Background: Between January and August 2011, the Enumerative Institution was commissioned by the planning bureau of an unnamed municipality, hereafter referred to as The City, for the creation of a land use map. After several months and a major revision, the project was shelved by the client and described, off-the-record, as “unusable.” The final version of the map and reflections on the project notes are presented here, for better or worse, as a case study. 17 January We were commissioned to create a Land Use map. It was a straight-forward request from The City and a project we initially thought might not be interesting enough to pursue. In response, we proposed that such a map might be interchangeable with already existing zoning maps (these were readily accessible by The City). Though there would admittedly be some gaps between zoning and land use, perhaps these would be negligible enough depending on the intended purpose. We also suggested that the numerous planning studies executed by The City and its external consultants could be translated and spatialized to achieve the necessary maps. While we would happily provide this spatialization, we acknowledged that The City most likely had someone better suited for the task on staff. In our estimation, a new survey was excessive and our services were not required. 21 January On 20 January, The City responded with a clarification. Though our proposal for alternative approaches to the land use map were good suggestions, and indeed had been implemented in the past (politely, outdated), the planning bureau was operating with new ambitions and required a custom mapping strategy. For the previous nine months they had been working with two prominent engineering schools on a series of projects funded by a major telecommunications provider, and during this time The City had secured vast amounts of data regarding people’s locations and activities within the municipal boundaries. Using an exotic low-altitude sensor The City could now correlate an individual’s GPS data with aerial observations collected at a highly-periodic interval: every 90 minutes. With this new data, they were hoping to create a highly detailed land use map with a density of categories “not yet reasonably attempted.” They were also interested in defining the “constitutive infrastructure” for these activities. It was admittedly exciting to have access to the dataset they had described but from our position both of the stated deliverables were too vague; we were unsure on the number of desired categories and their composition, or what was meant by “constitutive infrastructure.” 8 February On 8 February we accepted the commission after several meetings with the planning bureau and their engineering consultants. Regarding our questions about the land use categories, The City offered the Standard Land Use Coding Manual (SLUC) as a precedent. Completed in 1965 by the Bureau of Public Roads and the Urban Renewal Administration, the SLUC contained over 700 land use categories limited primarily to economic activities. The City believed that this level of detail was consistent with the intended scale of the Land Use map and the amount of data they were able to collect with their new sensors. We agreed that a more detailed inspection of human activities implied an increased number of land use categories. On the other hand, we cautioned that the closer the map approached 1:1 scale, the more the land use study would resemble Ethnography— the proliferation of observed land use activities would undermine the purpose of the classification scheme. The SLUC provided a useful compromise. Despite its limitations we agreed to use the scheme for the project. The City was less prescriptive on the topic of “Constitutive Infrastructure” : in the course of our work, we were simply asked to analyze and represent the infrastructural features that shaped land use activities. 24 April We delivered the first land use maps on 21 April. Their overall development took nearly three months. We spent the first two weeks familiarizing ourselves with The City’s new datasets. Most of that time was spent waiting; the file sizes were very large and moving the data between machines and between software took more of our attention than we had anticipated. It was three weeks before we saw the first pass spatialized and only then could we start analyzing the patterns for indications of land use. Despite The City’s initial claims that previous zoning and land use maps were irrelevant to the present project, we found them very useful. When treated as base maps, their categories and boundaries often clarified the patterns we were observing in the data and aided in our decision-making. It was a simple process of layering previous maps with the new data to understand what it was showing us. We established a workflow and proceeded this way for several weeks. The resultant maps were rendered at a scale of 1:20,000 and contained 300 SLUC categories. They were delivered as a bounded volume on standardized 24 x 36” USGS topographic quadrangles. After a positive assessment from The City, we were asked to complete a ground-truth session on select neighborhoods to test the accuracy of our maps. Given the resolution of their data, they were certain we would confirm our categories on the ground quickly; the session would add that much more validity to the maps and to the new datasets. 29 April The results of our 27 April ground-truth session were completely unexpected; we had a difficult time understanding what we had observed. To begin, we chose what appeared in the data to be a busy intersection, with heavy, predictable foot traffic and commercial activity. It was a site that had been highly legible in the data and was classified without any debate. What we found was confusing— an area that should have been full of activity was populated only sporadically and without order. We waited 90 minutes for the next aerial sensor fly-over; regardless of the site’s puzzling state, our job was to correlate the readings from the sensor with ground observations. As the sensor approached, we were surprised to find numerous people also approaching the intersection and surrounding streets from all sides. Each of them walked along or cut across the site at the moment of the sensor’s measurement and diverged as quickly as they had come. This had a remarkable result. For the moment of aerial observation, the street had the density and activity we anticipated to find all along. We observed this phenomenon four more times that day and it occurred in just the same manner. With an apparent lack of intentionality on the part of the citizens, they all seemed to coincidentally arrive at the site at those moments of measurement while going about their daily business. We were not sure how to document these findings or what they meant for our maps. Our attempts to write them in the standardized ground-truth template made it all the more confusing. The procedure states that the observer should document when is seen on the ground at the moment of the sensor fly-over. Based on this requirement, what we observed at the moment of fly-over was as anticipated. Given this confusion, we submitted our forms but requested an extension to the project in order to explore these issues. We described them as “data correspondence errors.” 2 May The City requested clarification on our last report. We assured them that their data was not wrong in a strict sense. In fact, it was technically more precise than we had anticipated at the moment of sensing. The data, so far as data can, documented the state of things on the ground at the moment of the sensor fly-over. In our view, the issue came from the patterns of land use implied by the sequential sampling of the site. We stated that these patterns were at best misleadingly prosaic and that the actual behavior of the citizens was far more complex, improvisational, and difficult to classify. In the interest of developing a model—and by extension a land use map— that better correlated the data to the ground, we requested several more weeks of accuracy studies. These were granted without further inquiries. 