BEGIN:VCALENDAR VERSION:2.0 PRODID:-//AMSE//Event Calendar//FR CALSCALE:GREGORIAN METHOD:PUBLISH BEGIN:VEVENT UID:event-13045@www.amse-aixmarseille.fr DTSTAMP:20260429T082452Z CREATED:20260429T082452Z LAST-MODIFIED:20260429T082452Z STATUS:CONFIRMED SEQUENCE:0 SUMMARY:big data and econometrics seminar - Agathe Fernandes Machado DTSTART:20260203T130000Z DTEND:20260203T143000Z DESCRIPTION:Algorithmic fairness refers to the set of principles and techni ques aimed at ensuring that the decisions produced by an algorithm are fair and non-discriminatory toward all users\, regardless of personal character istics such as gender\, ethnicity\, or other so-called sensitive attributes . Its assessment can be carried out at multiple levels: on the one hand\, a t the group level\, by comparing a model’s predictions across different g roups defined by sensitive variables\; and on the other hand\, at the indiv idual level\, by focusing on a specific individual from a minority group an d asking counterfactual questions such as: “What would this woman’s sal ary be if she were a man?” To evaluate algorithmic fairness at the indivi dual level\, we adopt the notion of Counterfactual Fairness proposed by Kus ner et al. (2017). This approach relies on the mutatis mutandis principle\, in contrast to the ceteris paribus principle: rather than checking whether a model’s prediction for an individual remains unchanged when only the s ensitive attribute is modified while keeping all other explanatory variable s constant\, we ask whether the prediction remains the same when only the v ariables not causally influenced by the sensitive attribute are held consta nt. The definition of Counterfactual Fairness relies on Pearl’s (2009) ca usal inference framework and involves computing an individual’s counterfa ctual in which the sensitive attribute is modified\, assuming prior knowled ge of a causal graph over the model’s explanatory variables. In this stud y\, we link two existing approaches to derive counterfactuals: intervention -based approaches on a causal graph with quantile preservation\, as propose d by Plečko et al. (2020)\, and multivariate optimal transport introduced by Lara et al. (2024). We extend the concepts of “Knothe’s rearrangemen t” and “triangular transport” to probabilistic graphical models and e stablish the theoretical foundations of a counterfactual approach\, called sequential transport\, to discuss individual-level fairness.\\n\\nContact: Sullivan Hué: sullivan.hue[at]univ-amu.frMichel Lubrano: michel.lubrano[at ]univ-amu.fr\n\nPlus d'informations: https://www.amse-aixmarseille.fr/en/ev ents/agathe-fernandes-machado LOCATION:Îlot Bernard du Bois - Salle 24\, AMU - AMSE\, 5-9 boulevard Maur ice Bourdet\, 13001 Marseille URL;VALUE=URI:https://www.amse-aixmarseille.fr/en/events/agathe-fernandes-machado CONTACT:Sullivan Hué: sullivan.hue[at]univ-amu.frMichel Lubrano: michel.lu brano[at]univ-amu.fr TRANSP:OPAQUE END:VEVENT END:VCALENDAR