Highlighted Selections from:

The Cultural Labor of Surveillance: Video Forensics, Computational Objectivity, and the Production of Visual Evidence


DOI: 10.1080/10350330.2013.777593

Gates, Kelly. “The Cultural Labor of Surveillance: Video Forensics, Computational Objectivity, and the Production of Visual Evidence.” Social Semiotics 23.2 (2013): 242–260. Print.

p.243: This essay argues that the status of video evidence as an index of real events--a sign or representation that offers a direct, empirical connection to material reality--is the result of an intentional process of production. This process involves the repurposing of new technologies borrowed from the domain of creative media production in order to transform a chaotic field of raw surveillance video into useable evidence. In addition to the exchange in technologies, an unavoidable epistemological and interpretive exchange takes place between evidentiary uses of surveillance video on the one hand, and the now prevalent forms of surveillant narration found in both fictional and reality-based storytelling. But despite this exchange in meanings and technical systems, considerable effort has gone in to establish formal standards for the evidentiary uses of surveillance video that distinguish the discovery of video evidence from the production of creative content. Building on Daston and Gallison’s historical study of the prevailing "epistemic virtues" that have defined objectivity over time, I argue that what we see emerging in the field of forensic video analysis, as a means of establishing its scientific and legal status, is a commitment to a new epistemic virtue of "computational objectivity." -- Highlighted may 1, 2014

p.243: A photo of two of the alleged 9/11 hijackers passing through airport security in Portland, Maine, on the morning of the attacks appeared in 1 October 2001 issue of TIME magazine.Another photograph, allegedly depicting three of the London subway bombers on a trial run before they carried out the attacks on 7 July 2005, was published in The New York Times two and a half months after the bombings. -- Highlighted may 1, 2014

p.244: Images like these are readily recognizable as still shots captured from surveillance video -- the perspective, usually shot slightly from above and at an angle, a banal location, a grainy or blurry quality, and often a time¡date stamp and a camera number imprinted on the image. When reproduced as press photos, the images typically include an image credit printed in tiny type below the photo. In the London subway photo, it read "Metropolitan Police, via Associated Press," (Lyall, 2005) suggesting both a path of information flow and an institutional relationship. This and other contextualizing metadata (the caption, the headline, and the text of article) assign specific meaning to these images, but the meanings of their visual content alone are often radically indeterminate. Nothing in these photographs themselves can definitively identify the people depicted, for example. Surveillance images can be of fairly poor quality, and it can be difficult to see people’s faces. If facial recognition technology had been installed with the surveillance systems that captured these images, it is unlikely that it would have matched the faces with accurate identities since the technology does not work well in uncontrolled settings like the ones shown. -- Highlighted may 1, 2014

p.244: I want to call attention to the amount of work that goes into establishing the credibility of these images and others like them -- the discourses, procedures, technologies, and forms of labor that work to define the images as authoritative proof. -- Highlighted may 1, 2014

p.244: Images like these have a grainy, seemingly unproduced claim-to-authenticity -- an "aesthetics of objectivity." These images are metonyms of video surveillance, invoking the realtime visual mediation of real events. When they depict something of interest, they not only represent that event but also the fact that it was caught, "live," on camera. But despite their seemingly "raw" quality, considerable work goes into producing these images, into choosing "the perfect moment" as well as clarifying the image and making relevant details visible. The surveillance images that circulate as press photos are not untouched real-time images, but they gain authority and credibility by invoking a sense of real-time surveillance. -- Highlighted may 1, 2014

p.244: A still from a surveillance video, circulated in the press, not only makes an evidentiary statement but it also allows a symbolic form of passive public participation in the investigation. Readers are invited to take the perspective of surveillance workers, observing the real-time surveillance feeds, to literally see the scene through the surveillant police gaze. -- Highlighted may 1, 2014

