We would like to thank Cristina Grasseni and Federico De Musso for their invitation and editorial work. The paper benefited from the generous comments of two anonymous reviewers, Tomas Percival, and from our colleagues in the Security Vision Research Project. This work was supported by funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (SECURITY VISION, Grant Agreement No. 866535).

“A wheel turns because of its encounter with the surface of the road; spinning in the air it goes nowhere. Rubbing two sticks together produces heat and light; one stick alone is just a stick. In both cases, it is friction that produces movement, action, effect.” (Tsing 2005: 5)

I. Introduction: Imagining a Collaboration

2Our team is diverse. The three members of this mapping project come from three different disciplines. It is however difficult to demarcate exactly where the borders between disciplines start and end, as we have each followed our own interdisciplinary trajectories. Francesco is trained as an international politics scholar, but his work is mainly influenced by sociology and anthropology; he also holds a degree in film and developed a practice as a filmmaker since he began working on his PhD. Ildikó is trained as a visual anthropologist, later her curiosity about vision as technology brought her into the field of science and technology studies (STS). Ruben, while originally trained as a filmmaker, combines a research-based approach with a digital arts practice, in which working with technology (programming) is a method of inquiry. Our interdisciplinarity is thus exponential (Lingard et al. 2007): an interdisciplinary assemblage of interdisciplinary trajectories. The sum of our trajectories and the networks of (scholarly) colleagues, disciplines and sub-fields thus shape both the conscious and the embodied intellectual inclinations with which we come to the team (Bourdieu 1992).

3Furthermore, while we strive for a more horizontal collaboration, the team has a formal hierarchy which cannot be ignored: Francesco is an associate professor, Ildikó is a post-doctoral researcher, Ruben is a PhD student. As the Principal Investigator of the project, Francesco secured the (external) funding and selected the team, rendering him ultimately responsible for the project’s finances and obligations. The project thus takes place in a context of project-based neo-liberal parameters, involving the short-term duration of the research, fixed-term working contracts, and deliverables (Richter and Hostettler 2015). The interdisciplinary nature of the team however immediately disturbs this hierarchy: de facto each of the team members is the authority in their field of expertise. Additionally, an explicit agreement is made from the start: the team is not gathered to execute a pre-determined vision, but rather each member is invited to contribute from their specific positionality to the common objective.

4In recent anthropological scholarship collaborations are explored not only as the fundamental mode of engagement in a field setting, for instance with research participants and interlocutors, but also as an emergent way of doing research in teams rather than alone, sharing data or thinking and writing together (de la Cadena 2015; Niewöhner 2016; Estalella and Sánchez Criado 2018; Boyer and Marcus 2020; Murphy 2020). These new engagements, often happening across various fields and disciplines, also engender new possibilities for conceptualizations and theoretical interventions (Boyer and Marcus 2020).

5Visual anthropology in particular has been grappling with questions around negotiation between different epistemic practices. Thinking about collaborations in terms of inter-spaces, Schneider and Wright (2010) for instance focus on interactions and appropriations between different (visual) fields. Differences are seen to be productive resources both in terms of conceptual interstices and material practices even when they are accompanied by a sense of strangeness or failure (Grimshaw, Owen and Ravetz 2010: 161). These differences can also lead to new practices that push the boundaries of each of the involved disciplinary fields further. These can be the expansion of cinema into digital platforms (Favero 2017; Weidle 2019) or bringing ethnographic materials into experimental spaces and art galleries (see for instance the Ethnographic Terminalia Collective in Smith et al. 2021). In this article, we contribute to these discussions by foregrounding frictions and telling the story of how we gave up collaborating and turned instead to co-laboration.

6To make a distinction between collaborations that are often problematized from an ethical or political perspective and the joint epistemic practices that often take the form of experimentations, Niewöhner (2016) proposes the concept of co-laboration. “With co-laborative, I mean temporary, non-teleological, joint epistemic work aimed at producing disciplinary reflexivities not interdisciplinary shared outcomes” (Niewöhner 2016: 2). Reflexivity here is understood as an experiential and experimental problem with an emphasis on mobility between fields, disciplines, methods and theories. Niewöhner also emphasizes the role of curiosity in how researchers with different backgrounds negotiate their interests in co-laborative spaces without necessarily working towards a single shared outcome. Creating these spaces often requires commitment and courage to leave one’s (disciplinary) comfort zone, by crafting experiments (Mann et al. 2011; Sánchez Criado and Estalella 2018; Bleumink et al. 2021) and embracing “radical not knowing” (Stengers 2018; De la Cadena 2021).

