Visual (data) observation in International Relations: Attentiveness, close description, and the politics of seeing differently

Abstract

Has Visual International Relations (IR) become too distant from the content of visual artefacts? This is a paradoxical question. Visual IR is a vibrant and pluralist field exploring visuals in innumerable ways. Nonetheless, the field tends to focus on ‘deep’ readings of the socio-political implications of visual artefacts at the expense of a close and attentive observation and description of the events, situations, or phenomena they may depict. Simply put, visual IR usually analyses visuals-as-visuals rather than seeing them as entry points for studying the social world. But might a video of torture teach us something about the practicality of torture? Might a video of peace negotiations teach us something about their successes or failures? Can we gain a fleeting glimpse of ‘reality’ within visuals? We address these questions by first situating our focus on close ‘visual (data) observation’ in conceptual conversation with the literature’s existing focus on deep interpretation. Second, we outline three approaches to visual observation as they are deployed outside IR. Third, we unpack how those approaches might be of value for IR, especially vis-à-vis the study of practice, materiality, and discourse. Finally, we conclude by asking if visual data observation can retain critical political potentiality.

The study of visual politics has matured into a diverse, empirically grounded, and conceptually sophisticated field. Empirically, the range of artefacts explored is now vast. Photographs, comics, videos, popular cultural visuals, computer games, artworks, memes, satellite images, digitally produced visuals, and more are all being analysed.¹ The conceptual approaches deployed to explore such artefacts also now cut across a heterogeneous set of disciplines including social theory, science and technology studies, art theory and aesthetics, film studies, literary theory, digital humanities, International Relations itself, and far beyond. Equally, there is now a growing – again too diverse to enumerate² – ‘repertoire of multiple, pluralist methods for the visual analysis of international relations’.³ This includes approaches centred around material-semiotics, discourse analysis, content analysis, (auto-)ethnography, data scraping, participatory photography, photo-elicitation, film-making, data visualisation, and many others. Simply put, visual International Relations (visual IR) is now a vibrant, diverse, and established field producing work of key conceptual, empirical, and methodological importance for the discipline and beyond.

In light of that intellectual vitality, this article asks a perhaps unusual question: has visual IR become too distant from the content of the artefacts that it analyses? Following Heather Love, we suggest that the field is preoccupied with ‘deep’ readings of the politics of the visual at the expense of a ‘close’ reading of the content of visual artefacts that sees them as ‘windows’ for the observation of the social world.⁴ In part, this status quo is a consequence of the intellectual genealogy of the field. As Andersen, Vuori, and Mutlu write, ‘images and visual analysis’ were ‘“latecomers” … in their introduction to the field of IR’ and ‘entered the field mainly through poststructuralism’s concern with countering domination, simulacra, and spectacle’.⁵ As a result, a significant majority of work deploys ‘forms of discourse analysis … [originally] modelled on the study of the text’ to explore visual artefacts.⁶ Such work embraces the polysemy of images in order to unpack their ambivalent political, social, economic, affective, circulatory, etc. effects, or to understand their role ‘in the development of techniques of knowledge’.⁷ Core here is a search for ‘deep political meaning’ within visuals, something that – we suggest – can distance analysis from what is actually depicted in images.⁸ Whether studied as an object, modality, or category of signs, visual artefacts are typically analysed to reveal something about something else (politics, sociality, etc.) or to foreground their polysemic status.

Our goal in this article is to suggest that the study of visual artefacts requires rebalancing. We aim to show that while the search for ‘deep’ meaning is crucial, that quest can be augmented if greater recognition is given to the idea that visual artefacts can also be deployed as modes of – ‘literally’ – observing the social world. For example, in her study of how images become globally iconic, Hansen describes the content of one photograph of torture at Abu Ghraib in terms of (1) the abstract aesthetic composition of its visual form, and (2) the ways in which social discourses have fixed particular ‘facts’ to this image.⁹ The concern is with how these images produce contested political meanings. By contrast, very rare is the reading that seeks to glean something about the practice of torture depicted within images like these.¹⁰ It is in this sense that Love refers to some forms of analysis as ‘deep’ in their search for structures of meaning but ‘distant’ in their comparative lack of concern for ‘attentively’ describing the specific details of an object of analysis in all their minutiae, contradictions, and frictions. Following this, our aim is to explore how visual IR can become both ‘deep’ and ‘close’ (or ‘attentive’) in its engagement with the visual.

For us, it is especially important that Love’s critique of ‘deep’ reading is linked to an interest in ‘observation-based social sciences’ such as ‘ethology, kinesics, ethnomethodology, and microsociology’, which ‘have developed practices of close attention, but, because they rely on description rather than interpretation … do not engage in the metaphysical and humanist concerns of hermeneutics’.¹¹ Unlike most work in visual IR, these observational social sciences seek to describe what people do, and how they do it, rather than necessarily aiming to assign deep (political or otherwise) meanings to these doings. Notably, key fields therein have long deployed visual methods – ranging from photography to video (recording) – to increase their capacity to observe and describe the social world. Interestingly, many such fields (especially ethnomethodology) share affinities with the semiotic (or material-semiotic) precepts central to work in visual IR. However, against the canonical lineage of semioticians (Barthes, Propp, Lyotard, etc.), they focus on observing the ‘real world’ of human practice, rather than describing the structure of representational artefacts (visual or otherwise). For example, Hansen’s description of the aesthetics of the Abu Ghraib photographs is a form of deep or ‘slow’ observation of these images-as-images, but not an observation of the practices of torture depicted therein.¹² It is in this sense that we speak of visual IR as being distant from the content of the artefacts it explores. We wish to ask if it is possible to learn something about the practices, events, and objects depicted in visual artefacts through their close observation and description, rather than an analysis (however intricate) of their semiotic structure, material-aesthetic form, or cognate features.

There are clear reasons why visual IR is reluctant to engage in such acts of ‘observation’ and ‘description’ alongside its usual interest in ‘interpretation’. It is often assumed that such work must embrace a positivist epistemology and naive empiricism that is at odds with both a critical reading of world politics and appreciation for the intrinsic polysemy of the visual.¹³ Indeed, the term ‘observational data’ is more readily associated across visual IR with quantitative political science and a search for causality, which indeed is sometimes the purpose to which visual data is put.¹⁴ The fear is thus that claiming we can observe the social world through visual materials is linked to what Bleiker¹⁵ termed a ‘mimetic’ view of social reality (a correspondence between reality and representation) concerned with the ‘positive creation of laws, models, concepts, and most importantly theories’’¹⁶ This clashes with the lineage of visual IR as a community of scholarship within ‘critical’ traditions.¹⁷ In this view, visual artefacts may be a form of data, but it is one that must be approached through deep critical reflexivity and interpretive methods.

