Computational ethnography: A view from sociology

Abstract

This commentary elaborates on the ideas and projects outlined in this special issue, from a specifically sociological perspective. Much recent work in sociology proposes ‘methods mashups’ of ethnography and digital data/computational tools in different and diverse ways. However, typically, these have taken the form of applying (with or without tweaks) the principles of ethnography to new domains and data types, as if ethnography itself is stable and immutable; that it has a universal set of methodological principles that unify ethnographic practice. Returning to anthropology (whence, arguably, ethnography originally came) is, therefore, a useful way to extend our methodological thinking to (re)consider what ethnography is and how it operates, and from there think more clearly about how it may be effectively combined with digital data/computational tools in an emerging ‘Computational Anthropology’.

This article is a part of special theme on Machine Anthropology. To see a full list of all articles in this special theme, please click here: https://journals.sagepub.com/page/bds/collections/machineanthropology

Emphasising the importance of internet-mediated communication in how we (social scientists) understand social life, Housley et al., note ‘these technologies and their allied data have the potential to “digitally-remaster” classic questions about social organization, social change and the derivation of identity from collective life’ (2014: 4). This research commentary extends the idea of ‘digital-remastering’ to the methods of social science, casting new light on long-standing issues. One such issue, around which sociology (as my disciplinary background) and anthropology (as the disciplinary context of this special issue) converge, is an inherent tension between (a) the context-specific and inherently-amorphous nature of (digital) ethnographic studies, and (b) the need to provide instruction as to how (digital) ethnographic research is to be done in relatively prescriptive terms. ¹ Foregrounding this long-standing tension around ethnography as a research practice provides a moment of pause in a seemingly-ever-emerging array of fields (social media analytics, data science, machine learning, natural language processing, corpus linguistics etc.) which necessarily move fast to meet the demands of a constantly-shifting digital social environment (Sujon, 2021). This is an attempt to ensure our methods are on a fruitful path by first taking stock of where we are and where we have been, rather than operating as though these matters are more or less settled already. Though in the face of the excitement around the ever-expanding possibilities of digital research this may come across as dreary Luddism, I propose that such an approach leans into (rather than shies away from) one of the core values of both anthropology and sociology; the placement of method front and centre, in this case by virtue of seeing the digital as a reinvigoration of the long-standing issues of our respective fields rather than a development beyond them.

The commentary begins with an account of how sociology presents ethnographic methods mindful of this tension, then traces the lineage of that tension back through its anthropological roots. From here we cast forward into ‘digital ethnography’, ² as a methodological domain of relevance to the broad theme of this special issue, to explore how this is (perhaps counter-intuitively, given the characterisation of digital methods as being transformative and innovative) carried forward largely unchanged. The upshot is that if this tension is to be apologised for (and many accounts presented herein do precisely that) then rethinking rather than reproducing the formats that necessitate this tension seems a long-overdue intervention, especially when we are in the midst of designing methods of enquiry for Computational Ethnography and have a relatively open field ahead of us in which to do so.

Ethnography in sociology

Ethnography is a vital component of sociology's methodological repertoire, with a long tradition of written materials to define it and guide practitioners in its conduct. A common theme across such materials (Atkinson et al., 2001; Brewer, 2000; Hammersley and Atkinson, 2007; O’Reilly, 2005) is that the cogency of the method is assured by an attentiveness to the impossibility of providing prescriptive guidance on what ethnography is and how to do it. Nevertheless, such accounts (paradoxically) get their pedagogical value through a suspension-of-disbelief around this impossibility, presenting the method via reference to universal essential qualities. ³ We might note this tension in Atkinson et al.'s reluctance to pin down ethnographic research to particular (sets of) practices: ‘we have explicitly avoided any typology or developmental schema for ethnography which assumes a linear model of progress, or tries to erect “pure” categories … [such models] can serve useful pedagogical functions, but can ultimately do violence to the complexities of research and its historical development’ (2001: 1-2). Yet immediately thereafter, Atkinson et al., note that ‘There are, of course, broad family resemblances between the various methods and applications that have characterized ethnographic research over the years’ (2001: 2); chiefly, ‘a commitment to the first-hand experience and exploration of a particular social or cultural setting on the basis of (though not exclusively by) participant observation’ (2001: 4). Hence, what matters is not so much the activities of field research, but the preservation and representation of a context of lived experience by whatever means necessary: ‘It is only by being in context, being there to talk with and listen to the people you are researching as they experience things and as they go about their daily lives, that you can get them to tell you about how they feel and think’ (O’Reilly, 2005: 10). Yet as Holstein and Gubrium (2007) elaborate, representing context as a static backdrop against which lived experiences play out is a necessary evil in written accounts of research. Hence, as Hammersley and Atkinson acknowledge, ‘ethnographic research is guided by a distinctive analytic orientation; yet one which contains some important, unresolved, and perhaps unresolvable, tensions’ (2007: 236); not least that it is impossible to pin down what ethnography's distinctive analytic orientation actually is. Accordingly, the notion of writing texts on ethnographic methods seems to be, by the admission of the authors themselves, an unavoidably unavailing enterprise – one of writing formalised accounts with the right hand, whilst simultaneously erasing them with the reflexive left.