7The main argument we develop in this paper is that in an interdisciplinary and participative collaboration – as audio-visual research projects often are – ontological, epistemic and methodological considerations are more than just features of the intellectual coherence of a research project. They are the sites where researchers embedded in different traditions and trajectories articulate a politics of academic collaboration; in which conflicts around intellectual and emotional attachments as to what constitutes “good” science, appropriate methods, and the proper use of one’s time and effort can be negotiated and resolved. They are also the sites where professional identities are re-articulated and re-situated into their respective ecologies of practice (Grasseni and Gieser 2019). In other words, conceptual choices are not only what make the research, they are also the social glue that allows for researchers with conflicting (professional) identifications, habituses and emotions to come together around a shared space that fosters collective research. From this perspective, the question then becomes: how are discussions around the scientific dimensions of research also discussions about how to work together?

8As this article will show, the collective decision, in our particular case has been to imagine a space of intellectual autonomy while retaining the collective goal. Rather than attempting to reconcile epistemological, theoretical and methodological differences into a single coherent project, we turned instead to thinking about our “co-laboration” as emerging from the interstices, intersections and overlapping points of contact between autonomous epistemic spaces. Borrowing from the work of Karen Barad (2007), we conceptualized this approach as “diffractive”, as it creates a space for the various projects to be read “through” one another.

II. The Birth of a Centipede

9The initial plan for the mapping project was to collect exploratory data, both qualitative and quantitative about actors that develop or use security technologies based on computer vision, such as the ones used in biometric border technologies (Amoore 2011), smart CCTVs deployed in the context of “smart cities” (Kaufmann, Egbert, and Leese 2019) or for the purpose of social media content moderation (Gillespie 2020).

11The group thus set up a server, installed a Semantic MediaWiki instance (a version of the software used in Wikipedia) and started to collect data on the basis of open source information found in online newspapers, official press releases, manufacturers’ products advertisements. The software chosen had a feature that satisfied the initial ambitions of the project: while allowing to store and query structured data, it also allowed for narrative descriptions. It was a truly quali-quantitative data collection device (Venturini and Latour 2010). The dataset was structured around a certain number of entities: institutions, people, deployments, technologies, cities, countries. Each of these entities was given a list of attributes: creation date, geographical location, type of technology, etc. Over the course of the data collection, these attributes were tweaked — new attributes were added, properties changed — in response to the cases we collected. We decided Ruben would create an initial network visualization of the data we had collected so far, in order to obtain a preliminary view of how the data was related: which were the central companies? The most important clients? The most used and recurring technologies? And there it was:

12“This looks like a centipede” Ildikó exclaimed in awe when Ruben first presented the graph. While Ruben and Francesco were glad to obtain a visual representation of the data they had been collecting through a relatively known method of network representation, for Ildikó, this image triggered a different line of reasoning. As we know from Goodwin (1994), our ways of seeing are socially and professionally embedded. Our eyes have been trained in different epistemic traditions (Grasseni 2009) therefore we, very concretely, saw different things. Such network visualizations were not new to Ruben’s and Francesco’s eyes. For Ildikó however, who never worked with quantitative methods before, this image seemed to come alive as a “centipede”. Ruben’s visual output, programmed with the D3 JavaScript library, allowed the nodes to be dragged around making the network float and stretch. It was also not a static graph. As more data was entered into the Wiki, or occasionally deleted, the centipede grew or shrank. This movement was very literally caused by the changes in the database. And as such, it steered the attention away from the finished and cleaned database towards the process of assembling and shaping it. Imagined by Ildikó as a centipede, this data visualization therefore became a curious visual object, one with an added narrative potential. It could tell the story of a database in its becoming, making visible not only its feeding and growing but also its shrinking when data was deleted or re-tagged. It also told a story of the unfinished nature of the database not to be taken as a pre-existing object out there but an object created and shaped along particular theoretical and methodological curiosity.