Against such a reading of the place of ‘observation’ and ‘description’ in visual analysis, we follow Neal’s call to ‘reclaim empiricism from the scientific IR tradition’.¹⁸ If we take empiricism as simply concerning ‘the collection of data, information, or empirical material of some kind’, then all visual analysis is empiricist.¹⁹ As we will show, while deploying observational tools is thus indeed empiricist, it is not necessarily naively empiricist. Instead, we emphasise that ‘sometimes describing something’ even ‘without [fully] explaining it’ can ‘say something politically and intellectually important’.²⁰ To do so, we wish to reiterate that our account here attempts to correct an overemphasis on interpretation, rather than distract from the value of the deep and the discursive: we seek a distinct account of how these research traditions can co-produce knowledge. In this, our hope is to show how embracing attentive modes of observation can be allied with deep interpretive practices to produce new forms of politically relevant knowledge. Indeed, Love stresses that engaging in (visual) observation is crucial for ‘their capacity to link together and open up otherwise disparate and ignored features of the world’.²¹ Central to our argument is thus the claim that engaging in forms of visual observation has the radical potential to orient IR towards previously unseen, invisibilised, and sociologically surprising aspects of the world. These reorientations will naturally require the kind of deep interpretive tools that IR is already adept at deploying. In short, our wager is that allying visual IR with this distinct tradition of (visual) observation may ultimately help us see things differently.

With all this in mind, we now seek to justify an expansion of the use of visual data observation in IR. We do so in four main parts. First, we explore what visual observation constitutes concretely by comparing two approaches to analysing the same video: one within the tradition of visual IR and another by scholars from the life sciences, anthropology, sociology, criminology, and psychology. We offer this comparison to affirm that visual observation and description can be achieved in a manner that is neither naively empiricist nor positivist. Second, we disentangle different approaches to such visual data observation, constructing a typology of three approaches that cut across epistemological, ontological, and methodological divides. This allows us to situate our argument within traditions that have the greatest affinities with work in visual IR. Indeed, parts of our argument draw on work in adjacent fields to IR – especially peace and conflict studies and sociology – where the attentive analysis of video data has rapidly grown.²² Third, we discuss three interconnected ‘focal points’ for deploying visual observation in IR: the study of practice, materiality, and discourse. For each, we emphasize how visual observation can (1) open up previously unnoticed areas of inquiry; (2) assist in overcoming methodological dilemmas; and (3) work in productive and pluralist complementarity with other modes of doing visual IR. Finally, we discuss several core dilemmas that remain within visual data observation, especially those which relate to the positionality of the observer, and the question of whether or not ‘observation’ and ‘description’ can contain critical normative-political potential. Ultimately, we conclude by stressing the need to engage in the uncomfortable trans-epistemological work of bridging distinct approaches to visual analysis within IR, allying different traditions to augment the value and promise of each.

A foray into visual (data) observation: From the deep interpretation of polysemy to the close and attentive observation of life

In his compelling analysis of ‘military techno-vision’, Saugmann explores a Wikileaks-released video entitled ‘Collateral Murder’, which depicts a 2007 incident in Baghdad where a US military helicopter killed a group of civilians.²³ He notes that the presence of digital recording mediums within military technologies has now created ‘a visual archive of the violence they deploy’.²⁴ In his analysis of ‘Collateral Murder’, however, Saugmann focuses on the impossibility of using the video as a form of reliable evidence (of war crimes). He does this by first noting that Wikileaks claimed that the video shows ‘this is how it was’ through the invocation of an ‘aesthetic immediacy’ via an ‘indexical-iconic reading of the visual signifiers as causally connected to and resembling the signified’.²⁵ He then unpacks how the US government challenged this claim that the video could ‘resemble’ the signified, instead describing it as ‘unimportant’ and ‘unreliable’ due to a lack of context and understanding of the reality of combat. In consequence, he stresses how even videos that seem to provide unambiguous evidence of reality are encoded within a ‘semiotic fog’. In short, his analysis is an astute reading of how aesthetic regimes are constituted so as to divide what is sensible within (world) politics, intervening in how visuals impact on emotional, political, circulatory, affective, and economic discourses via their entanglement with the technological, the political, and the sensory.²⁶

But Saugmann’s analysis does not tell us much about what is depicted in the video. For our discussion, we thus want to contrast his analysis to that of Elsey, Mair and Kolanoski.who explore the same video through the prism of ethnomethodology.²⁷ At first, similar introductory remarks are made vis-à-vis the polysemic meanings of the video, stressing that ‘Collateral Murder is much watched but not so readily grasped’.²⁸ Indeed, the authors note that viewers are too often ‘guided quickly away from a consideration of the particulars of the situation itself [as depicted in the video] to a consideration of underlying causes’.²⁹ This highlights that controversy surrounding the video largely relates to it being situated as a piece of ‘evidence’ of responsibility for a crime, inflecting analysis of the video to aspects not necessarily present within it: the politics of counter-insurgency, US political and military doctrine, racialisation, hegemony, etc.³⁰ In contrast, Elsey, Mair and Kolanoski. seek to ask ‘in this case [not all cases] what was involved in their [the soldiers’] assessment of the situation’ that led them to fire on civilians. Put differently, the concern in this analysis is whether we can observe data in the video useful for describing how the practical actions of the soldiers depicted came about, rather than subsuming those actions into broader socio-political discourses. Methodologically, they thus ask ‘just as importantly, how can this situational work be recovered from the video?’.³¹

To answer this question, Elsey, Mair and Kolanoski. stress that they do not rely solely on the video. Rather, they draw on ‘vital context and background to the contents of the leaked video’ in their observations.³² This is important. Their observations of the video are not naturalistic but draw on background research giving insights into the ‘gestalt contextures’ underlying human action.³³ Second, following ethnomethodology, they parse the video into a minutely detailed transcript of what can be observed to have occurred. Given the complexity of the video, they draw on what they term ‘a combination of best hearings and visual sense-making practices’.³⁴ An intricate description of the ‘moment-to-moment interactional organization of this violent incident’ is then offered, followed by two conclusions:

1. (1) Collateral Murder shows the ‘episodically organized’ nature of violence, in which the multiple attacks on civilians in the video cannot be treated separately but are intrinsically linked: the ‘situational assessments’ of the soldiers were longitudinal, informed by elements preceding and following the moments in which they fired on civilians. In this, the soldiers developed a ‘scheme of interpretation’ synchronously to the situation³⁵;

2. (2) The violence was also affected by ‘how the pilots acquired their local intelligibility against the background not only of the mission but also of the phase of military operations in Iraq more broadly … this was one counterinsurgency mission amid [others]’. Simplifying, the ‘episodically organized’ nature of this specific situation is also shown to have been indexically linked back to other situations within the counter-insurgency campaign ongoing in Iraq.

