WHAT PERSONALISATION CAN DO FOR YOU! OR, HOW TO DO RACIAL DISCRIMINATION WITHOUT RACE

Between 2016 and 2020, Facebook allowed advertisers in the United States to target their advertisements using three broad “ethnic affinity” categories: “African American,” “U.S.-Hispanic,” and “Asian American.” This paper uses the life and death of these “ethnic affinity” categories to argue that they exemplify a novel mode of racialisation made possible by machine learning techniques. These categories worked by analysing users’ preferences and behaviour: they were supposed to capture an “affinity” for a broad demographic group, rather than registering membership of that group. That is, they were supposed to allow advertisers to “personalise” content for users depending on behaviourally determined affinities. We argue that, in effect, Facebook’s ethnic affinity categories were supposed to operationalise a “post-racial” mode of categorising users. But the paradox of personalisation is that in order to apprehend users as individuals, platforms must first assemble them into groups based on their likenesses with other individuals. This article uses an analysis of these categories to argue that even in the absence of data on a user’s race—even after the demise of the categories themselves—users can still be subject to techniques of inclusion or exclusion for discriminatory ends. The inductive machine learning techniques that platforms like Facebook employ to classify users generate “proxies,” like racialised preferences or language use, as racialising substitutes. This article concludes by arguing that Facebook’s ethnic affinity categories in fact typify novel modes of racialisation today.

Socio-Economic or Emotional Predictors of Populist Attitudes across Europe

Previous research on predictors of populism has predominantly focused on socio-economic (e.g., education, employment, social status), and socio-cultural factors (e.g., social identity and social status). However, during the last years, the role of negative emotions has become increasingly prominent in the study of populism. We conducted a cross-national survey in 15 European countries (N=8059), measuring emotions towards the government and the elites, perceptions of threats about the future, and socio-economic factors as predictors of populist attitudes (the latter operationalized via three existing scales, anti-elitism, Manichaean outlook, people-centrism, and a newly developed scale on nativism). We tested the role of emotional factors in a deductive research design based on a structural model. Our results show that negative emotions (anger, contempt and anxiety) are better predictors of populist attitudes than mere socio-economic and socio-cultural factors. An inductive machine learning algorithm, Random Forest (RF), reaffirmed the importance of emotions across our survey dataset.

Learning temporal rules to forecast instability in continuously monitored patients

Inductive machine learning, and in particular extraction of association rules from data, has been successfully used in multiple application domains, such as market basket analysis, disease prognosis, fraud detection, and protein sequencing. The appeal of rule extraction techniques stems from their ability to handle intricate problems yet produce models based on rules that can be comprehended by humans, and are therefore more transparent. Human comprehension is a factor that may improve adoption and use of data-driven decision support systems clinically via face validity. In this work, we explore whether we can reliably and informatively forecast cardiorespiratory instability (CRI) in step-down unit (SDU) patients utilizing data from continuous monitoring of physiologic vital sign (VS) measurements. We use a temporal association rule extraction technique in conjunction with a rule fusion protocol to learn how to forecast CRI in continuously monitored patients. We detail our approach and present and discuss encouraging empirical results obtained using continuous multivariate VS data from the bedside monitors of 297 SDU patients spanning 29 346 hours (3.35 patient-years) of observation. We present example rules that have been learned from data to illustrate potential benefits of comprehensibility of the extracted models, and we analyze the empirical utility of each VS as a potential leading indicator of an impending CRI event.