13In retrospect, interpreting the data visualization in terms of a “centipede” was also a way to establish a common language within the team, creating a bridge between a quantitative method output and methodological inclinations for a qualitative, sensorial rendering of social life. As Ildikó wrote in her notes at the time: “The centipede is our conversation partner. It might help us to carefully attend to our methods so as to draw attention to the frictions, conflict and ambiguity that otherwise would be cleaned out from the data.” Revisiting these notes now makes it clear that, already at this stage of the project our weekly debates that revolved around the theoretical and methodological underpinnings of the project were foreshadowing our frictions. However, at this stage the centipede still mobilized our imagination. Even though it appeared to us as clean lines on our monitors, it had the status of a preliminary object, a sketch. It therefore allowed for the freedom to read different things into it. For Ruben it was a lens into the data structure and the collection process, Francesco saw the first draft of a social network analysis, and Ildikó sensed its narrative potential. It was a shared epistemic object (a being even) around which we could rally and on which we agreed: “this is our mapping project”.

III. The Rigidity of Projection

14The collaboration seemed to be going well. Francesco, Ildikó and Ruben collected data, the centipede grew, and Ruben offered several visualization options. As the team was working with the centipede, it received an offer from a policy actor - the Green Party of the European Parliament. The Greens were commissioning a report on the uses of biometric and behavioural surveillance technologies currently in use in Europe. They were looking for academics working on the issue, who would be able to “map the field”. The commission seemed to match our own research to such an extent that the team decided to accept it. After the network visualization, and some trials with an alluvial diagram (used to illustrate the relations between various categories), Ruben thus started working on a geographical map with the policy report in mind (Ragazzi et al. 2021).

15Here a first conflict arose: as the geographical map had to be part of a policy document, its primary function was no longer to explore data, but rather to convey information for a publication and potentially serve as a (visual) tool for political intervention. Initially, the (still rather vague) idea had been to generate various visualizations of our dataset, and program an interface that would allow it to transition from one into the other, based on the parameters given by the user. Therefore, Ruben had chosen not to use an “off the shelf” pre-existing library for the visualizations: such an option would constrain our options in the future, while still requiring similar work to process the data. But now a concrete map had to be created. The requirements for a digital sketching device and an end-product made for publication are substantially different. Instead of a sketch, composed of a few broad strokes which convey the meaning and intention of the map, now all the details needed to be specified. For example, how details on the map (dis)appear at various zoom levels, and how legibility is preserved. It needed to function without glitches, across various platforms and devices. This meant that Ruben spent long hours coding simple functions for the interface. He began expressing his frustration at spending so much time coding with the purpose of presentation rather than research. He had to address a flow of requests from his project members, instead of interrogating the meaning of the lines and dots as they appeared on the screen. Collecting and visualizing data in this specific context did not allow us to interrogate what Security Vision is, Ildikó and Ruben felt. Rather than functioning as a tool for thought, the map presented what for us constitutes Security Vision. It did not question it. There were little insights to be gained.

17Ruben and Ildikó also had another concern that would become more and more visible once the map was shaped. For them, a reflexive approach to data visualization and research —explicitly or implicitly— should interrogate not only the contents of the dataset, but also the way it is constituted; its structure and modes of collection (Marres and Moats 2015; Martin-Mazé and Perret 2021). The Semantic MediaWiki we had been using was not a neutral or transparent ordering tool, but rather an epistemic ‘device’ that “assemble[d] and arrange[d] the world in specific social and material patterns” (Law and Ruppert 2013: 230). A visualization of our dataset therefore, needed to be reflexive of how entities were included or excluded; how our results relied on the stories the Google and DuckDuckGo search engines surfaced; how our own biases and assumptions shaped what we chose to include or exclude. Should it not be these practices that are questioned? The risk, according to Ruben and Ildikó was to replicate the acts that have haunted the quantification and vectorization of the world which are also so fundamental to the digital security practices that we are studying (MacKenzie 2017; Rieder 2020).