There are a few important points to note about the comparison sketched above. The first concerns the goal of analysis. Analytical work such as Saugmann’s can be conceived as ‘studies about’ a phenomenon, which trace the (constitutive) effects of an artefact. By contrast, descriptive work such as that provided by Elsey, Mair and Kolanoski constitute observational studies of the ‘work’ involved in achieving a particular task. To be clear, each perspective is vital. Nonetheless, we would accord with Garfinkel. Lynch, and Livingston that studies about visual artefacts are relatively ‘commonplace’ whereas studies of the ‘work’, ‘action’, or ‘forces’ depicted in visual artefacts are relatively ‘rare’.³⁶ The value of expanding the range of studies engaged in work such as that epitomised by Elsey, Mair and Kolanoski lies in its ability to unpack the ‘quiddity’ (or ‘just-whatness’) of human practice. For example, it is notable that while Saugmann’s text explores the ‘fog of war’ argument offered by the US military to defend its actions, we do not gain a sense of what such a fog of war constitutes. Elsey, Mair and Kolanoski demonstrate the ‘reality’ of such a lack of contextual certainty at a granular empirical level. In doing so, however, they do not necessarily affirm that this is a ‘defence’ of the soldiers’ actions. Indeed, their second finding could be used – as we will discuss in the conclusion – as a means of political intervention critical of the US military: why/how was US counter-insurgency strategy designed/implemented in way that it created a sustained ‘local intelligibility’ that predisposed soldiers to actions such as these? Most importantly, it allows us to ask these questions not abstractly but on a bedrock of firm(er) observational evidence.

To return to Love, observational work of this kind embraces an ‘alternative ethics’ of attentiveness.³⁷ Observation requires us to ‘disregard … not … take seriously, how closely or how badly the object corresponds to some original design – particularly to some cognitive expectancy or to some theoretical model’.³⁸ As such, there is an implication that we must ‘validate actors’ own statements about their behaviour rather than to appeal to structural explanations’.³⁹ We should take the helicopter soldiers seriously in their words and actions, rather than immediately reducing them to cogs in military machines. An underlying assumption here is that there is often a ‘mundane absence of explicit contextualisation as a routine feature of sensemaking’.⁴⁰ Most mundane (quotidian, everyday) practical activities lack a clear context or ‘script’ to follow of the kind that would require us to turn to deep levels of interpretation for an explanation. Instead, many situations – such as being in a helicopter gunship – involve intense local and improvised forms of problem-solving, given a ‘stable context’ is exceptionally rare in life. Erin Manning makes this point evocatively in her discussion of human movement, noting that even when carrying out a task as seemingly simple as ‘walking’ in and out of a metro train, we are obligated to engage in intense problem-solving activities to avoid bumping into others, an achievement that requires attentive observation to be understood.⁴¹

This subtle art of attentiveness at the heart of the observational sciences, which we believe is of equal value to IR, thus focuses on teasing out how human beings interact with other human and non-human beings in ways that ultimately generate the social world. This can be achieved in very different ways. But the ultimate ethos is that it is not always enough to ‘go beyond the surface’ of the visual to understand what it depicts. In this regard, it is important to reconsider Bleiker’s words that ‘our effort to make sense of’ a political event ‘can never be reduced to the event itself’.⁴² Key in this phrase is the term reduced to. We accord with the view that visual analysis must not be reduced to the observational or the descriptive. To do so would be to risk a naive empiricism and a denial of politics. At the same time, we do not believe that our efforts to ‘make sense of’ the world should too quickly depart from ‘the event itself’ when more attentive and potentially enlightening methods for observing that event are possible. We should reduce visual IR neither to ‘deep’ interpretation nor to ‘close’ description. Indeed – as we will see later – our view is that allying these two sensibilities more closely can open up for radical new forms of socio-political critique of the kind that reductionist approaches, in either direction, may find it difficult to leverage.

Three ways to observe visual data

The example of visual data observation sketched above is situated within the ethnomethodological tradition. But it can take many other forms. Before we discuss the potential contributions of the close observation and attentive description of visual data for IR, it is thus necessary to disaggregate these approaches. To do so, we describe three common but contrasting approaches – the behaviouralist, the performative, and the ecological (see Table 1). Our goal is twofold. First, we wish to stress that visual data observation can be naively empiricist, especially within the positivist tradition, and distance our argument from that work while also stressing that IR’s reluctance to engage in observation unnecessarily cedes territory to this approach. Second, we introduce the performative and ecological approaches as being of particular value to IR, while also stressing areas where work in visual IR overlaps, but also seek to emphasise how these overlaps tend to follow the epistemological/political preoccupations of the field just described, potentially problematically limiting IR to ‘comfortable’ terrain within acts of visual observation.

Table 1. Ideal-type approaches to visual data observation

Behaviouralist observation

A behaviouralist approach to visual data observation embraces a positivist epistemology. Nassauer and Legwie offer a formal definition describing how this approach explores:

situational dynamics and behaviours using video or other visual data to understand how people act and interact, and which consequences situational dynamics have for social outcomes. This perspective helps with understanding the rules and processes that govern social life, both in everyday encounters and in extreme situations.⁴³

In this view, visual material allows us to identify social ‘rules’ or ‘processes’ through the observation of micro-level situations and the interactional dynamics of their practices. While these foci are shared across different approaches to visual observation, the behaviouralist approach also believes visual observation can capture ‘natural behaviour’.⁴⁴ This results in a quasi-naturalistic ontology in which, for example, much emphasis is placed on ‘face and body posture’ in order to identify ‘universal emotions’ that are posited to be indexically linked to physical/physiological features.⁴⁵ Interactions, equally, are often seen in terms of physical bodily movements and communicative patterns that have discrete meanings that are more or less easily generalisable across situations and ‘social contexts’. To understand these ‘natural’ patterns, positivist approaches reconstruct ‘a situation step-by-step’ to ‘analyze its inner dynamics’ by sequencing the actions depicted in videos sequentially to code interactions and draw out general patterns.⁴⁶ The goal is to draw causal links, similarly to process tracing. Such a method is widely deployed in psychology, sociology, and some applied sciences.⁴⁷ Its use is rare within visual IR (as understood herein), though it is increasingly deployed within positivist variants of political science.⁴⁸

The behaviouralist approach is important to consider largely as a representation of what we are not advocating for here. Indeed, the approach risks implying a naive empiricism where visual artefacts are seen as ‘objective’ data through which the polysemy of the social can be distilled. Indeed, behaviouralist approaches can be remarkably inattentive in their modes of observation – risking becoming neither ‘close’ nor ‘deep’ – given a privileging of large-N studies, an emerging desire to automate (through machine learning) the analysis of videos, and a silence on methodological questions surrounding the place of the observer in analysis, as seen in the notion that ‘universal emotions’ exist and a wider lack of reflexive engagement. At the same time, however, the long-standing critique of this naive view that ‘seeing is believing’ has risked – we think – obscuring the potential that different approaches to visual observation might have within IR. It is that potential we, thus, now turn towards.