Ethnography in anthropology

Tracing back further, to longer-standing debates within anthropology whence ethnography has its origins, we can explore more specific concerns raised by methodologists of anthropology (which have subsequently filtered through to sociology and ‘new’ forms of enquiry such as ‘digital ethnography’). Particularly, the interpretation of culture as the material that ethnographers are to write about, participant observation as a method of doing ethnographic work, and writing as a form of representing that work.

Being no anthropologist myself, and wilfully taking the role of ‘ignorant newcomer’ ⁴ as a ludic Socratic-cum-ethnomethodological intervention designed to probe issues that have often been glossed over in favour of breathlessly greeting the arrival of digital methods (Brooker et al., 2019), it is apposite to focus some attention on texts which, though gauche in their field, form its roots. ⁵ With this in mind, Geertz arguably has the definitive statement on the role of ethnography within anthropology:

In anthropology … what the practitioners do is ethnography. And it is in understanding what ethnography is, or more exactly what doing ethnography is, that a start can be made toward grasping what anthropological analysis amounts to as a form of knowledge. This, it must immediately be said, is not a matter of methods. From one point of view, that of the textbook, doing ethnography is establishing rapport, selecting informants, transcribing texts, taking genealogies, mapping fields, keeping a diary, and so on. But it is not these things, techniques and received procedures, that define the enterprise. What defines it is the kind of intellectual effort it is: an elaborate venture in, to borrow a notion from Gilbert Ryle, ‘thick description’. (Geertz, 2000: 5-6).

To Geertz, thick description is an account of a setting comprising description-with-context. However, despite Geertz’ disavowal of prescriptive methods for doing the work, he nonetheless still attempts a taxonomy of how ‘thick description’ should be produced: ‘there are three characteristics of ethnographic description: it is interpretive: what it is interpretive of is the flow of social discourse; and the interpreting involved consists in trying to rescue the “said” of such discourse from its perishing occasions and fix it in perusable terms … But there is, in addition, a fourth characteristic of such description, at least as I practice it: it is microscopic [i.e. it focuses on the everyday]’ (2000: 20-21).

Having the aim and focus of the field established thusly gives room for others to then dig into the practical activities of producing anthropology's ‘thick descriptions’. These are often outlined in reference to component methods such as participant observation – a ‘central and defining method of ethnographic research’ (Musante, 2005: 251) if not the totality of it. Here, again, we see a tension between the impossibility of defining best practice (since context-specificity must be preserved) and an attempt to define it nonetheless. Perhaps this is represented at its starkest in Bernard's account, which begins with a (necessarily) ambiguous statement that ‘participant observation is not an attitude or an epistemological commitment or a way of life. It's a craft’ (Bernard, 2011: 258), followed by setting out a series of what appear to be highly specified rules for its artisanal practice:

Fieldwork comes in three flavours: ‘(1) complete participant, (2) participant observer, and (3) complete observer’ (2011: 260). It ‘traditionally takes a year or more because it takes that long to get a feel for the full round of people's lives’ (2011: 261), and the fieldwork process can be boiled down into stages with distinctive boundaries (‘1. Initial Contact’ (2011: 284), ‘2. Culture Shock’ (2011: 285), ‘3. Discovering the Obvious’ (2011: 286), ‘4. The Break’ (2011: 287), ‘5. Focusing’ (2011: 287), ‘6. Exhaustion, the Second Break, and Frantic Activity’ (2011: 287) and ‘7. Leaving the Field’. (2011: 288))

Clifford argues that a focus on writing, as the core means of representing such empirical endeavour as outlined above, is also not exempt from tensions of this kind:

A ‘focus on text making and rhetoric serves to highlight the constructed, artificial nature of cultural accounts. It undermines overly transparent modes of authority, and it draws attention to the historical predicament of ethnography, the fact that it is always caught up in the invention, not the representation, of cultures’. (1986: 2)

Hence, rather than providing an objective medium through which insights can make it from the world to the page, ethnographic writing is a constructive activity – not disconnected from a ‘real’ world, but also constitutive of it. Though the social construction of things of all kinds is mundane to sociology and anthropology alike (Hacking, 1999), the relevance of this (mundane) point in the present argument is that it pauses us on the inherency of this tension in the conduct of ethnography – if our objectives (to capture/produce ‘thick descriptions’), our practical activities (participant observation) and our representational forms (writing) preclude the possibility of definition by virtue of their being ethnographic, nonetheless, definition seems to be an inevitable aspect of what ends up being produced.

So what? The implications for digital/computational ethnography

Thus far, the notion of a specifically digital ethnography has been largely absent. However, the argument can now point to where the inherent tension of writing about ethnography and ethnographic methods has been reproduced in latter-day ‘digital’ innovations, with the effect of obscuring the role of methods in (digital) social research by failing to recognise an opportunity to clarify such matters. There is a diversity of practice in digital ethnography; the result of exponential complexities arising from the settings in/on which it is applied (Pink et al., 2015). In this sense, it not within the scope of the present paper to resolve such questions as to whether or not an offline/online distinction is still relevant in this area (though see Brooker (2014) for a take on this). Rather, the aim is to review, as a Wittgensteinian project of exploring the multiplicity of referential terms, a non-exhaustive range of well-trodden approaches to applying ‘ethnography’ digitally to explore the variety on offer. This is done with an eye for where these practices draw through longer-standing tensions between ethnography's nomological slipperiness and the pedagogically-driven need to reify it nonetheless. The value here is in seeing the diversity of approaches in one place as opposed to considering them in the isolation that the conventions of methodological publications often entails.

Old methods, new topics

An initial brace of early adopters of digital ethnography (e.g. Hine, 2000, 2006; Kozinets, 2010; Markham, 2004; Murthy, 2008) characterised the practice as a continuation of old methods in new domains. As Murthy notes, ‘As ethnography goes digital, its epistemological remit remains much the same. [Digital] Ethnography is [still] about telling social stories’ (2008: 838). No surprise then that in such studies we find the traditional tensions of ethnography are just as applicable: for instance, after Kozinets outlines ten criteria of good ‘netnographic’ studies (coherence, rigour, literacy, groundedness, innovation, resonance, verisimilitude, reflexivity, praxis and intermix) he summarises by saying ‘Some of these criteria … even contradict one another. These contradictions signal the unlikeliness and perhaps unworthiness of simple off-the-shelf solutions. They invite netnographers to probe and struggle in a focused and guided way, and to discover their own path’ (2010: 162). This conception of digital ethnography, then, operates much the same as usual, and the usual contradictions apply.

‘Online’ and ‘offline’ as integrated whole

Others have sought to lean into how online practices intersect with offline practices in cyclical loops, each contextualising and shaping the other (e.g. Baker, 2013; Hallett and Barber, 2014; Lingel, 2017; Orgad, 2006; Przybylski, 2020). For instance, Lingel (2017) studies three different countercultural communities – body modification enthusiasts, punk rock musicians/fans in Washington, and drag queens in Brooklyn – placing their internet usage within a wider ecology of identity-building and community-organising practices. On this ‘flat hierarchy’ model of online/offline, Orgad notes ‘the relationship between online and offline is often interpreted as the relationship between phenomenon and context … Moving from online to offline helps us … to break down this dualism and see how each configures the other’ (2006: 63-64), thereby rendering online worlds accessible to ethnographic investigation as traditionally conceived (albeit carried out in ways that tie specifically to the platforms in question).