18Ruben and Ildikó felt that the more the two-dimensional map of Europe solidified, the more it resisted any measure that challenged its “view from above”. Donna Haraway warned against “the god trick of seeing everything from nowhere” (Haraway 1988: 581). This is an optical illusion, she argued, that hinders, rather than serves objectivity. Hence her call for situated knowledge: “positioning is, therefore, the key practice in grounding knowledge organized around the imagery of vision […]. Positioning implies responsibility for our enabling practices” (Haraway 1988: 587). Drawing on Haraway’s work, in science and technology studies, a tradition Ildikó was most familiar with, Annemarie Mol and John Law (2002) propose the trope of the path instead of the map

‘as a way to describing the world while keeping it open, ways of paying tribute to complexities, which are always there, somewhere, elsewhere, untamed: to list rather than classify; to tell about cases rather than present illustrative representatives; to walk and tell stories about this rather than seek to make maps‘ (Mol and Law 2002: 16-17).

19Collecting, cleaning and labelling data about companies, deployments and data-sets entailed a meticulous work of analytically decoding information retrieved from news media and company websites and re-coding them into the Semantic MediaWiki categories. This process of translating data (Latour 1999) from online content (articles, company websites, policy documents) into our data-set involved a set of decisions about what to include and exclude, or how to fashion our data so as to fit the knowledge infrastructure of the MediaWiki. While collecting the data, these limitations were not so problematic, as the MediaWiki clearly exposed the framework of data collection. The data structure was visible, as well as the author of the data and the history of edits. As such, the data in the MediaWiki reflected our search for how to define “security vision”. For example, in our initial data collection we included cases of automatic number plate recognition (ViNotion 2018). These could be seen as either part of, or outside security infrastructure, depending on how one cuts out their research object. As the geographical map emerged, eventually we decided to leave such cases out. All the painstaking work of creating, cleaning and shaping the data, all the decisions of what to include and what to leave out were suddenly rendered invisible.

20This circumscription of our subject meant not only cutting out certain initially considered instances, but also solidifying blurry boundaries. For example, leaked documents suggested that several law enforcement agencies from the European Union were on a client list of the United States based facial recognition software company Clearview AI. However, the extent of their use of the software was largely unclear, which lead us to include merely two of these cases in the map, Interpol and the Swedish Police. Interpol for instance, despite claiming that “there is no formal relationship between Interpol and Clearview,” in another place admitted that they had used the software as part of a 30-day trial account (Mac et al. 2020). The visualization of these instances as dots and lines in our map solidified fuzzy and uncertain sources into “facts” or “truth”.

21Francesco was aware that these concerns loomed large over the data collection objective. He was however unconvinced by the doubts of the rest of the team. He saw them as secondary to the task of making security assemblages that are often purposefully left in the dark visible, reversing a status-quo in which transparency is only applied to the data-subjects, while professionals in position of authority can claim a right to secrecy (Pasquale 2015; Isin and Ruppert 2020). Of course, he agreed, we need a reflexive take on our data collection process. Yet the “god trick’s view” argument was not convincing for him. The problem of reflexivity could be addressed by making sure that the position of the researchers within the broader set of actors involved in the world of computer-vision based security technologies (universities, companies, clients) was clearly outlined in the mapping process itself. For Bourdieu, indeed

‘scientific objectivation is not complete unless it includes the point of view of the objectivizer and the interests he may have in objectivation (especially when he objectivizes his own universe) but also the historical unconscious that he inevitably engages in his work. By historical, and more precisely academic, unconscious (or ‘transcendental’), I [writes Bourdieu] mean the set of cognitive structures which can be attributed to specifically educational experiences and which is therefore to a large extent common to the same (national) educational system or, in a more specified form, to all the members of the same discipline at a given time.‘ (Bourdieu 2003).

22For Francesco, the methodological concern with the ambition of mapping therefore was not a good argument to give up on the project of establishing a map of “who does what and where”. This was indeed an important step which would later allow for the outlining of the main lines of structuration of the social universe we were trying to construct as our object of analysis. Giving up the idea of mapping, he believed, would mean falling into an epistemic fallacy according to which, as no map is “objective”, no map is possible. 