27 May We spent three weeks mapping The City at 1:1 scale. As we anticipated, it was impossible to parse land use into standardized categories from this work. We originally assumed issues would arise from the varied and ever-changing personal activities that individuals engage in. The difficulties, however, came from the citizens’ strategic relationship to mapping and measurement itself. In the numerous interviews we conducted we found that an overwhelming number of citizens were aware of The City’s data collection program and its intended uses. Many were better informed than our team in the technical aspects of the new air-borne sensors and several even knew that our firm had been granted a contract to create land use maps. We found two former students of Nyqvist and many avid readers of Whittaker, especially his latently political work, Calculus of Observations, in which, among other claims, Whittaker calls for a mathematics of “reading between the lines.” Most importantly, we learned about the practical relationship to this knowledge that had developed within the community. They had designed path-planning and scheduling algorithms that optimized their routes through the city. Whatever their plans or activities, their paths were pre-computed to be located and measured in specific locations during a sensor fly-over. Once their paths were planned, they needn’t even think about it as they went about their daily routine. These tools had been in use for several months and had reorganized the space of the neighborhood in ways that excited the citizens. From the choices of the path-planning algorithm, they found themselves passing through blocks they hadn’t explored previously while interacting with a new set of neighbors. Of course our team understood that these effects could be pleasing, but they were simply that, effects, of a system that had very sophisticated technology. Why go through all of this trouble? The citizens with detailed knowledge of the software expressed a firm interest in maintaining a gap between The City as it is managed and The City as it is lived. They believed that a dataset producing no spatial outliers would limit developmental and marketing pressures to specific locations. If the new sensors replayed banal activities happening precisely where The City expected them to occur, it would reinforce a simple view of urbanity while protecting a thriving and potentially experimental culture. Real urban life could exist while defying measurement. Ironically, intense sensing from above gave the citizens this cover. It was a repeated claim from those we interviewed: high resolution data, confirming exactly what the planners expected to find, meant The City could continue to be managed from afar in the abstract space of data, while still being lived in an experimental fashion, on the ground. 11 July The findings from our field study radically altered our approach to the land use maps. We began again with a new set of criteria, and a stronger sense of what the “constitutive infrastructure” could be. We finished the design in six weeks and delivered to The City on 5 July. (Photographs of the final land use map are included with this article.) 18 July Two weeks after delivering the map, we received a negative response from The City stating that we had not fulfilled the requirements of the project. This was especially problematic given the additional time we had requested for the field investigations. The reply culminated in a curt and damning question: “Is this thing even a map?” We attempted to address their concerns with a description of the new map’s operation and how it related to the project brief as we understood it: The land use map consists of three main components: the Scene, the Sensor, and the Display. The operation of each component is highly dependent on the others. The Scene is projected vertically onto any large surface. The Sensor films the Scene at a rate of 1 frame-per-second. The video from the Sensor is fed to the Display. An Observer positioned in front of the Display can view, in a single glance, the entire flow of information from Scene to Sensor to Display. (Note that the distance between the Scene and the Sensor is proportional to the distance between The City’s air-borne sensor and the ground; this is a scale map.) The video content of the Scene presents an abstract aerial view of a ground condition using nested circles of varying radii. A direct observation of the Scene makes it readily apparent that this is in no way a photograph but a highly synthetic image of geometric shapes. Across the surface of the Scene flow hundreds of algorithmic Citizen-Agents. They are projected at a rate of 24 frames-per-second. Observing the Agents directly they appear to circulate across the Scene randomly. This is not the case. The Agents have a special relationship to the Sensor: Their Paths have been computationally-evolved to deviate maximally from the Scene roads for 23 frames, yet cross a road at the precise moment that the Sensor records the Scene. The Display is connected to the Sensor. The Display is updated once a second with the Sensors newest measurement. The video of the Scene that appears on the Display is a photo-realistic image of hundreds of agents moving in an orderly fashion along the Scene’s major roads. An Observer watching the Display can also see the Scene. The graphics used to render the Scene were chosen to interact with the Display’s color and resolution. The image produced on the Display looks more realistic to the eye than its source Scene image projected on the wall. The Observer also sees the difference between the apparently chaotic motion of the Agents in the Scene and the orderly result on the Display. Our new Land Use map met the requirements of the brief to the best of our skill and knowledge. We documented surprising new uses of space that were not indifferent to the technologies of measurement applied to them. The sensing, data analysis, and resultant urban decision-making are the “constitutive infrastructure” that shape these land use behaviors. With this contract, we were presented with a continuous chain of information and influences: the conscious and unconscious spatial strategies of citizens; aerial sensors and their techniques of measurement; the translation of data to visualizations; the limitations of an interpreter’s eye; the material properties of interim land use maps; committee decision-making procedures; the techniques of urban policy implementation. By requesting a map of contemporary land use and the infrastructure that shapes it, it is this web of interactions that we have been asked to map. To address these concerns in total we were forced to move away from static, two-dimensional techniques and instead create a new optical machine. We encouraged The City to consider the new map with the same amount of openness and enthusiasm as they treated their new airborne sensing platform. 11 August We received our last reply from The City. They claimed that they had a better understanding of our map and how it responded to the project requirements. They found the description of citizen strategies “puzzling but fascinating” and agreed to spend more time studying our findings. We were told to be prepared for follow up conversations on the appropriate next steps. Three months later I saw a new Land Use map of The City. It was rendered at 1:10,000 scale using a subset of SLUC categories.