence of points: a behavioral composite into which movement and intention are inferred.

A typical drone requires 19 analysts. A single drone outfitted with "Gorgon Stare" technology -- which can capture live video of an entire city -- requires 2,000 analysts. This advanced video capture system, paralleling the teeming array of snakes emerging from the head of the mythical creature referenced in its title, has a spherical array of nine cameras -- five electro-optical and four infrared -- emerging from the underbelly of its platform. Its software compiles the various camera views into a broad, continuous mosaic, of which personnel on the ground or elsewhere can simultaneously grab slices -- its analytical requirements farmed-out, in real time, to a network of analysts and computing platforms.

Like the winged fusion of human, beast, and machine that is its namesake, with claws of steel, bulbous head, and large unblinking eyes, the gorgon drone is equipped with a deadly, commanding stare: it looks at you, but you cannot look back, lest you be turned to stone. Although menacing in its demeanor -- fangs bared and nostrils flared as it readies to inhale and consume -- it is understood to have protective qualities for its deployers. Heir to the symbolic apparatuses of myth, its figures are everywhere present on objects, documents, ideologies, and units of value exchange, manifesting hopes of warding off evil. Its potential for domestic protection is not lost on the U.S. Departments of Defense and Homeland Security, who are exploring its panoptic potential for border security: Gorgons now joining the cavalcades of machine beasts flying high over the desert borderland, with no illegal activity going unchecked, be it immigration, drug trafficking, or the very flows of terror.

The technology is not without its substantial problems. While it may "see" over a wider swath of territory, it does not necessarily understand the significance of what moves within it. The challenge remains that of tracking vehicles, objects, and humans on the ground with a higher degree of precision, in ways that lessen the demands on human personnel. The challenge with UAS in general is to amplify the overall intelligence of the system -- heightening the level of skill and expertise that the affiliation can engender. This often takes the form of enhancing the capacity of tracking and search algorithms, along with the network processing capability required to parse and coordinate the data. It involves increasing the ability of UAVs to sense, reason, learn, and make decisions, and to collaborate and communicate, with a minimal degree of direct human involvement.

In popular terminology, this is called automation or autonomy. When approached within the terms of the "salvage operation," however, where agency is situated in shared composites of intelligence and skill -- affiliations among all manner of actors operating at various scales, magnitudes, and degrees of complexity, whether at the level of hardware, software, flight crews, or institutions -- these discourses of autonomy are resisted. The unmanned system does not eliminate the human so much as redistribute the agencies of warfare. The capacities of sensing, dispatching, analyzing, and alerting -- the intelligence and skill required to interpret and store information and act on the results -- are shared by an affiliation of actors, however algorithmic, organic, or systemic. The focus is on their performative practices within the functional organization of the system. It is a matter of how they are maintained as dynamically stable entities -- sustained, naturalized, and rendered discrete -- and the programs through which this is accomplished.


Image analytics software is used for the recognition of objects, vehicles, and people. Even the most rudimentary drones, such as the Orbiter, have some form of algorithmic tracking, even if it is only basic motion detection. The software takes input from cameras, then recognizes and identifies the objects in each frame to learn what activity normally takes place within the area under observation. With the norm firmly in place, the software then aims to detect activity that deviates from it: the exceptional occurrence that stands out from the domain of the ordinary. Algorithms screen out non-critical movement activity and foreground the critical, in order to maximize the attention spans of observing experts. Once alerted to aberrant movement activity, its nature and intent is to be inferred by these experts, who then decide what action to take.

The norm is based on the practiced rhythms and regularities embedded in everyday space -- the patterns inscribed in the timespaces and infrastructures of observed populations through travel routines, social habits, building configurations, and communication forms, as these are aligned with the rhythms and regularities of the observing institutions of the UAS, cohering through flight patterns, transport timetables, interchanges, regulations, monitoring systems, base locales, and operating habits. It is a calibration among systems and data derived from both the object of observation and the observing institution itself -- the stabilization of dynamic fields that have been limited and brought into correspondence with one other. The UAS learns the everyday norm as it co-constitutes it. Embedded regularities coalesce in the "air" and on the "ground": embodied practices respond to the regularities that they help sculpt, in ways that can further stabilize or destabilize their defining platforms and programs.

It is always a question of the priorities that come into play: the patterns and flows that are deemed most appropriate to the circumstances, as they are stabilized and maintained in practices. Some aspects of practices, prioritized as such, stabilize into higher-order principles, and once sufficiently stabilized, they replicate, become redundant and standardize, at various scales, across various platforms of endeavor. They endure as programs -- dynamic sites of social negotiation and organization. The terms of these programs, potentially, are equal, though they are always hierarchized, however temporarily: some are more enduring, influential, and intimate than others, some more effective and relevant, in part because they have more valued affiliations. Their standards, perpetuated, set forth the terms of social negotiation, and in so doing, they compel external actors to move in accordance with these terms. Movement is affiliation, and affiliation happens in program.