p.245: In his history of police photography, John Tagg (1988) notes that crime reporting was part of the manufacture of the criminal threat that provided an alibi for the professionalization of the police forces in the UK in the nineteenth century. Today, with a professional law enforcement apparatus fully institutionalized in all developed countries, crime reporting continues to serve a legitimizing function for the police. But rather than establishing the professional police force as a social necessity, which is no longer considered a reasonable question, crime reporting (and crime drama) works to legitimize policing strategies and adoption of new technologies. Each time a surveillance camera image circulates in the press, it does double duty, representing the event in question while at the same time justifying police deployments of video surveillance systems. -- Highlighted may 1, 2014

p.245: In this paper, I consider the emerging field of video forensics as one technology of CCTV optimization. -- Highlighted may 1, 2014

p.245: In other words, surveillance video in itself has little or no evidentiary value; video evidence must be produced from a chaotic field of raw surveillance footage. The status of video evidence as an index of real events -- a sign or representation that offers a direct, empirical connection to material reality -- is the result of an intentional process of production, a process that requires new forms of technical expertise and police work, as well as the adaption of new technologies borrowed from the world of creative media production. -- Highlighted may 1, 2014

p.245-246: The emerging field of forensic video analysis is one site where an epistemic virtue of "computational objectivity" is taking shape: the belief that neutral scientific image analysis can be achieved by translating certain forms of professional trained judgment into computational processes or, in this case, through the application of computational techniques by police professionals retrained as video specialists. I also suggest that, regardless of how important the virtue of objectivity is to forensic video analysis and the production of visual evidence, there is no avoiding the exchange in technologies, practices, and meanings that takes place between the official domains of investigative practice on the one hand, and the less official domains of surveillant narration found in film, television, and new media. -- Highlighted may 1, 2014

p.246: In Christopher Wilson’s (2000) cultural history of policing, he examines the narrative authority that police held in the United States throughout the twentieth century -- how dominant ideas, values, and assumptions about crime and policing circulated back and forth from police policy and practice to crime reporting and the popular genres of crime fiction and True Crime storytelling. -- Highlighted may 1, 2014

p.246: As Wilson argues, police power is derived to a significant extent from their narrative authority or their ability to tell authoritative stories about crime and about their own proper role as agents of the state and as arbiters of law and order. -- Highlighted may 1, 2014

p.246: In a study of the institutional structure of police work conducted in the 1990s, Richard Ericson and Kevin Haggerty (1997) found that the police did not spend the bulk of their time "fighting crime," as portrayed in media crime dramas, but, instead, on administrative matters or "knowledge work." For this reason, they argued, the police are best understood as knowledge workers within risk management systems: law enforcement officers perform the labor of collecting and processing information by and large to support the risk management needs of other institutions, such as the insurance industry. Institutional demands for knowledge of risks have led to the rise of a risk-communication system of policing, which now largely determines what the bulk of police work actually involves. -- Highlighted may 1, 2014

p.246: Where Ericson and Haggerty argue that the police face external institutional forces and constraints that shape their work in ways over which they have little control, Christopher Wilson sees the police as somewhat more autonomous producers and promoters of "cop knowledge" or police-oriented epistemologies that hold considerable sway in narrating forms of social understanding. -- Highlighted may 1, 2014

p.247: When considering the types of "new media work" that the police perform today, what is clear is that the respective roles of the police as generators of data and producers of culture are not mutually exclusive. To understand the relationship between these two indistinct responsibilities of modern police, it helps to understand the ways in which at least some of the work that police do today can be characterized as cultural or "immaterial labor." -- Highlighted may 1, 2014

p.247: What is conceptually understood as immaterial labor pervades police work, especially insofar as "the skills involved ... are increasingly skills involving cybernetics and computer control" (Lazzarato 1996). -- Highlighted may 1, 2014