Performative observation

A performative approach to visual observation opposes any notion of capturing ‘natural behaviour’. Instead, the approach sees visual data as a means of uncovering the ideationally structured nature of all sociality.⁴⁹ In this view, visual material is analysed to ‘gain a far more nuanced idea of how participants derive meanings’.⁵⁰ Visual observation can help ‘escape text- and talk-based approaches’ to understanding ‘the meaningfulness of social life’ by providing an additional but ontologically distinct avenue through which to interpret the meaningfulness of social action.⁵¹ Importantly, the search for meaning is not restricted to the deep structures that IR usually focuses on. Instead, these approaches – common in anthropology, geography, and sociology – also consider how meaning is constructed in an embodied, localised, and interactional sense: at the micro level. For them, the specificity of the visual as a medium is key for opening our sensibilities to a wider set of affective capacities, which cannot be captured textually or verbally.⁵² Nonetheless, to address questions about the positionality of the observer, performative approaches prefer to employ video materials in combination with other methods.⁵³ The goal is to incorporate ‘video into research methodologies, and to do so in conjunction with other corporeally invested research methodologies, and other, fleshier, recording “machines”’.⁵⁴ This includes the use of participatory photographic and documentary methods in which researchers and interlocutors co-produce video materials, upon which they collectively reflect back. A striking example is the documentary film The Act of Killing, where director Joshua Oppenheimer co-produced a surreal assemblage of video materials with victims and perpetrators of the Indonesian genocide.⁵⁵ Oppenheimer and his subjects re-enact practices of violence at their most minute level but also watch back those re-enactments, comment on them, and situate them in broader discourses. In a less dramatic example, Spinney describes using a ‘ride-along’ technique to explore how cyclists make sense of their activities:

‘By using techniques such as (video) freeze framing and slow motion alongside an interview format, the researcher can more easily elicit previously under-explored meanings of the practice of cycling relating to the sensory, vulnerability and technology.’⁵⁶

At first glance, the performative approach finds echoes in current work across visual IR, especially a growing focus on the value of film-making. Harman’s pathbreaking work in this area, for instance, describes how film-making can make visible the invisible and foreground the agency of the subaltern.⁵⁷ Film can be a ‘co-produced feminist method’ connecting the everyday and invisibilised to structures of power, in large part because it achieves this through ‘showing rather than explaining politics’.⁵⁸ Echoing this, Callahan suggests that films offer ‘an appreciation of the power of the nonlinear, nonlinguistic, and nonrepresentational aspects of knowledge’.⁵⁹

Such work is vitally important. However, IR’s affinities with a performative approach tend to remain enclosed in a focus on ‘deep’ political meaning. Harman, for instance, stresses that ‘film provides a new way of showing how power works to broad audiences’.⁶⁰ For his part, Callahan describes film as especially useful for exploring ‘the sensible politics of Self/Other relations’.⁶¹ In each case, it is emphasised how the visual achieves this through a focus on ‘the role of person-to-person relations, the importance of the everyday, and the value of emotions and embodied knowledge’, but the emphasis remains on fitting those micro-level schemas into wider ‘macro’ (or ‘deep’) categories.⁶² In this regard, deployments of performative approaches within IR tend to veer back to the safe terrain of the representational and the connotative in ways that, unlike other deployments of performative observation, risk reducing those local relations and embodied experiences to manifestations of higher-level structural factors. Equally, film-making in IR typically plays with what Van Munster and Sylvest term ‘arrangements of perceptibility’ that are ‘laying claim to or destabilizing truth or reality’.⁶³ In this, visual materials are produced to ‘show’ audiences something, rather than being considered objects of analysis of value for accessing the social world, as objects through which sociality and discourse can be (materially) observed. Instead, they tend to be subsumed within a broader and sometimes rather too ‘applicationist’ deployment of one or another variant of discourse analysis.⁶⁴

Ecological observation

A third understanding of visual observation exists: what we term the ecological approach, which can arguably be traced to the ethnomethodological tradition encapsulated in work like that of Elsey, Mair and Kolanoski.⁶⁵ Ecological approaches are in no way naturalistic, nor positivist, but some of their methodological precepts (such as the minute temporal and spatial sequencing of behaviours, practices, and events) may seem behaviouralist at first glance. This is because ecological approaches distinguish between the ‘literal’ (naive) observation found within behaviouralist approaches and what Garfinkel termed the documentary method of interpretation.⁶⁶ This involves deploying observation to search for an ‘an identical homologous pattern underlying a vast variety of totally different realizations of meaning’.⁶⁷ Put differently, if the performative approach focuses on the micro level to identify situated meanings that are often particular and non-generalisable, an ecological approach seeks to identify meaningful patterns across distinct situations. These meanings, however, are not naturalistic generalisable laws, but equally constructed and performed meanings. Most usually, ecological approaches associate such patterns with the idea that all social situations must be ‘worked towards’ at a micro level, and so certain practices of ‘local ordering’ must exist. As Garfinkel put it, if ‘the orderly character of everyday life is something that people must work to achieve, then one must also assume that they have some methods for doing so … members of society must have some shared methods that they use to mutually construct the meaningful orderliness of social situations’.⁶⁸ It is these ‘methods’ – these ethno-methods (‘human-methods’) or ‘interaction ritual chains’⁶⁹ – that the ecological approach hopes to access. In the classical ethnomethodological description, this refers to the idea that:

rather than deriving constitutively from a shared, underlying code, the order of practice is an ‘ongoing accomplishment’ by actors constantly monitoring the joint definition of the situation, updating it to accommodate resistances and inscribe them into its logic as constraints … Rather than pre-existing performances, the logic of practice … [is always] indeterminate.⁷⁰

The challenge here is to identify patterns that give order to the indeterminacy of life, something that self-evidently exists. In this, the ‘analytic mentality’ of the ecological approach is ‘radically different from asking the question What. It is the search for the How, for the modus operandi of the production, of the emergence, or the process of the formation of a gesture’, situation, or practice that it explores.⁷¹ If behaviouralist approaches believe visual observation can identify the ‘why’ of behaviour (e.g., the idea of universal emotions), and performative approaches seek to identify the ‘what’ of meaning seen in visual data, then the ecological approach tends to focus on ‘how’ social practices are achieved in situ. We term this the ecological approach because it focuses on the ‘full’ components of social situations: the full ecology of relations between particular human actors, material tools, surrounding technologies, and other factors which may or may not be implicated in how social events occur. The importance of ‘sequencing’ these behaviours (as seen in the example of ‘Collateral Murder’) and observing them through video lies in this understanding that ‘performances’ are not solely dictated by broader social scripts that are necessarily ‘invisible’ in particular situations, whether these are articulated as ‘natural laws’ or instead as structuring social discourses constructed over history, nor entirely uniquely situated meanings that require a focus on ‘strong objectivity’ to tentatively uncover performatively.⁷² Instead, the ecological approach gestures towards a recursive relationship in which each situated performance of an action (e.g., firing on civilians from a helicopter gunship) is unique and indeterminate, but also connected to previously experienced situations, in ways that require we consider both immanent situations and their structural context holistically (i.e., ecologically).

