Our joint article Mitigating Discrimination in Insurance with Wasserstein Barycenters with François Hu and Philipp Ratz is now published, in the Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2023) volumes.
Assurance, IA, biais et équité
Pour compléter mon exposé auprès de l’Association des Masters d’Actuariat, je vais poster un billet, qui reprend les slides de mon intervention, avec un peu du discours qui va avec. Tout d’abord, mille excuses, mais les slides sont en anglais. Et ce que je raconte est très librement basé sur des travaux récents qu’on a pu mener, avec pas mal de monde, dont Laurence Barry (de la chaire PARI), Marie-Pier Côté (professeure à l’université Laval, à Québec), Olivier Côté (doctorant que je co-supervise avec Marie-Pier), Agathe Fernandes Machado (doctorante à l’UQAM), Ewen Gallic, (maître de conférences, à Marseille), François Hu (ancien stagiaire postdoctoral et aujourd’hui consultant en France), Philipp Ratz (ancien stagiaire postdoctoral et aujourd’hui consultant en Suisse), ainsi qu’Ana Patron Pinerez (ancienne stagiaire qui venait de Colombie) et Mulah Moriah (doctorant en France, que j’ai suivi pendant son mémoire de fin de maîtrise).
Exposé auprès des étudiants (et anciens étudiants) de l’AMA
Ce midi (heure de Montréal), je donnerais un exposé virtuel sur le thème « IA, biais et équité (en actuariat et en assurance) » (les slides sont disponibles via le lien). C’est un honneur de donner la première conférence de l’AMA, Association des Masters en Actuariat, qui sera diffusée en direct aux différentes promotions des formations de l’AMA, réunies dans leurs universités !
Comme me le disaient Chi et Samuel, “pour les anciens ou extérieurs qui veulent suivre sans être en présentiel, cela sera possible sur Briques2math“, la chaîne YouTube. Je préciserais juste que personne ne sera en présentiel, puisque je serais à Montréal, j’enseigne demain matin…
Ah oui, l’exposé sera en français… mais j’aurais des slides en anglais
Journée d’Économétrie Appliquée, à Montpellier (France), fin mai
Le 22 mai, la journée d’économétrie appliquée sera organisée à Montpellier. Je n’ai pas encore trouvé le site dédié, mais avec Ewen Gallic, on a été invités en tant que “keynote speakers”. A suivre donc
KNN and K-means in Gini Prametric Spaces
With Cassandra Mussard, intern last summer, and Stéphane Mussard, we uploaded a paper entitled KNN and K-means in Gini Prametric Spaces on ArXiv.
This paper introduces innovative enhancements to the K-means and K-nearest neighbors (KNN) algorithms based on the concept of Gini prametric spaces. Unlike traditional distance metrics, Gini-based measures incorporate both value-based and rank-based information, improving robustness to noise and outliers. The main contributions of this work include: proposing a Gini-based measure that captures both rank information and value distances; presenting a Gini K-means algorithm that is proven to converge and demonstrates resilience to noisy data; and introducing a Gini KNN method that performs competitively with state-of-the-art approaches such as Hassanat’s distance in noisy environments. Experimental evaluations on 14 datasets from the UCI repository demonstrate the superior performance and efficiency of Gini-based algorithms in clustering and classification tasks. This work opens new avenues for leveraging rank-based measures in machine learning and statistical analysis.
Beyond Human Intervention: Algorithmic Collusion through Multi-Agent Learning Strategies
Our paper, Beyond Human Intervention: Algorithmic Collusion through Multi-Agent Learning Strategies, with Suzie Grondin and Philipp Ratz is now available online
Collusion in market pricing is a concept associated with human actions to raise market prices through artificially limited supply. Recently, the idea of algorithmic collusion was put forward, where the human action in the pricing process is replaced by automated agents. Although experiments have shown that collusive market equilibria can be reached through such techniques, without the need for human intervention, many of the techniques developed remain susceptible to exploitation by other players, making them difficult to implement in practice. In this article, we explore a situation where an agent has a multi-objective strategy, and not only learns to unilaterally exploit market dynamics originating from other algorithmic agents, but also learns to model the behaviour of other agents directly. Our results show how common critiques about the viability of algorithmic collusion in real-life settings can be overcome through the usage of slightly more complex algorithms.
Optimal Transport on Categorical Data for Counterfactuals using Compositional Data and Dirichlet Transport
Our recent paper, Optimal Transport on Categorical Data for Counterfactuals using Compositional Data and Dirichlet Transport, with Agathe and Ewen is now online
Recently, optimal transport-based approaches have gained attention for deriving counterfactuals, e.g., to quantify algorithmic discrimination. However, in the general multivariate setting, these methods are often opaque and difficult to interpret. To address this, alternative methodologies have been proposed, using causal graphs combined with iterative quantile regressions (Plečko and Meinshausen (2020)) or sequential transport Fernandes Machado et al. (2025)) to examine fairness at the individual level, often referred to as “counterfactual fairness.” Despite these advancements, transporting categorical variables remains a significant challenge in practical applications with real datasets.
In this paper, we propose a novel approach to address this issue. Our method involves (1) converting categorical variables into compositional data and (2) transporting these compositions within the probabilistic simplex of \mathbb{R}^d. We demonstrate the applicability and effectiveness of this approach through an illustration on real-world data, and discuss limitations.