p.248: In other words, while it is important to consider the evolving forms of new media work that the police engage in today, it is also important to consider a parallel development: the use of surveillant forms of narration in a wide range of cultural forms, both factual and fictional. In fact, it is difficult to draw a clear distinction between the uses of surveillant narration in real versus fictional storytelling. On the one hand, surveillance video lends dramatic weight to court proceedings, news stories, and reality-based programs. On the other hand, a visual rhetoric of surveillance is now commonplace in fictional drama, typically used to invest the story with a high level of realism or to convey information about the temporal dimensions of the story, such as using a recorded surveillance video as a form of flashback or to show two scenes happening simultaneously. The symbiotic relationship between the uses of a surveillant mode of visual address to represent both reality and realist fiction cannot help but encourage a similar sort of interpretive exchange at the level of reception among audiences. As Nick Groombridge (2002) puts it, the "separation between the rational/bureaucratic elements of CCTV and the affective/ aesthetic/entertainment [aspects] can no longer be sustained" (37). -- Highlighted may 1, 2014

p.249: Cinema’s use of surveillant narration as a form of realism has its inverse in the techniques that forensic video analysts are developing to process and analyze surveillance video. Video forensic techniques enable their users to make authoritative claims about the indexicality of surveillance images, helping video analysts transform a chaotic and meaningless field of recorded vision into coherent and direct references to real people, places, objects, and events. But much like with cinema, a significant amount of production labor goes into transforming surveillance video into evidence and ensuring that images maintain a seemingly "unproduced" quality. Concerns about lighting conditions, placement of cameras, camera angles and movement are not alien to users of video surveillance systems. These users are also becoming acquainted with editing and postproduction techniques, adapting them to the needs of visual evidence production. -- Highlighted may 1, 2014

p.250: Fictional depictions of video forensics may or may not create a "CSI effect," leading to unreasonable expectations about what can be accomplished with digital imaging techniques and visual evidence. Regardless, these dramatic portrayals should not be dismissed as simply misleading, nor should concern focus narrowly on their impact on juries. Instead, cinematic and televisual depictions of forensic video analysis point to the epistemic and interpretive exchanges that take place between police narrative strategies and the broader domain of cultural production about crime, social disorder, and police power. -- Highlighted may 1, 2014

p.251: The photo of two hijackers passing through airport security on the morning of 9/11 mentioned earlier, for example, was produced with frame averaging, using data from seven different video frames to bring out more detail. -- Highlighted may 1, 2014

p.251: Similar imaging techniques were used by ABC News to show "a pair of gashes or welts" on George Zimmerman’s head the night he shot and killed teenager Trayvon Martin, marks that were initially not visible in a surveillance video taken at the police station as he arrived that night. -- Highlighted may 1, 2014

p.251: It must also be understood in terms of the extent to which forensic techniques facilitate the opposite aim of visual opacity. At the very least, forensic analysts must make particular choices about what visual information to focus on and what to ignore, informed by their own professional trained judgment. -- Highlighted may 1, 2014

p.252: When ABC News aired their enhanced video of George Zimmerman being brought into the police station, for example, they froze the video on an image of Zimmerman from behind, using a highlighted circle to enhance the marks on his head that were not visible in earlier footage. In this way, they directed viewers’ attention to the precise visual details deemed relevant to the story. The rest of the image was darkened so that other visual information would not distract viewers’ from seeing the newly visible evidence and making the correct determination (i.e., that Trayvon Martin had inflicted the injuries on Zimmerman, thereby supporting the police narrative of events and their decision not to charge him. -- Highlighted may 1, 2014

p.253: In descriptions of video forensic systems like dTective, one would be hard-pressed to find evidence of William Mitchell’s (2001) argument that new digital technologies are "relentlessly destabilizing the old photographic orthodoxy" and subverting traditional notions of photographic truth (223). Instead, in these promotional descriptions, we find an emphasis on the ways new digital techniques amplify the powers of surveillance and open up new dimensions of reality to observation. -- Highlighted may 1, 2014

p.253: Here I want to suggest that the use of digital imaging technologies to make visible what is invisible and invest images with indexicality points to a new conceptualization of objectivity. This way of thinking about objectivity holds that neutral, scientific results can be achieved through the application of computational forms of analysis -- automated, algorithmic techniques performed by computers. -- Highlighted may 1, 2014