A new balance

At present, empiricist variants of political science are increasingly deploying variants of the behaviouralist approach. For its part, critically inclined visual IR has made some early steps towards integrating performative modes of observation into analysis, but otherwise remains reticent. This is a missed opportunity. Performative approaches to visual observation outside IR are interesting for combining a focus on the discursive, ideational, invisible, etc., with the close description of the everyday practices depicted in video material that the ecological approach tends to privilege. At the same time, we would encourage a new balance in which in particular the value of ecological observation is considered. This would allow a bridge between the ‘deep’ reading of visual artefacts and the ‘close’ or ‘attentive’ observation of their contents. Indeed, the ecological approach is especially promising for allowing the development of what Collins calls a ‘sociology of the non-obvious’.⁷³ In its most attentive forms, visual observation requires that the observer develop what Bourdieu termed a ‘rupture with the presuppositions of lay and scholarly common sense’⁷⁴ and/or engage with Foucault’s dictum that we ‘must … pretend not to know’ what we are observing by engaging in acts of ‘seeing’ rather than ‘recognizing’.⁷⁵ Despite raising many questions about the position of the observer, it is this attempt to withdraw as far as is possible from presuming the meanings inherent in actions depicted in a video that we think is especially valuable. Cultivating this attentive sensibility allows the possibility of fracturing our understanding of what constitutes social action, potentially nudging us to ‘think differently’ about world political dynamics tout court. In doing so, we hope to show in the next sections how novel – and potentially critical – insights about international relations might be conjured up.

Focal points for visual data observation in IR

How can visual data observation complement existing work within visual IR? As our discussion hopefully clarifies, our contention is that visual observation is of especial value for understanding micro-level political events and their connections to the global. Our claim is thus not that visual observation is of universal value to IR tout court. Instead, we now lay out three focal points where visual observation is especially useful: practice, materiality, and discourse. Importantly, these focal points are intertwined in any real-world analysis. Their separation here is thus carried out only as a heuristic, particularly because each is often associated with different subfields of IR that explore the kinds of micro-level phenomena that visual observation allows us to see in action. This includes practice theoretical approaches, variants of international political sociology, and science and technology studies-informed IR. Our third focus on discourse, however, is also taken up to show how observation of this kind can ‘link back up’ with a focus on deep meaning. In short, the re-balancing between deep reading and close observation that we call for here is designed to further the agenda of visual IR (and beyond) rather than displace existing approaches.

Practice

IR is increasingly concerned with what actors ‘do’, shifting to a micro-level appreciation of the everyday, the lived, and the practical.⁷⁶ Approaches include Bourdieusian or pragmatist variants of practice theory, deployments of ethnography, turns towards the everyday and vernacular, and the long-standing work of feminist theory.⁷⁷ Each approach is concerned with reconstructing the granular everyday of world politics in ways that go beyond macro-political explanations. This is achieved through a focus on the practical content of world political action, the ways those practices are often unreflexive, and the ways they entrench or loosen global power structures. But accessing practice is not easy: ‘how does one “capture” an international practice?’⁷⁸ For example, practice theorists note that their most common methods, interviewing and document analysis, stand only as ‘proxies to direct observation’.⁷⁹ Interviews typically result in the ‘verbalization of reflexive knowledge’ rather than the ‘background dispositions’ of practical experience. The phenomena of actual interest to practice theory ‘must [thus] be read between the lines and distilled from the analysis of practices’.⁸⁰ While participant observation is seen as a means of overcoming these issues, many phenomena of interest to IR are impossible or very difficult to access in this way. Even where they can be accessed ethnographically, no observer can ‘simply be dropped on site to neatly collect and meaningfully report on its practices’.⁸¹ Instead, a multiplicity of methods are always required.

How might visual data observation assist here? First, visual observation can be seen as a ‘second-best’ form of ethnography, given its capacity to capture everyday dynamics.⁸² Now, watching a video of an event cannot compare to being there in person. Instead, videos must be analysed as ‘documents which relate to a bodily material practice which is outside of the text itself but to which the text is related’.⁸³ Despite those interpretive limits, videos provide more than texts: a deeper, more granular, and ‘lived’ set of insights into depicted environments and the chance to experience a fuller affective and atmospheric sense of situations. Images ‘do’ something distinct ‘in the sense of an active notion of what visceral affect they can provoke’.⁸⁴ Equally, and reflecting the use of video recording in ethnography, visual artefacts are valuable as they can be replayed an infinite number of times to detect the intricate ‘micro-dynamics’ of situations. Likewise, visual artefacts allow the observation of tones of voice, music, bodily postures, and facial expressions, as well the rhythm of an interaction (corporeal or conversational).⁸⁵ This is especially true vis-à-vis accessing the ecological approach’s concern with the ‘how’ of social practices, situations, and events.

Indeed, because of their benefits, ecological variants of visual observation have been in other disciplines for some time. For example, ethnomethodologists use video data to study practices such as crossing the street or queuing in lines, while micro-sociologists study CCTV videos of street fights and riots.⁸⁶ The study of practice in IR might fruitfully follow these examples, given the rising availability of video data concerning many different phenomena of relevance to world politics. For example, within peace and conflict studies, Bramsen has analysed video recordings from the Northern Ireland Assembly to study agonistic dialogue.⁸⁷ Figure 1 thus depicts a practice of ‘self-irony’ where Member of Northern Ireland’s Assembly (MLA) Jim Allister has just exclaimed ‘I’ll spare you that’ in response to being offered the chance to speak. To contextualise that micro-practice, Bramsen counts the number of times that members of the assembly laugh throughout the session (seven times), which she holds up against the number of times that laughing is reported in the transcripts of the 188 meetings held from 1998 to 2002 (six times), illustrating the subversive nature of this practice. Studying micro-practices such as these is something visual observation makes possible and which may help expand our understanding of the diversity of the social processes (the presence of humour, agonistic relations, etc. in the case cited above) that shape institutions or events.

Figure 1. Ministers laugh after an intervention in the Northern Ireland assembly.

