Classify, assess, prioritize

The distance which separates the ideal of a society of equals – where everyone’s place would only be defined by virtue and talent – of the empirical observation of racial, sexual or social origin is a persistent difficulty for the social sciences. It raises in hollow the question of the evaluation of individuals: what are the mechanisms which contribute to establishing distinctions between the members of a society and to attribute to them in this way a social value ?

At a time when our societies are always more digitized, Marion Fourcade and Kieran Healy are reinvesting this problem as classic as it is delicate by showing that the ordering of the social is now largely based on algorithmic operations of classification and rating of individuals. Deployed on a large scale by the public authorities such as private players, these classification technologies are not content to influence the place occupied by each, they also modify our own representations and moral judgments. By showing the importance they have acquired today, the two sociologists draw the contours of what they call “ ordinal company ».

Data capture

Understand the meaning given to the enigmatic concept of “ ordinal company »Invise to return to the upheavals generated by digital technologies. As we know, the multiplication of computers, smartphones And other connected objects has the effect of translating whole sections of our data into data. Interactions on social networks, travel made thanks to Google Maps Or the purchases made by bank card is all information that is collected by a myriad of organizations, both private and public.

Based on abundant literature relating to contemporary forms of technological surveillance, Marion Fourcade and Kieran Healy highlight the main developments that have led to the current abundance of data. If states and businesses have long collected information, especially about their workers, customers and administered, this process has gained considerable scale with the birth of IT. At the turn of XX^e A century, an additional decisive step was taken by certain digital companies, like Google which changed its business model by offering advertisers to target their advertisements on the basis of behavioral traces left by users of the site. Gradually spread the idea that any type of information – including the most insignificant – is a potential source of profit. In parallel, the progress made in the field of automatic learning (machine learning) made possible the analysis of large amounts of extremely varied data.

The book underlines the ambivalence of the very notion of data, the latter being more captured without the knowledge of the person to which he relates to freely conceded. The authors recall Rob Kitchin’s suggestion in this regard that it would be more correct to talk about capta what data. A chapter of the book is, however, to show that an important part of the digital economy is well underpinned by a logic of the donation which they describe as “ Maussian haggling “(“ MAUSSIAN BARGAIN », P. 43). By that, they designate the process under which Internet users agree that their behavioral traces are recorded and analyzed in exchange for free access to the services offered by large platforms.

Among the information collected, Marion Fourcade and Kieran Healy also insist on the importance of data indexical which make it possible to identify individuals and therefore link different data games. In addition to the identifiers traditionally used by states, such as the name, address, or social security number, a variety of markers such as the address IPthe serial number of a device, a loyalty card, or a biometric line is used to associate personal data from various sources. These indexical data are the basis of the digital tracing to which the data brokers proceed (Data Brokers) and large platforms.

The development of classification technologies

This massive collection feeds algorithmic models which claim to predict the future behavior of individuals by attributing a note to them or by storing them in a predefined category. Our societies are ordinal in that that citizens are constantly ordered from each other, in various contexts and on different measurement scales. As an example, private companies produce, in the United States as in many other states, supposed credit scores describe the solvency of consumers. The banks consult the scoring of a candidate for a loan before deciding to grant him. Many employers use automatic job software, while states rely on algorithms that assess the risk that a migrant is a potential terrorist or a social insured is fraudster. These classification technologies therefore underlie decisions that have a considerable impact on life opportunities (“ Life Chances ) People evaluated.

Most of the information that scores come directly from the digitized world and thus seem to offer direct apprehension of individuals. The treatment which is reserved for the latter is less based on their belonging to the traditional socio -demographic categories (such as diploma, income level, gender or ethnic origin), than determined by their own behaviors which are recorded by the countless sensors that saturate our daily lives. All these personal traces constitute a form of resources insofar as they condition access to goods, services or rights ; a resource that must be learned to manage in order to produce “ good data ».

Knowing what makes good or bad data, however, is not obvious given the opacity of algorithmic models. Not only are their designers reluctant to deliver the code, but the most advanced methods of automatic learning also make unintelligible and therefore inexplicable predictions. And despite the alleged surpassing groups to which algorithmic devices would proceed, the latter prove to be on the contrary tirelessly reproducing the inequalities which structure the social world.

Ordinal citizenship: a question of value

However, everything suggests that these ordinal technologies say something about the intrinsic value of the people they measure. One of the great forces of the book is precisely to show that these not only transform the way in which individuals are apprehended by platforms, businesses or states, but that they also upset representations, that is to say the way in which individuals relate to others and to themselves.

From the introduction of the book, the authors underline how decisive the evolution of the web was from this point of view: while the initial vocation of the canvas was to allow a decentralized sharing of information thanks to the interconnection of pages (which remained static), the Web 2.0 encouraged the active participation of Internet users – in particular via the evaluation devices and people – and foster the aspiration of everyone Compare, as well as being properly evaluated.

In this sense, the ideal of ordinal society is to be “ evaluated as an individual rather than as a member of a group, and gauged according to his actions and not his thoughts “(“ Citizens SHOULD BE ASSEESSED AS INDIVIDIENTIALS RATHER THAN AS MEMBERS OF Groups, and GAUGED NOT BY Their Thoughts But by Their Actions », P. 237). However, because they are based on numerous and behavioral data, the hierarchies that classifying technologies establish at the same time exact, individualizing and morally correct. They are carrying a double claim to objectivity and justice which makes them particularly difficult to challenge.

The diversity of ordinal methods

The question arises whether all forms of evaluation contribute in the same way to the measurement (and government) of individuals. The book thus seems to place the note awarded by a consumer to a hotel or taxi driver on the same level as the credit score established by a bank about a borrower. If it is in both cases of a quantified assessment, it is useful to highlight what distinguishes them. The algorithmic scores aim for an objectivity that comes from their statistical base, which is supposed to guarantee their predictive value. On the contrary, the evaluation by consumers of goods or services is the expression of subjective preferences and constitutes a measure of reputation and not a prediction of behavior. In the same way, the pairing methods – such as Parcoursup in France which combines establishments of higher education and students – meet specific objectives and are based on their own techniques (typically the algorithms based on the theory of stable marriages of scabies and Shaley). In this sense, it would undoubtedly be fruitful to emphasize the diversity of ways of comparing, assessing and sorting individuals through digital technologies.

These remarks do not, however, invalidate the link that the book establishes between these different types of evaluation that they do not invite to extend its careful analysis. In this regard, and although the examples mobilized come mainly from the United States, The Ordinal Society Makes an important contribution to the numerous works that study the digitization of our societies and will not fail to nourish subsequent research in this area.