Pedro Saleiro: “Fairness in Algorithmic Decision-Making”

Pedro Saleiro, Senior Research Manager at Feedzai and an expert on Ethical issues and Artificial Intelligence, made a presentation based on his chapter in this forthcoming book “Big Data and Social Science: A Practical Guide to Methods and Tools”. To see the chapter, click here.

Here you have a summary of the chapter:

“Interest in algorithmic fairness and bias has been growing recently (for good reason), but it’s easy to get lost in the large number of definitions and metrics. There are many different, often competing, ways to measure whether a given model is statistically “fair” but it’s important to remember to start from the social and policy goals for equity and fairness and map those to the statistical properties we want in our models to help achieve those goals. In this chapter, we provide an overview of these statistical metrics along with some concrete examples to help navigate these concepts and understand the trade-offs involved in choosing to optimize to one metric over others, focusing on the metrics relevant to binary classification methods used frequently in risk-based models for policy settings.”

 

Lea Heyne and Luca Manucci: “A new Iberian exceptionalism? Comparing the populist radical right electorate in Portugal and Spain”

Lea Heyne and Luca Manucci, from ICS, presented a paper with the following abstract:

“Since their transition to democracy in the 1970s, Spain and Portugal have been ‘immune’ to the success of populist radical right (PRR) parties, because of the stigma associated with the long-pasting right-wing authoritarian past and other unfavorable opportunity structures. Since 2019, however, the two Iberian countries are no longer negative cases given that Chega’s leader, André Ventura, was elected in the Portuguese parliament, while Vox has become the third most voted political party of Spain. Using new online survey data from the Spanish and Portuguese national elections in 2019, we investigate the similarities and differences between the electorates of Vox and Chega. We find that the Iberian PRR electorate is mostly in line with the characteristics of the PRR electorate in Western Europe when it comes to socio-demographics, political dissatisfaction, media diet, and the rejection of immigration and feminism. Interestingly, several elements seem to set apart the type of support for Chega and Vox: none of their electorates can be classified as typical globalization losers, and they both capitalize on country-specific issues: national unity in Spain, and welfare in Portugal. We conclude that the failure of PRR parties in Iberia has ended for peculiar reasons, and therefore the success of Vox and Chega might be temporary and due to volatile conditions.”

 

Eva Anduiza and Roberto Pannico: “Timing partisanship: stability, strength and sway of party identification with new parties”

Roberto Pannico, from ICS, presented a paper. The abstract reads:

“Time has been considered an essential element in the construction of party identification, both for classic and for revisionist accounts. Partisanship is reinforced by time, regardless of whether it is based on a social identity or on a running tally. But when new parties appear, time is, by definition, short. In this paper we explore the dynamics of party identification with new parties. Using panel data collected in Spain during a period of profound party system change, we estimate the extent to which partisanship with new parties is stable, intense and works as a heuristic for preference formation and vote choice. Our data suggest that time is not a necessary requirement to develop relatively stable, strong and meaningful partisanship. These findings seem particularly relevant in a context where new parties are on the rise.”

 

Emerson do Bú: “The Intergroup Time Bias and Its Psychological Mechanisms in the Medical Context”

Emerson do Bú, a doctoral student at ICS on Social Psychology, presented some results from his PhD project. The abstract of his project reads:

“The Intergroup Time Bias (ITB) effect is an implicit discriminating behaviour characterized by the individuals’ motivation to invest more time evaluating ingroup than outgroup members. This research program aims to investigate this phenomenon in the medical context. We specifically propose that (a) physicians invest more time evaluating and making a diagnosis of White than Black patients, and that (b) this occurs throughout aversion and attention mediation processes. We will test these original hypotheses throughout five experimental studies using implicit measures. Study 1 aims to test whether the ITB occurs when physicians form an impression of patients. Studies 2 and 3 go further by testing the hypothesis of an ITB effect when physicians elaborate diagnoses and indicate treatments for black and white patients. Studies 4 and 5 will test if the influence of patients’ skin colour on ITB is mediated by aversion and attention towards Black and White targets.”

 

Rui Costa-Lopes, André Mata and Cristina Mendonça: “Real people or mere numbers? The influence of kill-save ratios and identifiability on moral judgments”

Rui Costa-Lopes,  from ICS, André Mata and Cristina Mendonça, from Faculty of Psychology – ULisboa, presented a paper. The abstract reads:

In moral dilemmas, decision-making can be based on more utilitarian or deontological reasoning. In a classical trolley dilemma, the indecision lies between choosing to sacrifice one person to save five (utilitarian decision) vs. not sacrificing a human life in any circumstances (deontological decision). In two experimental studies, we manipulated the number of people to be sacrificed (1 to save 5 vs. 3 to save 5) and whether personalizing information about them was presented. Results provide the first evidence of how the importance of kill-save ratios and identifiability of the potential victims are contingent on one another. Specifically, this research shows that when individuating information about the potential victims is present in a trolley dilemma, participants are more reluctant to sacrifice three persons to save five than to sacrifice one person to save five. When such individuating information is nor present, the acceptability of sacrificing the victims does not depend on their number.​”