Beyond furthering empirical inquiry, visual data observation may also assist in expanding the conceptual coordinates of practice theorising. At present, most accounts minimally assume practices emerge based on a form of ‘rule-following’, allowing us to explain situated practices as more or less ‘(in)competent performances’.⁸⁸ References to habits, discourse, background knowledge, or scripts often risk conceptually formalising practices as objects that exist outside the moment of their enaction. It is perhaps for this reason that – paradoxically – most accounts of practice in IR are empirically light vis-à-vis the actual situational enaction of practices. For example, in their account of international intervention into Libya, Adler-Nissen and Pouliot’s empirical discussion is limited to the observation of a series of events (British drafting of a resolution, Lebanese mediation, etc.) and the temporal emergence of practices (framing, drafting, etc.) without unpacking the situational dynamics through which those individual practices were enacted.⁸⁹ As Walter writes, accounts like these are ‘forms of meta-commentary that invoke theoretical referents but cannot specify them in empirically substantive terms’ because they are rarely anchored ‘in the empirical texture of the [practical] episodes themselves’.⁹⁰ To return to our earlier discussion, even practice-theoretical accounts often assume that ‘context’ is relatively fixed at a micro level, leading to a habitual use of pre-existing scripts, rather than focusing on the ubiquity of local problem-solving. This may be one reason why much practice theory has coalesced around the study of diplomacy or cognate practices, where the habitus of practitioners is assumed to be relatively fixed given the institutional thickness of diplomacy as a field.

Following this, engaging visual data observation might allow IR to draw more closely on the insights of a microsociological focus on interaction and as such avoid the risk of practices becoming seen as exogenously imposed (by training, historical repetition, etc.) rules. By turning towards a micro-sociological focus on how practices are constituted through multiple situational interactions, practice theorising would gain a more contingent, fragile, and complex understanding of how practices emerge through contextually specific, embodied, and emotional processes. Consider, for example, the work of Randall Collins on (political) violence.⁹¹ Discussing instances of the massacres of captured soldiers during wartime, Collins describes in granular detail a phenomenon he terms a ‘forward panic’, which refers to the ways ‘tension and fear’ in conflict situations create an ‘onrushing flow of events in time’ marked by affects and emotions that builds a ‘tunnel of violent attack’ often resulting in atrocities.⁹² His ability to trace out this phenomenon rests on an extensive use of ecological visual data observation to infer the emotional state of participants, the ways they interact physically and verbally, and the overall rhythm of the situation concerned. The importance of micro-sociological insights like those of Collins is not that they deny that other factors can cause the emergence of these practices, but that they expand our repertoire of explanations for how practices become in situ and how those situational constraints might in turn provoke unexpected practices. The result might be an empirical and conceptual expansion of the coordinates of practice theorising.

Materiality

Across IR, there is now sustained interest in the relevance of material, technological, and other non-human objects. This is true both theoretically, where scholars employ ecological, post-humanist, new materialist, pragmatist, and cognate toolkits to map out the non-ideational components of international affairs, and empirically, where IR now examines the ways novel technologies (drones, artificial intelligence), architectural structures, everyday objects, and more, alter the behaviour of individual human beings and/or operate autonomously beyond the conscious control of human beings. While there are many conceptual variations in the approach taken, Pickering provides an especially evocative description of the underlying sentiment:

The world … is continually doing things, things that bear upon us not as observation statements upon disembodied intellects but as forces upon material beings. Think of the weather. Winds, storms, droughts, floods, heat and cold – all of these engage with our bodies as well as our minds, often in life-threatening ways … Much of everyday life, I would say, has this character of coping with material agency, agency that comes at us from outside the human realm and that cannot be reduced to anything within that realm.⁹³

There remain distinct problems in analysing the place of materiality in world politics, however, and in particular with unpacking the co-constitutive manner in which humans are ‘coping with material agency’ on a daily basis. Two trends seem to reflect these difficulties in particular. First, there is a tendency to focus on the most strikingly non-human and alarming technological forces, as seen in the literatures exploring autonomous weapons systems, climate change, or algorithmic governance. This often results in overly dramatic accounts of material and technological agency as objects that are evolving intrinsically separately from human life as opposed to being embedded like ‘winds, storms, [and] droughts’ in more banal interactions with humans. Second, where IR does focus on less exceptionalist material processes, it tends to do so through the lens of ‘infrastructures’, ‘assemblages’, or ‘networks’ as ordering devices that allow us to understand the importance of the non-human in structuring world politics. While useful frameworks, these concepts often abstract the place of material forces beyond the local, everyday, and micro in ways that make it difficult to analyse the quotidian interactional dynamics between human beings and material forces. Again, they are too frequently ‘deep’ readings of the place of materiality in politics, as opposed to close observations of how materiality impacts upon sociality in real-world scenarios.

Visual data observation is helpful here because it allows us to capture not only human interaction but also human–material interactions at a granular level, uncovering processes of co-constitution between the human and the non-human. Put differently, visual observation can help us analyse ‘the very ecology’ of activities ‘which includes the spatial arrangement of the participants, their embodied action, and their use of various material resources such as objects, documents, and technologies’.⁹⁴ Indeed, most of the theoretical insights that ground a broader social scientific shift towards considering the importance of materiality were generated through visual observation. This is best seen in the decades-old work of anthropologists and sociologists in studying the banal interactions between humans and machines.⁹⁵ In her classic Human–Machine Reconfigurations, for instance, Suchman describes how the basis for her theoretical work in this field found its genesis in her videotaped analysis of users of a then-novel technologically advanced photocopying machine, laying out how those videos allowed her to ‘argue that the machine’s complexity was tied less to its esoteric technical characteristics than to mundane difficulties of interpretation characteristic of any unfamiliar artifact’ and so that ‘however improved the machine interface or instruction set might be, this would never eliminate the need for active sense-making on the part of prospective users’.⁹⁶

Work like this is at the root of a majority of the theoretical toolkits deployed by IR to explore the materiality of world politics. But it is rare for IR to use similarly granular modes of analysis.⁹⁷ For us, greater use of visual observation would enable IR to de-dramatise its accounts of materiality by refocusing on the ways humans and non-humans co-produce reality in quotidian, banal, and everyday ways. Specifically, two approaches might be particularly useful: sequential and comparative approaches to considering the place of materiality. Sequential approaches (viz. Elsey, Mair and Kolanoski’s account of ‘Collateral Murder’) would focus on the ways in which material objects do or do not combine their technological affordances with human bodies and settings in order to produce particular outcomes. For example, Bramsen drew on the sequencing of video data (alongside interviews) to show that the presence of material objects, ranging from weapons to everyday things, not only enabled but also directed the form that violence took during the 2011 Arab uprisings.⁹⁸ Figure 2, for example, was analysed to show how what seem banal material objects – a garbage can and its contents – to encourage particular forms of violence in Bahrain, over others. Her findings there echo that of micro-sociologists in the study of domestic interpersonal violence.⁹⁹

Figure 2. Bahraini protestors throw garbage at riot police during 2011 protests. Along with emotional motivations driving the situation, the material object of the garbage can can be seen to shape and direct their violent acts.