23The sketch-like centipede had still allowed the freedom of interpretation. It allowed each of us to imagine, or read into it the different epistemic imaginaries. Ever since the moment the group started to work towards a concrete map, the abstract concepts—which we had seemed to agree on in our conversations—when transformed into concrete cases, started to materialize in such a way that the differences between them could no longer be ignored. The visual logic of the map suddenly made the edge cases explicit and started to expose the intellectual shortcuts that we had to make when collecting the data. The moment these shortcuts solidified in a visual object, the map, the compromises became all the more apparent. As we were all committed to the project we kept on coming together, discussing and debating. However, the more time we spent on trying to figure out a shared epistemic horizon, the less the centipede grew. While for someone else—or even at another time—these discussions might not be a dead end, we should have shifted registers. However, because of our desire to stick together as a team, we only dug ourselves further in.

IV. From 2D to 3D to Letting Go of Dimensionality

24The group did not immediately abandon the idea that the initial project could be preserved by embedding some of the doubts mentioned above into the data collection and visualization itself. One issue, the fact that most data’s certainty was questionable, could be made visible in the map. The group thus started actively discussing the introduction of the idea of “uncertain” data: not so much, as it exists in many quantitative disciplines in terms of error estimation, projection, or margin of error, but at a more fundamental level the fact that some information had a dubious status: rumour, gossip, conjecture. Given the degree of secrecy in the field of security, this seemed to be a key component of the data collection.

25Ildikó and Francesco became enthusiastic about the theoretical prospects and started wondering how to map the large amounts of unknowns around the uses of computer vision in the field of security. Ballestero’s (2019) work on the practice of "remote sensing", i.e. the practice of establishing a map of remote areas and their materials through technologies of long-distance survey provided a new metaphor for a few months around which the different traditions could perhaps converge. “Remote sensing” would be not only about uncovering, but also leaving room for qualified probabilities: gossip, speculations and uncertainties. How do we make the map address the complexity, the messiness, the ambiguities, that we knew were implicitly present in the data we had collected? Or better, how could we use ambiguity and uncertainty as the basis on which a representation is constructed (Drucker 2014: 127)? The visualization of our data as unambiguous nodes and edges on a bi-dimensional plane did not allow for uncertainty to come through. Francesco and Ildikó were imagining visual metaphors to pursue this theoretical proposition – “covering and uncovering”, “above and below the surface”. The idea of 3D renderings of digital surfaces appeared as a possible solution.

26However, as Ruben argued, the problem of the map did not lie in its number of dimensions. If such a three-dimensional visualization does indeed address the issues of complexity, should we not then add a fourth, fifth, ... or n-th dimension, in order to account for other grey areas in our collection? As Johanna Drucker (2014) argues, the mode of visualization is a crucial aspect of the knowledge it produces. The issue lay not in the number of dimensions at our disposal but rather in the act of projecting, of pinpointing an entity to a fixed position in the multi-dimensional space while invisiblizing the collection process. We needed to open up the space for interpretation, but interpreting space in a radically different way.

V. Diffracting Maps: Towards Epistemic Autonomy

27The group had come to an apparent collaborative dead-end. After a long summer break, we thus decided to change the parameters of the collaboration. It was a happy moment in September when we got together and put all our dissatisfactions with the map on the table. One issue we had been struggling with was how to define the contours of our object of analysis. Francesco believed that it was possible to start “in the middle” by identifying the main actors, their properties, and the different lines of opposition between them through the exploratory data collection. The boundaries of the field could thus be decided down the road, by looking at the logics of exclusion and inclusion which were enacted by the field itself (Bourdieu and Wacquant 1992). For Ruben and Ildikó however, our initial data collection based on web searches merely served to present our pre-existing notion of security vision, rather than to expand and challenge it.

28Ildikó emphasized that computer vision in the field of security was not “out there” to be collected as data, organized and visualized but it emerged through messy entanglements between human and non-human actors, materials, discourses, technologies, professional and bureaucratic practices, institutions, politics, and not least, our own act of mapping. If we took these entanglements seriously, security vision would emerge as a complex research object, inseparable from the practices in which it is situated (Haraway 1988). Francesco shared the concern that a visualization of a dataset is always the result of, and premised upon, the scientific construction of the object of research—he did not hold a positivist perspective according to which the map would be a simple “translation” of a world out there. However, whether this was relevant or not for the exercise of the mapping was a matter of emphasis on, or interest in, different ontological moments. It thus turned out that we had gotten caught up in fine tuning our arguments, while being oblivious to the fact that we diverged too much on the epistemological status, or goal, that the map had to serve to come to an agreement. While not fundamentally disagreeing on many important premises, each of us was ready to make different trade-offs. To give space to these different considerations, we decided to let go of the idea of an ontological and epistemological consensus. We decided to stop working collectively towards a unified, coherent, systemic visualization grounded in a single dataset. The group decided instead to pluralize the maps.

