Articulation, or the persistent problem with explanation

Abstract

Sociologists have long argued that explanation, as a form of knowledge, has serious limitations when it comes to understanding society. The case against explanation is one of the field's founding ideas, it is literally a foundational idea. It was by rejecting causalist forms of explanation that had been developed in the natural sciences that 19th century scholars and activists that we today call sociologists succeeded in articulating a distinctive realm of reality with relative autonomy from the state, the economy and the family: society (Wagner, 2000). Key to their achievement was the argument that the phenomenon of society is fundamentally different from nature. Scientists at the time expected nature to obey eternally valid laws, but society has a number of features that challenge this assumption. Social actors formulate norms and rules to justify their actions and to make sense of social reality. This means that norms and rules themselves may play an active role in the transformation of social reality. Society, in other words, is marked by reflexivity. Of course, a lot has happened since the 19th century and this very notably includes unrelenting efforts by social scientists to create forms of explanation that are capable of taking reflexivity into account. Yet problems with explanation have continued to make themselves felt in the social sciences and humanities. The problem, in a nutshell, is that explanation sets up the relation between social science and its object, society, in terms of representation, but the relation between knowledge about society and social reality is fundamentally an interactive one: the creation of knowledge about society far more often than not involves intervention in society. The creation of social scientific knowledge can rarely, if ever, by considered a purely representational affair. This obtains for practically all forms of knowledge about society - and as we shall see, about nature as well - but it causes specific problems for the explanation of social phenomena. Let me give an example from contemporary social science, broadly defined. Some years ago computational social scientists published research that showed that the high levels of political polarization that can be observed among communities on Facebook cannot be explained by the role of social media algorithms in the promotion of content. As they put it: “individual choices, more than algorithms, limit exposure to attitude-challenging content” (Bakshy et al., 2015, p. 1131). Such a claim asks us to accept a number of assumptions, most notably, that it is possible to disentangle the influence of individual user choices on news consumption on Facebook from the influence of platform settings such as the structure of news feeds.1 This assumption may or may not ultimately be methodologically convincing. But in grounding its main finding in this distinction, this study distracts attention from a more fundamental phenomenon: that “choice” in online platform settings is socio-technically constituted in a highly distinctive way, involving clicks on links that are dynamically served up by the platform based on social network analysis, among others. This type of “choice” presents a very different form of action as compared to say, choosing what article to read in a paper newspaper. However, and this is the key point, affirming such ontological complexity would no doubt be seen as reducing the “strength” of the explanation offered. As the recently deceased constructivist sociologist Aaron Cicourel (1964) pointed out many decades ago, to draw attention to the participation of the underlying apparatus of social research - in this case, Facebook data categories such as “clicks” and “friends” - in the construction of social reality is to challenge the representational understanding of social science in general and of explanatory social science in particular. Explanation requires the relation between social categories and social reality to be stable and one-way (unidirectional): it requires that social scientific categories first and foremost refer back to social reality. When the apparatus of social research is shown to interfere in the realities it purports to measure, it is clear that this does not quite obtain. A persistent problem with explanation, then, is that its validity seems to depend on the bracketing, externalising or trivialising of dynamics of reflexivity. If the proponents of the idea that “social science is explanation, or it is nothing” would get their way, and social science indeed would offer only explanations, and nothing else, this would surely end up restricting our capacity to interrogate the manifold ways in which social categories interact with social realities. However, most people who are interested and/or trained in sociology are well aware of the phenomenon of reflexivity, of the power of social categories to shape social reality. So why do so many sociologists today favour explanation over other, more open-ended forms of knowledge, like ethnographic description and theory-driven interpretation, methodologies which have been specifically designed to enable interrogation of the interactive relations between social categories and social realities (Krause, 2016)? I would like to argue here that there is another layer to this phenomenon of interactivity, one that has less to do with how norms, categories and methods shape reality, and more with how social science achieves what Norbert Elias (2011) called adequacy to social reality. For me, there is a danger that lurks in the commitment to “explanation” that is related but different from the problem that it legitimates or encourages indifference to reflexivity, the danger namely that it distracts from a key task and purpose of social science: articulation. The slogan “explanation or nothing” makes me think of a remark by the German media scholar Erhard Schüttpelz who once reminded me that “all of our concepts are going downhill all of the time.” At the time, we were speaking about the concept of “media bias,” and the ways in which understandings of such bias developed in the 1970s are no longer adequate, though still highly relevant, to today's digital society with its algorithmic organisation of content, where frames are consolidated through post-discursive, automated, processes of selective circulation. I continue to find Schüttpelz comment helpful more generally speaking, as it draws attention to a brute fact that deeply affects the relation