Risk, Society and Sociology of algorithms
« Risk is sometimes elusive since, by its very nature, it is the representation of a future danger, whose main change is its place in our societies. »
This case begins with an analysis of risk as an area of understanding, measurement and the factors that influence the perception and acceptability of risk. To do so, we will first seek to determine whether public policy is the preferred terrain for co-responsibility in managing uncertainties and whether there is specific public action for risk management. Under the cover of the concept of risk, there is a dualism between, on the one hand, the need for the evolution of knowledge and, on the other hand, the obligation to control its negative effects. Thus, starting from the definition of risk and the nature of its articulation with the notion of uncertainty in Ulrich Beck's sociology of the "risk society", we will identify the way in which this specific apprehension puts to the test certain foundations and interpretations of the dominant sociology of risk as it unfolds in the specific apprehension of the notion of risk in the Foucaultian-inspired theoretical framework of the literature on governmentality (governmentality studies). Then, in the second part, we will plunge into a second analysis of risk in the society of innovation.
1/ What is the sociology of risk?
The strength of a sociology of risk is to take full measure of the challenge of deconstructing the knowledge to which the production of a risk gives substance. If risk is something that does not exist in reality, it is nevertheless a way of ordering reality and rendering it in a calculable form. Bekkian sociology, which theoretically breaks with realism - or weak constructivism - is clear, and gives full measure to the plurality of the "nature" of risks by offering a flexible and fruitful framework. It is through the contemporary proliferation of social risks, and afterwards, of the history of science within science itself, that the description of what is recognized as a risk is described.
2/ Beck's meta-story
U. Beck's sociology is driven by the idea that it is possible to produce a general and comprehensive characterization of the contemporary period through a "meta narrative. This is part of a perspective of analysis of modernity and the process of modernization. Within this interpretive framework, modernity forms a kind of meta-concept that serves to describe a phase in the evolution of societies. The notion of modernity that characterizes the recent history of societies have thus been thought of as totally dedicated to modernity, and have turned it into an ideology. The theory of the "risk society" thus seeks to go beyond the debates of "post-ism": post-modernity, post-industrial, post-history, etc., and will attempt to provide a positive explanation of the present, grasped as a completed phase of the modernization process, a transitional phase of modernity, whether it takes the form here of an alarming critique of the dynamics that animate capitalism. If, for Beck, the theory of individualization has undoubtedly been the most discussed, as much for its stimulating perspectives and undeniable sociological imagination as for its alleged excesses - notably in the rapidity with which Beck announces the death of gendered classes and roles, and then the continuous forms of techno-economic progress - it is the theory of the globalized risk society that draws our attention as a starting point. Thus, we first note the keen interest in Beck's reflections on risk as a basis for critiquing dominant knowledge (scientific, expert and technocratic) and denouncing their active participation - both ideological and practical - in the production of what are now seen as the great current threats to humanity. However, beyond this contribution, it has been suggested that Ulrich Beck's realistic, uniform and totalizing approach has been relatively irrelevant in grasping the inclusion of a specific risk in a particular regime of government.
Governmentalist authors have been able to show how the socialization of risk has been amalgamated with general and abstract forms of social and political rationality that provide a conception and image of society to disciplines such as sociology, the "new social economy", and state doctrine down to politicians and lawyers. It is in this sense that the notion of society can be seen as much as an artefact of risk as vice versa. By government, we mean, from a "social" perspective, both government by the social, which is conceived as a sui generis reality, and government on behalf of the social (Rose 1996b, 349). Indeed, the governmentality approach forms a theoretical framework that generates a series of criticisms of Bekkian analyses, particularly with respect to political economy and institutions. This approach draws on Foucault's most recent work (2004a-b) and focuses on governmentality as a specific form of power exercise, and, on the other hand, it points to an era marked by the predominance of this form of power over others. In this sense, the Foucaultian narrative is also a meta narrative that characterizes the modern age, but the departure from Beck's theory is nonetheless notable from the outset. Where the German author sees "risk" as the product of a process of modernization, Foucault's challenge is to reconstruct the body of knowledge and calculations underlying the production of "risk" at the heart of transformations in the ways in which power is exercised.
a/ The sociology of quantification
Quantification has been a technology of the ruling elite, government and the territories and a means of individual conduct. Can it today become one of the components of a democratic society? Our era is characterized by a neo-liberal regime of knowledge production, where the traditional paradigm of science as a "public good", a source of progress and a pillar of the welfare state, is being replaced by the paradigm of science as a source of economic competitiveness. In fact, today, performance indicators are perceived as a source of economic competitiveness, including the competitiveness of the territory, which is linked to its attractiveness and its capacity to attract activities on its soil. However, we maintain that quantification and performance indicators are, in fact, a social practice that consists in putting into form many of the human conventions and that the constructivist perspective to which we adhere, allows us to recall that measurement is a social construct, and that it is precisely because the conventions from which it is derived are forgotten that quantification is naturalized and measurement reified in its uses.
b/ Statistical representation and the real
It should be remembered that statistics were first developed in conjunction with the State with a view to quantitative control of populations and intervention in the social field, but that it also had a managerial aim, which consisted in counting and classifying in order to organize and regulate social life. The sociology of quantification has taught us that statistical representations do not describe all reality in all its complexity, whereas its role is to establish units, standards, measurement methods and methods for calculating uncertainties. This change in perception has helped to promote the authority of the number in our contemporary societies, establishing "government by numbers" essentially based on the production of indicators, risk calculation and opinion polls. But measurement is also a resource that has made possible the development of science and technology leading to social and cognitive revolutions. New research would make it possible to understand that the tricks of algorithmic governmentality would no longer make it possible to consider quantification from the point of view of the sociology of statistics as an operation of instituting reality and transforming the world (Desrosières, 2002, Porter, 1996, Hacking, 1999 and Anderson, 1990).
Conclusion: Towards a Sociology of Algorithms?
A new field of study is emerging in the social sciences, specializing in the study of algorithms, whose legal and Foucaultian approach demonstrates that, starting from the notion of "algorithmic governmentality", profound transformations are made possible by machine learning. By showing that the "social norm emerges from reality itself" through the process of statistical learning, the Foucaultian approach enlightens the statistical sociologist as much as it troubles him. If the forms of quantification inherent in statistics embarrass criticism, should the sociology of statistics collapse? For however relevant it may be, analysis in terms of the dispossession of criticism limits the concrete possibilities of emancipation in the face of algorithms. Indeed, it locks the actors, and even the sociologists themselves, into a position of powerlessness. Is it necessary to abandon the sociological analysis of machine learning algorithms, and to rely instead on the sociology of quantification, on a sociology inspired by the law as the obligatory point of passage; and on a sociology that is involved and embarked on as a result of social reverse engineering ?