The norm is constituted through the categorization and standardization of information. A logical grid is summoned on which categories are made. The phenomena that appear in algorithms and databases must pass through this logical grid: they must pass through the standardized forms of information that the system, as a whole, admits. A particular scaffold gains influence over others, and in so doing, sets forth the terms of engagement and interpretation. Things fall into place. Objects are classified according to pre-programmed definitions and specifications, and rules are established that are tailored to these objects within the observed scene. Moving phenomena are stabilized, constrained, and defined in accordance with these database associations and the programs through which they are ordered. An object is disengaged from its maintaining network and a form, atmosphere, space, or force is normalized, no longer understood in terms of these operations.

The normalization of activity is what makes possible the detection of abnormal movement -- an event. An actor is taken up within the arena of attention as an exception because that which surrounds it has been standardized, regularized -- transformed into atmosphere.

Advanced-stage UAVs now incorporate cognitive architectures and machine learning capabilities that allow them to recognize and identify objects with a more complex and integrated capacity of expertise. The parameters for the algorithms to recognize behavior or objects need not be set in advance. A learning engine gathers information about dominant object content -- tracking, for each object, features like size, color, reflectivity, sheen, shape, and level of autonomy -- and forges object classifications without any pre-programmed definitions or specifications. The software analyzes the scene to learn and identify normal and anomalous behaviors by way of a constant study of the types of objects that exhibit those behaviors. It learns from experience, internally adapting to changes in the observed environment, detecting and classifying activity that was not previously defined or anticipated.

Drones like Teranis, developed by the British firm BAE -- among the world's largest military contractors -- combine these cognitive analytics with vehicle control systems. Integrated flight control systems are already in use with the Global Hawk, which, after its launch, carries out a pre-programmed mission by downloading GPS coordinates via satellite. However BAE has developed a comprehensive mission-handling system that combines image analytics with flight control. Its MANTIS and HERTI drones are said not only to fly themselves, but to conduct target searches on their own -- reducing the risk of human distraction and error as well as communications and data link requirements between the vehicle and the ground.

Lockheed Martin's Polecat drone was also said to be a fully autonomous system. Its prototype, however, crashed upon its unveiling. It plunged into a Nevada test range after a failure in the ground equipment caused its automatic, fail-safe flight termination mode to activate. The fail-safe mode is intended to minimize danger to civilians should the plane deviate irrecoverably outside its range boundaries. It is designed to prevent human operators from recovering control of the UAV. Perhaps the true meaning of an autonomous system can only be achieved in such a state of complete and utter relinquishment. Rendered powerless, the operators could only watch as the drone plummeted to earth.

Lockheed Martin did not divulge whether the source of the crash was human error or technological malfunction. The company did however attest to the reliability of the flight termination software, which according to them, performed exactly as expected.

The tailless, 90-foot wing-long Polecat looks like something straight out of science fiction. Had it not crashed -- as did its predecessor, the ominously named Dark Star -- it could have well been the flagship of advanced drone lust. Its material constitution was exemplary. It was built of composite materials rather than metal. It was comprised of less than 200 parts, many of which were built in rapid prototyping. The digital models of these component parts were produced with a computer-aided design system and output to a 3D printer, inside of which two powerful laser beams, steered by a computer, were finely focussed at a composite powder -- sintering it, layer by layer, to form complex, solid volumes. Until now, this technique was only used in the industry to make test parts, to assess their adequacy for the final job; the strength of parts produced in this way has improved to such an extent that this intermediate stage is not necessary. A wing strut, formerly only a prototype, becomes the real thing.

These 200 components were not riveted together but glued with adhesive. When the Polecat slammed to the ground, its identity abstracted in a burst of composite materials, electronics, sensors, and smoke, the original contours of these components would not have been visible. They were, in any case, only provisionally-stabilized platforms in a larger process -- one whose origin might just as well be located in composite powders or programs. Like all parts they coalesce as affiliations at various scales, magnitudes, and spans of endurance, their geometries abstracted and opened in the advent of their collision with other affiliations of more solidity, permanence, and force. Offered up anew in a spray of oil and dirt, burnt rubber and wire, recovery logistics and salvage discourse, their negotiations and consolidations continue.


Author's Note: All of the UAV crashes mentioned in this essay are based on actual crashes, although I have taken some degree of poetic license in their description. The details and dates of each are discoverable through a simple web search. I have chosen not to footnote them. I have chosen to dispense with footnotes entirely. The scholarly reader will see the intellectual debts of many of my concepts, which are drawn from vitalist philosophy, object-oriented ontology, and feminist science studies, along with theories of affect informed by the life sciences. They owe much to assemblage theory, actor-network theory, and agential realism, among other work that develops alternatives to modernist frameworks of subjectivity. I have avoided footnoting this material for several reasons. I want to combine traditional forms of scholarly research with approaches drawn from the worlds of visual art, literature, and performance -- inspired, for example, by writers who are experimenting with critical fiction from an anthropological perspective. I want to energize and amplify the performative dimensions of the worlds of human and nonhuman actors of which I speak. Such a course of study, rather than imposing argumentative convention, requires attunement to the unexpected priorities of the encounter and the novel forms of articulation that emerge. It challenges one to forge new concepts that may take on a life of their own, like good fictional characters, and at times exceed the boundaries of traditional forms.


Theory Beyond the Codes: tbc027
Arthur and Marilouise Kroker, Editors
Date Published: 11/2/2011