By contrast, comparative approaches focus on increasing the scope of claims made through sequential approaches vis-à-vis the salience of particular material objects by tracking the different effects that those objects have across a wider set of cases. For example, in his micro-sociological studies of the emergence of torture, Austin lays out the importance of particular material objects for the form that torture takes by comparing instances where particular objects are present or absent.¹⁰⁰ In one discussion, the presence or absence of a Kalashnikov assault rifle is shown – through an analysis of several hundred videos of torture in the Middle East – to significantly affect the kinds of violence inflicted on victims (even within what is nominally the same type of torture). The absence of the rifle leads to more ‘extreme’ or ‘frenzied’ forms of torture, with greater risk of death for the victim, whereas the presence of the rifle and its material affordances tends to allow for a more ‘controlled’ form of torture in which perpetrators maintain their composure. On that basis, Austin theorises the rifle as ‘combining’ with human participants to provide a kind of structure for their activities, a structure that falls apart (or must be otherwise assembled) in its absence.¹⁰¹ This theory is derived directly on the kind of attentive visual observation described here.

These approaches, or others, can deepen our conceptual understanding of materiality and its relation to world politics. Specifically, IR would gain a better grasp of how world politics is structured in part through what Ingold refers to as the ‘taskscape’ of life.¹⁰² As he writes, ‘just as the landscape is an array of related features, so … the taskscape is an array of related activities. And as with the landscape, it is qualitative and heterogeneous … the taskscape is to labour what the landscape is to land, and indeed what an ensemble of use values is to value in general’.¹⁰³ By viewing international practices ‘in action’ (through visual observation), the non-human is no longer a ‘separate’ thing – part of the landscape or technological sublime – but part of a heterogeneously negotiated sphere of sociality. In this view, for example, it would be possible to move beyond the idea that technologies contain particular affordances that are ‘inherent action potentials of a given technology’, which users may or may not ‘realize’.¹⁰⁴ This view maintains a separation between the human and the material rather than seeing how both humans and non-humans constantly reconfigure one another and that hence there are few ‘inherent’ (i.e., essential) potentialities in either category. Just as in the case of practice theorising, then, visual observation would introduce a more contingent and complex understanding of materiality/technology to the study of world politics.

Discourse

Can visual data observation transcend the micro? The current focus of visual IR on the ‘deep’ reading of visual artefacts relates to a fear that empiricist observation weakens the field’s commitment to comprehending global or structural facets of discourse, power, meaning, identity, and so forth. Thus, even if visual observation is not necessarily naively empiricist nor positivist, concerns remain. It is therefore important to stress that visual observation can help ground our understanding of how situated processes are recursively connected to macro-level discourses that extend across time and space. To see this, consider first the notion of norms. As Onuf writes, discussions of how ideas structure world politics are often empirically ungrounded:

Most constructivists hold norms to be formless (and their existence not contingent on their articulation), as such exhibiting the shape-shifting properties of fluids and gases. They are ‘in the air’; like ideas and expectations, they flow and float, if not always freely … they get into people’s minds through a mysterious process called internalization and then manifest themselves as proper or appropriate behaviour through another mysterious process called socialization.¹⁰⁵

One of the difficulties faced across the study of norms, discourse, meaning, or identity in IR is that explanations for the emergence, solidification, or rejection of these phenomena are typically made post hoc. Discourses are rarely studied (ethnographically, ethnomethodologically, or simply sociologically) as emergent and achieved things. One consequence of this is that human beings tend to be presented as either (1) relatively passive recipients of, and thereafter vectors for, ‘formless’ discourses; or (2) actively agential ‘entrepreneurs’ for particular discourses, given their possession of forms of power. Visual data observation has the potential to nuance this status quo. Take an example. In her study of soldiers who share ‘illicit images’ of violent abuse, Megan MacKenzie describes this process as emerging from ‘an established element of military culture’ and that the images ‘are central to the production of the band of brother internal military culture’ driven by discourses of ‘abuse, sexism, racism, and homophobia’.¹⁰⁶ Her argument is surely correct, but it also renders soldiers passive bodies through which these discourses flow, without grounding the reasons for the dominance of those discourses beyond their historical sedimentation. Likewise, responsibility is generally shifted to military or civilian officials who are said to be the main holders of the requisite agentic power to shift such an ‘internal military culture’ and hence those who must principally be targeted by counter-normative entrepreneurs seeking to shift the internalised nature of these norms.

By contrast, Austin and Leander draw on a performative/ecological understanding of visual observation to tease out the ways in which ‘illicit images’ can be read not only as passive inscriptions of meta-discourses (of military culture, etc.) but also as tools through which to understand how those meta-discourses are enacted and stabilised locally.¹⁰⁷ In the example they give, images of detainee abuse by both the USA and the Syrian Arab Republic are compared (Figure 3). The images, Austin and Leander begin by noting, depict the same practices: forms of the ‘extraordinary rendition’ of prisoners but are nonetheless most commonly discursively ‘read’ quite differently. In the case of images of Syrian abuse, the soldiers depicted carrying out violence are seen as active war criminals, whereas in the US case, they are considered to be ‘following orders’ (imposed from above). Their paper seeks to understand how the micro-practical and material achievement of the same task in each case is integral to these wider macro-level discourses. To do so, they study images (in the case of the USA) and videos (in the case of Syria) of the practices in question, teasing out how each ‘situation’ was achieved, the material limits dictating what was possible to the violence workers in each case, etc. Ultimately, they suggest that it was only possible to depict the cases differently due to material ‘inequalities’ in the respective environments in which the practitioners were operating. The USA’s comparative material advantage (advanced tools, weapons, planes, etc.) allowed them to ‘cleanse’ their practices in ways that – to some degree – were not possible for the Syrian soldiers involved. This, then, radically alters the aesthetic and affective interpretation of the practices publicly, feeding into wider discourses of a more or less ‘civilised’ set of practices carried out by the US and ‘barbarian’ practices carried out by Syria.¹⁰⁸

Figure 3. Photograph depicting the ‘rendition’ of prisoners to Guantánamo Bay by the United States, analysed by Austin and Leander, ‘Visibility’.

This is just one example of how ecological forms of visual data observation can reveal how wider discourses are situated within the complex contingencies of local events, captured on camera, rather than being ‘formless’ constructs imposed ex nihilo. But it helpfully demonstrates the value of the ‘re-balancing’ between deep and attentive modes of analysis that we are advocating for here. Austin and Leander’s discussion becomes possible through multilevel modes of observing visual images ecologically. In doing so, it shows how visual observation allows us to move ‘beyond the textual’ towards appreciating the ‘unsayable’ aspects of world politics or ‘that which cannot be expressed with words’.¹⁰⁹ It allows us to see sociality differently, creating a rupture with both lay and scholarly common sense. But understanding the politics underlying these surprising revelations from acts of visual observation then requires the tools of deep reading more common within visual IR at present. Each approach thus works with the other recursively in ways that allow us indeed to see the world differently.