The ‘online’ as its own distinct space

Others still (Boellstorff, 2008; Boellstorff et al., 2012; Rutter and Smith, 2006; Snodgrass, 2005) have applied ethnographic thinking to sites which have no ‘offline’ analogue. For instance, Rutter and Smith's work on forum-type newsgroup user communities, of which they note: ‘By definition online ethnography describes places that are not spaces. Disembodied persons people these places. Such facts are the fuel of the cyberpunk imagination’ (2006: 84). However, despite the newness posited in such studies, again the old principles and tensions of ethnography apply – as Boellstorff et al., note, ‘fieldwork methods, developed phenomenologically for “being there” in real-time, existential communities in the here and now, can be applied faithfully and with ingenuity to virtual worlds and communities enabled by contemporary information technologies’. (Boellstorff et al., 2012: xiv). Hence, what is at stake in these studies is the newness and ingenuity required to situate a method in an unfamiliar context, rather than anything transformative of the method itself.

A possible way round: concluding remarks

For the purposes of an emerging Computational Ethnography which might wish to ‘digitally-remaster’ (Housley et al., 2014: 4) its methodological foundations and their internal tensions from the ground up, what might be done? The twin requirements of context-specificity and methodological prescription will never fully balance. Yet, we might note, we (anthropologists and sociologists) are not the only arbiters of methodological insight here, as indicated by an emerging body of research that attends to the ways in which the subjects of our study – people, platform users, communities etc. – leverage ethnographic-esque methods in the management of their own group members, mediated communications and other online activities. Noortje Marres has called this a redistribution of methods, within which:

Digitization enables a broadening of the agencies playing an active role in the enactment of social methods, broadly conceived: in this context, a wide range of actors including platform users and analytic devices like search engines come to play a part in the collection, analysis and presentation of social data. And this redistribution of methods in digital social research opens up a space of intervention for social research. (Marres, 2012: 160)

On this, Marres and Gerlitz note that these ‘are emerging methods that we – as social and cultural researchers – can't exactly call our own, but which resonate sufficiently with our interests and familiar approaches to offer a productive site of empirical engagement with wider research contexts, practices, and apparatuses’ (2016: 27). Working within the tradition of ethnomethodology – itself the kind of ‘critical friend’ to sociology's formal analytic methods (Garfinkel, 2002) – Housley et al. (2017) and Brooker et al. (2017) further elaborate on the notion of drawing peoples’ own digital methods from the world, and examining and applying those in our research efforts rather than seeking to understand the world by laying methods over the top of it. The distinction is perhaps subtle; between (a) beginning research with a methodological guiding principle (context-specificity) driven by disciplinary tradition, and (b) seeking to derive the haecceities of that initial methodological principle from the methods by which members themselves come to make sense of their social worlds. Though the surface look of a proposal to replace one set of prescriptions (of ethnographic method) with another (ethno-methods (Garfinkel, 1967, 2002)) may raise an eyebrow in the context of the current argument, what is being proposed is that our methods of knowing the social world should come from those whose day-to-day life produces it, rather than imposing a framework a priori. The value of this, for a field seeking to resolve the tensions borne out of the application of a priori methodological practices, will be in grasping an opportunity to draw on the multiplicity of methods that already exist and work as ways of knowing the world outside of the academy. Painting from a broader palette of methods is, perhaps, something anthropology (with its foundational reliance on ethnography) can then take from sociology; equally, being truly attentive to context by taking methodological cues from outside of disciplinary commitments altogether is, perhaps, something sociology can take from ethnomethodology. If such a point of restart should be attractive to Computational Ethnographers (irrespective of their backgrounds), this would suggest that ethnomethodology's approach to delivering knowledge of its methods – exhortations to go and see for yourself the infinite variety of everyday local methods of being in the world through collections of empirical demonstrations (cf. Garfinkel, 1967, 2002), with no prescriptions attached – could prove an apposite way to probe our methodological toolkits as they are in the process of being designed and constructed.