How Smart CCTV is made Smart: A Story about the Smartboxing of Image Processing Algorithms
We are proud to present a short blog post from Christoph Musik, whose new article, “It’s the Camera! A Special One: The Smartboxing of Image Processing Algorithms and their Uncertainties in Media Representations of Surveillance Technology” appears in the new open issue of Surveillance & Society.
///
Have you ever asked yourself what makes Smart CCTV, intelligent video surveillance systems, data or homes either ‘smart’ or ‘intelligent’? What features transform an ordinary object into a smart or intelligent one? Most of you will guess that these added terms refer to digital technologies. And those of you with a little more interest in technology studies will know that different forms of computer algorithms are responsible this transformation. In the case of ‘Smart CCTV’ Image Processing Algorithms (IPAs) are used to make standard CCTV to smart CCTV. These IPAs process, sort, and understand the sequences of images produced by cameras in an automated way, without human intervention.
Eight years ago, I started researching the basic but fundamental question of how computers or machines are taught to be able to see. I soon arrived at the preliminary conclusion that all these approaches can be understood as processes in which IPAs are developed, produced and implemented in devices or systems. To find out more, I visited computer vision research laboratories and found, as might be expected at research institutions, that many of their answers to my question were of a tentative nature. Inside the labs, one could further observe that IPAs were usually referred to in terms of their uncertainties, such as false-positive or false-negative cases. Consequently, the focus of their work was on trying to make IPAs ‘more robust’ and better working.
I also discovered — to my surprise — that most of the Austrian research projects on automated image processing were still at a prototype or field test stage. It was more by accident that I came across a nationwide operating system that contains, at its heart, IPAs designed to recognize patterns: the so-called ‘Automatic Toll Sticker Checks’ (“Automatische Vignettenkontrolle”, in short: AVK) on Austrian motorways. This system has been in operation since 2007 and, as I learned more about it, I realized that IPAs and their inevitable uncertainties were completely disregarded in public discussion about the system. The technology was perceived as something close to magic, a phenomenon I call ‘smartboxed’ in my research, in reference to Bruno Latour’s blackboxing.
By smartboxing, I refer to the process by which complex technological surveillance systems — such as AVK — including all scientific, technical, and operational practices, are made invisible, while at the same time being labelled and accepted as a ready-made smart or intelligent entities. In smartboxing, the focus is exclusively on the successful output which is duly accentuated, while never referring to what makes it smart or special and in which ways.
In my view, the concept of smartboxing and the article describing it are important contributions on at least three levels. On an academic level, it helps us analyze and understand the process that increases the gap between the perception of a technology by the public and the complex socio-technical and ‘constructed’ nature of it. On a sociopolitical, level my article will contribute to the demystification of smart technologies and algorithms and, thus, provide important insights that will guide their future implementation and use in more realistic and modest ways. Finally, on a personal or individual level, readers might feel empowered in their ways of encountering, using, opposing, or refusing smart technology.
///
You can read “It’s the Camera! A Special One: The Smartboxing of Image Processing Algorithms and their Uncertainties in Media Representations of Surveillance Technology” here.