p.253-254: In their historical study of the prevailing "epistemic virtues" that have defined objectivity over time, Daston and Galison (2007) use the term "mechanical objectivity" to refer to the type of objectivity associated with photography and other visualizing instruments developed in the nineteenth century. Proponents of "mechanical objectivity" subscribed to the belief that mechanical devices could be used to produce scientific images that were uncontaminated by interpretation, in contrast to the artistically rendered, "true-to-nature" illustrations that populated scientific atlases. Photography promised to remove the individual scientist’s judgment, and the biasing hand of the illustrator, from scientific image making. But photography and other mechanical visualization techniques never made good on this promise -- the problem of image interpretation persisted -- and what emerged in the twentieth century, as an acceptable avenue to objectivity, was the epistemic virtue of "trained judgment." -- Highlighted may 1, 2014

p.254: What we see emerging today is another kind of objectivity -- a "computational objectivity," or an avenue to objective analytical results that aims to translate certain aspects of trained judgment into computational systems. Along with the effort to achieve "computational objectivity," and to promote it as a new epistemic virtue, there is a renewal of the suspect promise that the biased and imperfect perceptual capacities of human beings can be eliminated or designed out of computational forms of image analysis. -- Highlighted may 1, 2014

p.254: Computational models of vision, for example, promise to automate aspects of what Charles Goodwin (1994) calls "professional vision," or "socially organized ways of seeing and understanding events that are answerable to the distinctive interests of a particular social group" (606). -- Highlighted may 1, 2014

p.254: The belief in computational objectivity is a powerful counterweight to the challenge that digital techniques pose for the principle of photographic truth so important to the evidentiary value of images, whether in science, journalism, the law, or other domains. But the belief that the biased and imperfect interpretive capacities of human beings can be designed out of computational forms of analysis depends on a black-boxed view of technical systems. This view obscures both the culturally specific models of human perception that get programed into automated systems, as well as the central (if changing) role of human labor in the design and operation of technical systems. In fact, the translation of professional trained judgment into computational forms in the field of forensic image analysis is only a partial project. The role of human perception and professional trained judgment not only persists, but is itself being pushed in new directions, with new types of expertise taking shape in complex relationship with evolving professional norms, legal standards, and technologies and cultural forms. -- Highlighted may 1, 2014

p.254: "Computational objectivity" can be said to encompass not only the application of algorithmic techniques to specific images, but also the computational organization of large collections of images in order to associate them with other data, identify patterns and relationships, and make images available for later use. -- Highlighted may 1, 2014

p.257: Timelines can be constructed to track a suspect’s movements in video gathered from multiple sources, or to show how an event or series of events may have played out, piecing together video with other evidence gathered in one or multiple investigations. The timelines and reports produced in this way provide authoritative documentation for courtroom presentation, lending a form of "documentary verification" to investigators’ narratives of events. -- Highlighted may 1, 2014

p.257: What kinds of narrative possibilities are introduced by the intersection of surveillance image databases and the storytelling requirements of investigators and prosecutors? It is not simply the ability to extract more information from images that is significant. Equally important is the capacity for organizing and reorganizing imagery, editing shots together from multiple cameras, editing together different moments in time and different locations -- in general, the ability to apply nonlinear editing techniques to the task of constructing coherent narratives from a chaotic field of surveillance footage. The very real creative possibilities opened up by new database-narrative assemblages means that the emerging field of video forensics must carefully define the limits of its creative potential. Efforts to distinguish forensic video analysis from the domain of creative media production can be found in the promotional language used to define video forensics systems, as we have seen. They can also be found in the new field’s commitment to the emerging epistemic virtue of "computational objectivity". -- Highlighted may 1, 2014

p.258: An analysis of the social and cultural implications of surveillance video’s evidentiary uses should not be limited to its role in particular investigations. More broadly, we need to make sense of the ways that the field of video forensics and other types of media expertise invest police institutions with renewed narrative authority in the new media landscape, giving the police added ammunition in their constant battle of interpretations over the prevailing ways of seeing crime, social disorder, and police power itself. -- Highlighted may 1, 2014