between the social sciences and society, which is that the world, society, is continously changing, and this changing world continuously renders our existing vocabularies, and forms of explanation, deficient. In other words, we should expect our existing social theories to be losing their adequacy to our present social reality. Schüttpelz slogan reminds us that many of the phenomena that it is our job to understand are invisible or badly named, and are difficult to even observe let alone measure. Take the example of the security state, which in its current technological form - which is marked by the deployment of a data-intensive apparatus of monitoring and control - developed in the post-war period, from the 1960s onwards. As the French sociologist Dominique Linhardt (2008) has shown, the very observability of the security state during this period was dependent on targetted interventions in society: it partly depended on the interventions of militants such as the Rote Armee Fraktion in Germany and Provo in the Netherlands whose demonstrations brought the police out onto the streets for all to see. It was these interventions that rendered the phenomeon of the security state publicly visible, putting the repressive apparatus of the state monitoring activists onto the front pages of newspapers. But the observability of this type of information-based security state also depended on the invention of a new kind of sociology, which had to rework its concepts of power, in order to be able to demonstrate the significance of the technological security apparatus to the structuring of wider societal relations between population and state during this period. The point is, in a case like this, it is clear that both a phenomenon in society - the security state - and the categories we use to understand this and associated phenomena - state power - underwent transformation during that relevant period. This is not a context in which “explanation” works very well. To “explain” the security state requires the very phenomenon to be observable and identifiable in the first place, and a lot of work has to be done before we get to that point: the work of articulation. If we go along with Erhard Schüttpelz and recognise that events in the world render our ways of knowing deficient all the time, then it is clear that our work as sociologists must be re-constructive, we will need to be able to revise and rethink our categories, so we remain capable of naming phenomena in the world adequately, so that we can make them count. To give another example from my own field: when in the 1980s feminist technology studies developed the notion of “invisible labour” through fieldwork studies in workplaces like airports and offices (Star, 1991; Suchman, 1996), they transformed the wider concept and understanding of innovation. At the time, innovation was a restrictive notion - and it still is - which allocates a disproportionate amount of agency to the engineers who create technology, while defining the rest of us, workers and consumers, as more or less passive “users.” Feminist studies of invisible labour in technology-mediated work used ethnographic description to demonstrate how often female workers accomplished tasks - performing flight checks, copying documents - that were later acribed to the smooth running of “technology.” It was these descriptions of invisible labour that allowed this phenomenon to be named and defined. Only then could the work begin of incoporating sensibility to unrecognised work into the social study of innovation and its “apparatus of explanation,” and gain the efficacy it has today, for example, in the form of a critical exposé of the invisible labelling that workers in Kenya, India and other countries perform, so that “generative AI” can do well at image recognition (Catanzariti et al., 2021). To emphasise articulation as a key task and purpose of social inquiry is certainly not to imply that explanation is without value. Finding ways to measure the prevalence of invisible labour in society so as to be able to demonstrate its existence, explain its workings, and connect its persistence with wider underlying dynamics in the technological economy, is crucial. The problem, that is, is not with explanation as such, it is with the “or nothing” bit in the slogan above. The adequacy of our explanations to the world is dependent on the prior work of articulation, and it is this dependency that is at risk of being disregarded when it is suggested that explanation is all we need. An important part of the task of social science is to develop vocabularies, formulate categories, cultivate sensibilities, so as to render ever-changing phenomena in society observable, explorable, and communicatable. To name, to formulate, to label, takes us halfway to understanding. It is to draw what Alfred Schutz (1970) calls the “isohypses of relevance,” enabling some entites, some dimensions to stand out, to gain traction, in our engagement with the world, while submerging others. This is the work of articulation, as opposed to explanation. Without it, our explanatory apparatus will not just keep going down hill. Its adequacy will eventually be lost. There is an odd irony in the suggestion that sociologists should today consider putting all their eggs in the basket of “explanation.” We are living at a time in which the natural sciences are finally beginning to come to terms with the interactivity of science, as they grapple with the Anthropocene. Today, more and more colleagues in the natural and technical sciences are gripped by the realization that science has intervened in and indeed harmed and damaged the world in ways to which scientific epistemology previously rendered scientists blind. Representationalism places on science the methodological requirement to regard as external to the scientific mission - which is to represent reality - how its apparatus of representation impacts the world. Representationalism is part of what enabled scientists to cultivate indifference to how their work transformed the world physically, socially, materially, environmentally: think of the stress experienced by lab animals and lab technicians (Friese & Latimer, 2019); the PFAS molecules which are today found in ground and sea water, and known to cause cancer, and which for decades were represented in terms their qualities of “water-proofing” and “stick-resistance” (for carpets). There is a more complex story to be told here about the relation between science and innovation, but the problem is similar to the