See https://github.com/fer-agathe/transport-simplex for the codes
Sequential fairness with multiple attributes, talk at the L^P seminar
François presented our joint paper “A Sequentially Fair Mechanism for Multiple Sensitive Attributes“, written with Philipp Ratz, at the L^P seminar(ISFA-CNAM-HEC Lausanne) today
Thanks François ! (more to come soon)
Assurance des catastrophes : les routes de l’enfer sont pavées de bonnes intentions
C’était le titre initial de l’article qu’on avait écrit avec Laurence Barry (co-titulaire de la chaire de recherche PARI – programme de recherche sur l’appréhension des risques et des incertitudes – placée sous l’égide de l’Institut Louis Bachelier en partenariat avec l’ENSAE/CREST et Sciences Po), la semaine dernière et qui a été publié sur le site du quotidien Le Monde,
Je mettrais l’article original en ligne plus tard, mais en attendant, je peux mettre en ligne un brouillon de la partie sur ce qui se passait en Californie (et qui présente l’avantage d’être en partie sourcé). Pour le cas français, je peux remettre un lien vers un court article qu’on avait écrit il y a quelques mois Rapport Langreney : lutter contre le désengagement des assureurs dans la couverture des risques climatiques
Continue reading Assurance des catastrophes : les routes de l’enfer sont pavées de bonnes intentions
The Insurance Market in the Era of Digital Transitions
A few months ago, we spent time with Raphaël Suire to write a short article on the insurance market, or more specifically, “The Insurance Market in the Era of Digital Transitions: Relationships Between Insurers, Big Tech, and Insurtechs“. The report is now available, on the webiste of the Society of Actuaries.
The digital revolution has profoundly transformed market dynamics, particularly within the insurance sector. This transformation encompasses the infrastructure and technologies that facilitate information exchange, the emergence of new business practices, a deluge of data, and the rise of innovative players capitalizing on these changes to deliver unique value propositions to customers. Traditional insurance companies face significant challenges and opportunities as they navigate competition from established Big Tech firms and agile insurtech startups. This study examines the disruptive nature of digital advancements, compelling historical players to confront the innovator’s dilemma (Christensen, 1997): whether to adapt and develop established practices or invest in new strategies to leverage digital opportunities. In doing so, they also come up against smaller, more agile start-ups. We highlight the necessity for insurance actors to rethink their roles in light of new market entrants and the evolving landscape shaped by Big Tech’s data monetization strategies. To analyze these dynamics, we propose an original framework in the form of a triangle of possibilities, which positions various market players and elucidates their strategic movements, innovations, and possible partnerships. This framework also aids in identifying competitive advantages and development trajectories, ultimately offering scenarios for the evolution of traditional insurance players in a digital and data-driven era.
A fair price to pay: Exploiting causal graphs for fairness in insurance
Our paper “A fair price to pay: Exploiting causal graphs for fairness in insurance“, with Olivier Côté and Marie-Pier Côté just appeared in the Journal of Risk and Insurance,
In many jurisdictions, insurance companies are prohibited from discriminating based on certain policyholder characteristics. Exclusion of prohibited variables from models prevents direct discrimination, but fails to address proxy discrimination, a phenomenon especially prevalent when powerful predictive algorithms are fed with an abundance of acceptable covariates. The lack of formal definition for key fairness concepts, in particular indirect discrimination, hinders effective fairness assessment. We review causal inference notions and introduce a causal graph tailored for fairness in insurance. Exploiting these, we discuss potential sources of bias, formally define direct and indirect discrimination, and study the theoretical properties of fairness methodologies. A novel categorization of fair methodologies into five families (best-estimate, unaware, aware, hyperaware, and corrective) is constructed based on their expected fairness properties. A comprehensive pedagogical example illustrates the implications of our findings: the interplay between our fair score families, group fairness criteria, and discrimination.
Le nouveau numéro de Risques
Le dernier numéro de l’Actuariel est paru
Tout nouveau tout chaud, le dernier numéro de l’Actuariel est paru, avec en particulier un article Open finance : Big bang annoncé dans l’assurance pour lequel j’avais eu long entretien. Au delà de quelques idées qu’on retrouve ici ou là, je peux mentionner une petite phrase, qui semble avoir retenu l’attention…
« Le débat est posé de manière sournoise en faisant croire aux citoyens qu’ils bénéficieront de produits plus personnalisés, sans rappeler que l’assurance est bien souvent un jeu à somme nulle et que si certains paient moins cher, cela signifie que d’autres paient plus » indique Arthur Charpentier, professeur à l’Université du Québec à Montréal et actuaire agrégé.
Time for a break 🎄
Optimal vaccination policy to prevent endemicity: a stochastic model
Our paper optimal vaccination policy to prevent endemicity: a stochastic model, written with Félix Foutel-Rodier and Hélène Guérin was just published in the Journal of Mathematical Biology
We examine here the effects of recurrent vaccination and waning immunity on the establishment of an endemic equilibrium in a population. An individual-based model that incorporates memory effects for transmission rate during infection and subsequent immunity, is introduced, considering stochasticity at the individual level. By letting the population size to go to infinity, we derive a set of equations describing the large scale behavior of the epidemic. The analysis of the model’s equilibria reveals a criterion for the existence of an endemic equilibrium, which depends on the rate of immunity loss and the distribution of time between booster doses. The outcome of a vaccination policy in this context is influenced by the efficiency of the vaccine in blocking transmissions and the distribution pattern of booster doses within the population. Strategies with evenly spaced booster shots at the individual level prove to be more effective in preventing disease spread compared to irregularly spaced boosters, as longer intervals without vaccination increase susceptibility and facilitate more efficient disease transmission. We provide an expression for the critical fraction of the population required to adhere to the vaccination policy in order to eradicate the disease, that resembles a well-known threshold for preventing an endemic state with an imperfect vaccine. We also investigate the consequences of unequal vaccine access in a population and prove that, under reasonable assumptions fair vaccine allocation is the optimal strategy to prevent endemicity.