30Producing these diagrams relays the question of what constitutes security vision to practitioners in the field. Although we should be careful about contrasting the affective nature of a hand-drawn diagram to that of a digital drawing, their “sketchy” nature suggests their status as a conceptual aid. “The diagram here is a strategy of experimentation that scrambles narrative, figuration—the givens—and allows something else, at last, to step forward. This is the production of the unknown from within the known, the unseen from within the seen” (O’Sullivan, 2016, p.17). Moreover, diagramming, O’Sullivan explains, allows for a composite practice in which drawings from “different milieus”, from different frameworks, can be superimposed on to one another. These diagrams might help to work out possible relations and divergences between the various imaginaries of security vision. It refuses the projection of a bird’s eye view, and opens up for an ontological multiple.

31Agreeing on our desire to work from ontological and epistemological autonomy, we were thus faced with the question: how does one articulate this paradoxical proposition of unity and diversity in our mapping project? The work of Karen Barad (2007) provided us with a starting point. Our multiple engagements, we reasoned, would allow us to produce diffracting maps. A diffractive methodology, Barad explains, takes the various epistemological fields seriously in their own right, as each generates their own insights. The diffraction emerges when works from various fields are brought together; like light waves, their insights will amplify and dampen each other, forming new patterns—new insights. In our work, introducing not only other maps, but accepting the diversity in the ontological, epistemic and methodological assumptions that come with different practices of map-making (or path-finding) would move us forward. We thus pluralized the map making: each of the mappings addressing the question of “what is security vision?” on its own terms.

32While it is too early to draw conclusions about what these diffractive readings will yield, we can share some preliminary insights based on our first sketches. We thus diffractively read two maps. The first map (Figure 5) is the network analysis that Francesco is working on, based on data from the Semantic MediaWiki. This map represents clearly demarcated entities (i.e. a “company”, a “deployment”, a “technology”) as “nodes” (the dots in the network) and a simplified set of relations (“is part of”, “is located in”, “is managed by”) as “edges” (the lines between the nodes). It represents entities by different spheres, additionally characterized by text labels. The visualization deviates a bit from the traditional network visualization of nodes and edges by the introduction of a plane, that divides the three-dimensional space between “known” entities (for which we have information) and “unknown” ones (for example, we know a municipality deployed a surveillance system, but we do not know which one). The second map will be constituted by the diagrams produced by Ildikó and Ruben’s interviewees. In these diagrams, the relations between the drawn entities, i.e. companies, ethical concerns, private/public spheres, grants, the public and the mechanisms of computer vision cameras, all find an co-existence on the two-dimensional plane.

33A diffractive reading, Barad says, is about “reading insights through one another in attending to and responding to the details and specificities of relations of difference and how they matter.” (Barad 2007, 71) In other words, while each map enacts “security vision” on its own terms, it is the difference between these maps that attunes us to how these maps differentiate – how they come to matter. The network analysis, gives a sense of scale provided by a simple quantification: it helps us to identify the most present and recurring companies, which countries have the most deployments, which technologies are used most and which technologies are (still) rare. Moreover, the map renders visible how security vision happens across various spaces. Even when a deployment seems to involve many local actors, we can always see an intertwined transnational network behind it. Furthermore, the plane of the unknown entities gives presence to those cases and institutions that were inaccessible during the research – it gives prominent place to the dead-ends which so strongly shape our object of research. While we cannot project the lines and narratives of the diagrams into the space of the network analysis, they draw our attention to the complexity of all the social, economic and political relations between these entities. They thereby affectively charge the relations of the network analysis. Lastly, in practice these maps are deeply entangled: for example, the cases and institutions that appear in the Semantic MediaWiki and the network analysis proved a valuable resource to find interview subjects, whereas the narratives that flow from the interviews lead to unexpected actors in the network analysis – such as the use of residents’ domotica in the case of the Burglary Free Neighbourhood (see Figure 4). While these are just initial thoughts, they show the productive potentials of diffractive reading, of looking at the two maps through one another.