problem of methodologically ordained dis-interest in socio-technical complexity that I mentioned in the introduction. Damaging effects produced by the apparatus of scientific research on the world used to be referred to as “unintended consequences,” but they are today understood by an increasing number of natural scientists as a siginificant factor in the degradation of nature, animal and human life. Isn't it odd that at the very time that scientists are beginning to recognise interactivities between science and the natural and social world, we are debating whether the primary role of the social sciences is to represent the world? Why? Why would a sociologist today accept that “explanation” is the only serious game in science, at a time when the sciences themselves are finally waking up to the manifold problems with “explanation” as a scientific framework? Granted, one of the benefits of “explanation” is that it has a strong “theory of change” associated with it. Explanation offers us a tight, sequential relation between knowledge and action, whereby the representation of an underlying reality creates the basis for transformative intervention in society, for social change. By comparison, a theory of knowledge that puts articulation centre stage offers a far more contingent, multi-factored and messy view of how change happens in the world, and it requires an openness of inquiry to engage and continually attune understanding to a changing social world. My sense is that it is the strength and simplicity of the associated theory of change that the methodological paradigm of “explanation” promises, that explains much of its attractiveness to sociologists today. Perhaps it can be seen, too, as a response to the utterly restrictive expectations that are placed today on the social sciences. Explanation is able to offer knowledge of long-term processes and enduring underlying causes, which is badly needed in the face of muliplying social, environmental and political crises. It offers a robust alternative to the instrumental conception of social research, which is increasingly imposed on us by state and industry, and which requires that social science delivers on short-term policy needs and value for money, and which can seem to value knowledge of society only to the extent that it is able to produce demonstrable changes in behaviour and bottom-lines (Kelly & McGoey, 2018). However, precisely at this moment, when the social sciences face challenges from multiple fronts - from short-termist expectations of impact, to policy-makers endearement with behavioural science as provider of readily applicable solutions, and the eagerness of some computational scientists to serve as the go-to provider of “social explanations” - it is important that we don't feel pressured to retreat into the narrow methodological frame of “explaination” and to risk inadvertingly reasserting a false hierarchy between explaination and articulation. There is a vital difference between an instrumental social science whose value and legitimacy primarily derive from its ability to service short-term policy needs, and a sociology capable of affirming interactivity between social science and the social world. This difference needs to be urgently clarified, if we are to avoid sleepwalking into a future of shrinking sociological repertoires, and importantly, if we are to get better at engaging, informing and indeed leading the debate that is finally getting underway across the sciences, about what it means to affirm interactivity between knowledge and its objects, not only ethically and politically, but methodologically speaking. To sum up, my real disagreement today is with the “or nothing” element in the idea that social science is explanation or it is nothing. I don't have a problem at all with explanation as a form of social knowledge taking its rightful place among other forms of knowledge in the social sciences. I have a problem with the suggestion that explanation is superior to these other forms, and that other ways of knowing society need to make way for it. My problem is with the hubris of a social science that assumes that the work of articulation, of making phenomena perceptible, can be economised on. It cannot. Explanation depends on articulation: we can only posit and define phenomena and their underlying causes once the struggle to name, to make count - the struggle for the existence of a category - has been fought. The apparatus of explanation that we all rely on - the categories, standards, measures and data formats that together constitute the apparatus of social science - is always at risk of perpetuating outdated, inadequate categories. To the extent that "explanation" assumes that the work of articulation - what matters? how to name this? - can be taken for granted, it risks to become complicit in eroding conditions for articulation. The “or nothing” in the above motion wrongly implies that the renewal of our vocabularies, concepts and methods is of secondary importance. I would therefore like to turn the proposition around. “Social science is reflexive or it is nothing.” Sociology has impressive analytic and methodological resources at its disposal for creating knowledge under conditions of interactivity, and it is this capacity that should be celebrated and requires our endorsement, not just for the sake of advancing the social sciences but all sciences. Social studies of science and technology have shown how interactivity does not just operate on the plane of categorization, but also through indicators, infrastructures and indeed the world, as measures from GPD to the social media “clicks” and the have to the of realities in their affirming such interactivity effects does not making explanation the there are forms of explanation that the changing relations between social reality and the of social reality as their object, as in the work of Norbert Elias In sociological studies of how Elias a of explanation in which change in the explained partly by the dynamics of the I is that our categories, the apparatus that we rely on to render social phenomena and can be understood as a part of these As sociologists have pointed out since at the what and are relevant to the - and - of a social phenomenon is partly at in the of its articulation. is that this to our our categories and the measures that social science on. too, partly depend on and cannot be taken for to and for helpful and to the other in the “Social science is explanation, or it is nothing” for their This debate place online on