Mariana Pires Miranda: “Early Release from Prison in time of COVID-19: 4 Determinants of Unfavourable Decisions towards Black Prisoners”

Mariana Pires Miranda,  from ICS, presented a paper from the following authors: Mariana Pires Miranda, Rui Costa-Lopes, Gonçalo Freitas & Catarina Carvalho. Here is the abstract:

“On the onset of the COVID-19 pandemic, the overcrowding in prisons led to efforts to decarcerate. This study analyses the support for such public measures, bringing to the table the cognitive and ideological factors known to create and maintain racial biases in the criminal system. Participants were asked to express their level of agreement with the early-release of hypothetical prisoners. Results showed participants to be less favourable to the early-release of Black compared to White prisoners, when they had committed a stereotypically Black crime. As expected, the crime congruency effect did not replicate in the case of the White prisoners. Moreover, the difference between the agreement with the release of the Black vs. the White prisoner when both committed a stereotypically Black crime was higher as the level of endorsement of Meritocracy increased. Contrastingly, anti-egalitarianism only predicted an overall disagreement with prisoners’ early-release. This paper highlights the cumulative explanation by different levels of analysis of this current problem.​”

 

Jorge Vala: “Introduction – Threats: An indispensable Debate”

Jorge Vala,  from ICS, presented the introduction of the book “Societies Under Threat – A Pluri-Disciplinary Approach” edited by Denise Jodelet, Jorge Vala and Ewa Drozda-Senkowska. The summary reads:

“This Seminar aims to stimulate the discussion about a concept – threat. Threat still has little emphasis in the analysis and theorization of social sciences which concern the changes that have taken place in the last twenty years, and those that we imagine will still occur. For a long time, risk was considered an adequate and sufficient concept to address situations marked by uncertainty and unpredictability. Threat is not an alternative to the concept of risk but it may allow a broader focus encompassing phenomena analysed under risk and also extends to phenomena ignored by risk analysis. Thus, this seminar asks, can the concept of threat bring us anything new? Three topics will be addressed: conceptual challenges; the potential operative role of the concept in different domains; reactions to threats.​”

Pedro Magalhães and Mariano Torcal: “Ideological Extremism, Perceived Partisan Polarization, and Democratic Support”

Pedro Magalhães,  from ICS, presented a paper. The first paragraph of the introduction reads:

“What are the implications of ideological polarization for the democracy legitimacy? Sartori noted long ago that when the ideological distance between the “lateral poles” of a party system is very large, “covering a maximum spread of opinion, (…) [t]his is tantamount to saying that cleavages are likely to be very deep, that consensus is surely low, and that the legitimacy of the political system is widely questioned” (Sartori 2005 [1976]: 120). Sartori’s prescient concern with the implications of ideological polarization for regime support has experienced a revival in recent years. Polarization has been argued to constitute, in different ways, a potential danger for democratic stability and, particularly, for citizens’ principled commitment with liberal democracy, its principles, and practices (McCoy et al. 2018; Levitsky & Ziblatt 2018; Carothers & Donahue 2019; Svolik 2019; Svolik 2020; Nalepa et al. 2018; Graham & Svolik 2020).​”

Michael Lewis-Beck forecasts the winner of the next US election

Michael Lewis-Beck presented a forthcoming paper of his entitled: “U.S. Presidential Election Forecasting: The Economist Model”. The preview reads:

“In June of this year, The Economist began publishing regular forecasts of the outcome of the 2020 U.S. presidential election. In this article, Colin and Michael Lewis-Beck describe the model used, evaluate its potential strengths and weaknesses, and provide many perspectives on election forecasting models in general. They conclude with forecasts of the results of the vote in the upcoming November 3 U.S. presidential election.​”

 

Raquel Rego, Tatiana Pita & Cristina Nunes “The crisis of European tripartism: The case of gender equality on social concertation in Portugal”

Raquel Rego did a presentation concerning the contribution of social concertation through the social pacts on gender equality in the labour market.

Here’s the abstract:

“​​​Within the corporatist systems of the EU, social concertation is the body that promotes the national consensus among trade unions, employers and governments. Given the persistent inequality of the labour market, we have wondered what contribution it has been making to gender equality policies. We responded to this question by analysing the content of the social pacts signed in Portugal over more than 30 years of tripartism (1984-2019). The results show that gender equality has little weight in social pacts and that the European regulatory framework is decisive in the progress made. Our study thus reinforces the existing literature which points to the need for a reform of tripartism.”