34Thinking about it in retrospect, it was not only through the map-making but also through the process of writing this article, that we became aware of the intertwined dimension of our epistemic choices and the politics of our collaboration. For Francesco this echoed the analysis of Ernesto Laclau and Chantal Mouffe (1985) around the notion of “agonistic pluralism”. Laclau and Mouffe argue that the search for consensus in diverse societies inevitably leads to forms of domination by the majority identity group. Rather than seeking consensus, which in our case would inevitably lead to imposing one option over the other, our strength as a collective we reasoned, resides in the diversity of our identities as researchers and the various ecologies of vision (Grasseni and Gieser 2019) we brought to the project. Although we did not formulate it then explicitly as such, we reasoned along similar lines to Laclau and Mouffe when we decided that the best way forward was to work from zones of intellectual autonomy, along two parallel tracks: first within the confines of our epistemologically and ontologically coherent domain and second, in dialogue—not necessarily in agreement—with the other members of the group. Instead of collaborating we started to co-laborate (Niewöhner 2016). The idea of working at times together, and at times alongside, suddenly freed us from the conflicts that had slowed our collaboration. Our co-laboration would thus rely on constant engagement and collaboration from a position of heterogeneity and autonomy, with the understanding that our professional identities are never fixed: the process of co-laboration reshapes not only the boundaries between us but also who we are as social scientists.

35In the end, instead of one map, a constellation of maps will appear. These different maps—the network map, the geographic map and the diagrams—with their lines, shapes, and colours, will capture the relations between the actors of security vision in several two-dimensional planes. Yet all three maps are contingent on the methodological devices and the knowledge infrastructure they build upon. Although different, they still hang together (Mol 2002). The co-existence of these maps, and the diffractions they can cause amongst each other is reminiscent of the fact that we, in our co-laboration, did not subsume the virtues of our own trajectories as researchers to the common goal of a single output. Rather, the common goal was reformulated to accommodate the slight variations in vision and the epistemological consequences thereof.

VI. Conclusion

36In this collaborative mapping project, we bring together different approaches to mapping, so as to harvest the generative frictions between them. By taking both the modes of engagement and the materials and technologies involved in imagining seriously, creating and eventually multiplying our maps, we also attend to the politics of our collaboration.

37We started our research project by collecting data about the developments and deployments of computer vision in the field of security with the hope to produce visualizations that would help us articulate our research questions, and facilitate multi-modal engagements with the field. As we came to the project from different interdisciplinary trajectories our discussions quickly touched upon more fundamental theoretical and epistemological questions. These questions became even more visible when we started to get our hands dirty, collect data and produce visualizations. As dots and lines (the data entries and relations between them) became facts on a geographical map, the epistemic frictions among the three of us started to solidify as well. The visualizations were not mere outputs anymore but epistemic objects that fuelled our disagreements and became the obstacles for our collaboration.

38Data visualization is a practice rooted in a diverse set of intellectual and artistic traditions. Embedded in different ecologies of practice, these traditions also engender different ways of engagements (Grasseni and Gieser 2019) that were causing our frictions. To ignore this in an interdisciplinary team in order to pursue a narrowly defined vision of what should be visualized, thus ignoring the diversity of the team, is an arrangement likely to cause unproductive outcomes. In a more horizontal collaboration—as opposed to one in which a hierarchically organized structure is set up to execute an idea without much input—we chose not to work towards the goal of a consensual language, but instead towards the fostering of a conceptual (ontological, epistemological) space in which different academic, intellectual and artistic traditions could meet without being forced to blend. It is a space in which practice is understood as a form of conceptual work (and thus not reduced to execution) and where each researcher can interact without giving up what constitutes their values and beliefs. We had to take our frictions seriously and open up the collective space for intellectual autonomy in which instead of trying to reconcile our theoretical, epistemological and methodological differences we could find a way to co-laborate in the intersections and overlaps between autonomous epistemic spaces. In doing so, we embraced a diffractive approach.