From seeing-as-believing, through seeing-as-rupturing, towards seeing-as-critiquing

Visual data observation has the potential to work in complementarity with the existing pluralist repertoire of methods deployed across IR. Its especial value lies in the ways its cultivation of an ethic of attentiveness allows for forms of ‘close’ description of life, politics, and sociality that can complement the current preoccupation of visual IR on deep reading. All that being said, dilemmas remain. Some of these are pragmatic. Visual data is surrounded by secrecy and confidentiality concerns. Equally, the veracity of visual data can be problematical. In addition, the observation of visual data within IR in particular can raise substantial ethical concerns where that material depicts vulnerable persons, sensitive political events, or other such phenomena. These pragmatic concerns are serious. However, they are – generally speaking – dilemmas that are found across the spectrum of qualitative methods, and which can/should be addressed with the help of already existing toolkits. Beyond these issues, there exist nonetheless a series of far more complex dilemmas that surround the question of visual data observation.

Our critique of the behaviouralist approach to observation rests on its naive adherence to the view that seeing is believing. But neither ecological nor performative approaches are immune to the same error, and so maintaining reflexivity is critical. This is doubly true because of the specificity of the visual as a medium. The visual is visceral and affective, with the lives and loves of images misleading even the most attentive of observers.¹¹⁰ Equally, images are always attached to assumed contexts and iconologies, as well as being deliberately edited and curated to inject specific meanings, as Saugmann’s aforementioned analysis of ‘Collateral Murder’ makes clear. Addressing this dilemma requires care. Nonetheless, it also increases the value of our call for a new balance between deep and close or attentive modes of analysis. The ‘deep’ reading of visuals, and abstraction of their content through meta-discursive constructs or other means, is equally often based what Stoler calls our ‘epistemic habits’.¹¹¹ These are ‘ways of knowing [and seeing] that are available and “easy to think”’ … and which produce ‘permanent momentary items of [implicit] fact’. Indeed, despite its vitality, visual IR is often too applicationist, reducing the visual to the discursive, and erasing the materiality of discourse, through a naive deployment of discourse analysis.

In our view, the performative and ecological modes of visual data observation we have advocated for have ways of helping balance out some of these risks. Performative approaches rarely rely on observation alone but deploy these tools simultaneously in dialogue with those depicted in visuals (see the examples above) to attempt to glean a ‘strong objectivity’ that transcends the researchers’ own positionality and grounds discourse in bodies, affects, and atmospheres. For their part, ecological approaches rely on variants of Bourdieu’s notion of participant objectivation that seeks to break ‘with the unthought presuppositions of thinking thought … to rid themselves of their inbred scholastic bias’.¹¹² The goal is to generate a ‘rupture’ within our epistemic habits.¹¹³ While this is a general desire in much social science, ecological approaches have a particular potential to help achieve Foucault’s analogous dictum that we ‘must … pretend not to know’.¹¹⁴ In our opening example of Elsey, Mair and Kolanoski exploring ‘Collateral Murder’, for instance, the authors rely on the time-consuming and tedious task of minutely sequencing the actions depicted in the video. This task of transcribing and turning into ‘technical’ written language visual material can generate a distance or alienation from its affective content. To adapt Deleuze, the visual becomes ‘a foreigner in its own language’.¹¹⁵ Such distance, generated through attentive description, can direct us towards unseen, unnoticed, or invisibilised, and so analytically surprising, aspects of the world of relevance for all social science.

But again this requires an alliance between ‘deep analysis’ and ‘close observation’ of the visual. If all we engaged with was close observation of the kind advocated for here, the risk would be reducing visual IR to the ‘descriptive recapitulation’ of the artefacts it engages with that would be ‘devoid of significance’ for world politics.¹¹⁶ Nonetheless, even if such an alliance emerged, a final dilemma would remain. Could such an alliance be ‘critical’ in its analysis? Visual IR has also been reluctant to engage with visual observation for political and normative reasons. The preference for ‘deep reading’ also reflects a desire to intervene directly or indirectly in politics. For many, the idea of visual data observation is antithetical to such a goal. Indeed, even those outside IR who deploy visual observation are concerned this work might be ‘underestimating the question of the social embeddedness of’ the phenomena it explores.¹¹⁷ Simply put, then: can close and attentive practices of observation, in alliance with other forms of social scientific analysis, contribute to a politically meaningful mode of social science?

Speculatively, we wish to answer in the affirmative.

Consider Forensic Architecture (FA). Within IR, FA’s work monitoring the machinations of state agencies is seen as seeking to ‘make visible violence, war and environmental problems with the goal of providing evidence’ by engaging in a ‘performativity of picture-taking’.¹¹⁸ In this view, FA’s success is predicated on its capacity to capture state violence and inject those images into the world performatively. But FA’s work has more than one element. Most of its investigations begin with the collation of visual material (videos, satellite images), which are then observed by specialists who use this data to ascertain ‘what happened’ in a situation. But the success of its advocacy work has not rested on the findings or ‘truth claims’ of such highly technical acts of observation. Rather, FA shifts from this observational mode to critical aesthetic forms of engagement: translating observations into evocative aesthetic objects.¹¹⁹ Figure 4, for example, shows a reconstruction of a bombardment during the 2014 Israel–Gaza war created by FA. This is a synthetic – artistic and aesthetic – object derived from its earlier acts of observation that was combined with ‘deep’ contextual understandings and interpretations of the conflict and then exhibited – both literally and metaphorically – to a broader public to advocate for accountability. It is just one especially evocative example, we think, of the critical potential that lies in allying deep and attentive modes of analysing visuals.

Figure 4. A Forensic Architecture reconstruction of the 2014 bombardment of Gaza by Israel.

We see finally, then, how visual data observation can be allied with critical aesthetic practices of normative-political intervention and creative visual engagement. This is a mix of observation and empiricism, interpretation and criticality, objectivity and politicality. The challenge, as we see it, is thus not to ask if visual data observation can retain critical potentiality but instead to ask how this work of close attentiveness to the world can be entangled with existing methodologies, practices, and conceptual tools to augment the critical capacity of visual IR as a whole. Opening such a conversation will be complex. Indeed, this article combines the insights of researchers working at the intersections of different subfields of political science and adjacent fields with distinct epistemological and ontological commitments. What unites us, as authors, is an interest in how the close and attentive observation of visual material provides unexpected – surprising, even disturbing – disruptions of our theoretical, conceptual, and socio-political expectations. These surprises only really become clear when one indeed slows down and opens up to the polysemic messages that observing visual data generates. But even upon this shared terrain, epistemological differences remain. To some degree, thus, introducing visual observation of these kinds into visual IR requires a recognition that epistemological, ontological, and cognate frictions are less important, on one level, than collaboratively seeking to cultivate the full promise that visual artefacts hold for interrogating world politics.