Workshop on Fairness and Discrimination in Insurance

Overview

The Workshop on Fairness and Discrimination in Insurance is back for its third edition!

Given the rapid evolution of the field and the success of previous editions, the workshop will, for the first time, span two days: October 6-7, 2026. Researchers, regulators, actuaries, and legal scholars will gather to discuss fairness in insurance pricing and actuarial practice.

The workshop is supported by the Chair of Educational Leadership on big data analytics for actuarial sciences - Intact and is organized by professors Marie-Pier Côté (Laval University) and Arthur Charpentier (UQAM). It builds on two previous editions held in 2022 and 2024, the latter attracting more than 120 participants.

The program brings together academic, regulatory, legal, and operational perspectives on algorithmic discrimination and fair practices in insurance.

The Workshop on Fairness and Discrimination in Insurance will be offered in a hybrid format. Speakers will be present at Université Laval, offering in-person attendees the opportunity to engage with them and participate in the discussions. Remote participants via Zoom will only be able to attend the presentations. The majority of the talks will be presented in English.

Speakers

• Mathias Millberg Lindholm (keynote), associate professor, Stockholm University

• Mouloud Belbahri, ML scientist, TD Insurance

• Liz Bellefleur-MacCaul, senior data scientist - senior technical specialist, advanced analytics, Economical Insurance

• Arthur Charpentier, professor, UQAM

• Olivier Côté, PhD student, Université Laval

• Agathe Fernandes Machado, PhD student, Université du Québec à Montréal

• Charlotte Jamotton, postdoctoral fellow, UQAM

• Marie Michaelides, assistant professor, Heriot-Watt University

• Kathleen Miao, PhD student, University of Toronto

• Craig Sloss, enterprise analytics consultant, advanced analytics, Definity

• Stay tuned: other speakers will be announced soon!

Partners

• Actuarial and Financial Mathematics Laboratory

• Big Data Research Center, Université Laval

• Chair of fairness of predictive models: an application to insurance markets, Foundation for science

• Institute Intelligence and Data, Université Laval

• SCOR Corporate Foundation for Science

About the Speakers

Mathias Millberg Lindholm (keynote), associate professor, Stockholm University

Mathias Millberg Lindholm is an associate professor in mathematical statistics at the Department of Mathematics at Stockholm University. His research interest spans from applied probability to statistics and machine learning techniques, often with a connection to health or insurance applications. Prior to joining academia full time, he worked in the insurance industry as an actuary, with asset liability management, strategic asset allocation, and as a quantitative risk analyst.

Mouloud Belbahri, ML scientist, TD Insurance

Mouloud Belbahri is a Senior Research Machine Learning Scientist at TD Insurance, where he leads the AI and ML research program and collaborates with universities across Canada. He holds a Ph.D. in Statistics and conducts research on trustworthy AI, spanning fairness, causal inference, uncertainty quantification, and human-AI decision making. His work aims to bridge methodological rigor and practical impact in high-stakes real-world applications.

Liz Bellefleur-MacCaul, senior data scientist - senior technical specialist, advanced analytics, Economical Insurance

Liz Bellefleur-MacCaul (she/her) is a Senior Data Scientist at Definity Financial Corporation. With almost 10 years of experience, Liz has led the development of predictive models for both Personal and Commercial lines of business, as well as AI products that support call center, claims intake, and underwriting operations. In addition to her core role, Liz serves as the current Chair of Definity’s Bias & Fairness Subcommittee, leading the adoption of techniques to measure bias, assess fairness, and deliver bias-focused education that drives responsible predictive modelling and AI usage across the business. Outside of her role at Definity, Liz has presented on the topic of fairness at Casualty Actuarial Society meetings and webinars, FSRA Exchange, and the Canadian Automobile Insurance Rate Regulators (CARR) Conference. She is also a member of the CAS Canada Race and Insurance Pricing Task Force.

Arthur Charpentier, professor, UQAM

Biography to be provided.

Olivier Côté, PhD student, Université Laval

Olivier Côté is a Ph.D. candidate at the École d'actuariat of Université Laval, in Québec, under the supervision of Marie-Pier Côté and Arthur Charpentier. He holds a master's degree in actuarial science from Université Laval and is an Associate of the Casualty Actuarial Society. His research on fairness and discrimination in insurance pricing won the American Academy of Actuaries 2026 Award for Research. He has received numerous prestigious scholarships for his academic excellence and his leadership, among them the Hickman Scholarship of the Society of Actuaries and the doctoral scholarship of the Natural Sciences and Engineering Research Council of Canada. His thesis is carried out in collaboration with one of the largest insurance companies in Canada.

Agathe Fernandes Machado, PhD student, Université du Québec à Montréal

Agathe Fernandes Machado is starting the final year of her PhD in Mathematics at Université du Québec à Montréal (UQAM), under the supervision of Arthur Charpentier (UQAM) and Ewen Gallic (Aix-Marseille School of Economics). Her research focuses on Trustworthy Artificial Intelligence, with projects related to algorithmic fairness, the reliability of machine learning models, and interpretability methods. Prior to her PhD, she completed a double master’s degree in actuarial science and machine learning in France. Her master’s thesis in actuarial science focused on reinsurance and climate risks.

Charlotte Jamotton, postdoctoral fellow, UQAM

Charlotte Jamotton recently joined the Université du Québec à Montréal (UQAM) as a postdoctoral researcher. She earned her PhD in Actuarial Sciences from UCLouvain in Belgium. Her doctoral research focused on non-life insurance analytics. At the intersection of actuarial science and data science, her work explores how clustering techniques, Bayesian models, and fairness interventions can be applied to risk assessment and insurance pricing.

Marie Michaelides, assistant professor, Heriot-Watt University

Marie Michaelides is an Assistant Professor in the Department of Actuarial Mathematics and Statistics at Heriot-Watt University in Edinburgh, with broad research interests in dependence modelling, climate risk, and the application of machine learning methods to insurance. Her current work focuses on actuarial fairness and discrimination in insurance pricing and risk models. She holds a PhD in Mathematics from the Université du Québec à Montréal, alongside Masters’ degrees in Actuarial and Financial Engineering and Business Engineering from KU Leuven. She started her career as an actuarial consultant in Belgium, and she later held a postdoctoral position at Concordia University in Montréal before joining Heriot-Watt in 2025. She is a qualified actuary with the Institute of Actuaries in Belgium and an Associate of the Institute and Faculty of Actuaries (AFA).

Kathleen Miao, candidate au doctorat, University of Toronto

Kathleen Miao is a PhD candidate at the University of Toronto working under the supervision of Prof. Silvana Pesenti. Kathleen's research focuses on mathematical problems in insurance and risk management, dependence uncertainty, and robustness. Their work is supported by the Queen Elizabeth II/Reginald A. Blyth Graduate Scholarship in Science and Technology (Statistics). You can contact Kathleen at k.miao@mail.utoronto.ca.

Craig Sloss, enterprise analytics consultant, advanced analytics, Definity

Craig is an Enterprise Analytics Consultant at Definity Financial Corporation, where he has been a member of the Advanced Analytics team since 2014. Craig’s role involves providing technical guidance on model validation and governance to data scientists who work on predictive models supporting fraud detection, claims management, underwriting operations, and customer service interactions. As the founding chair of Definity’s Bias & Fairness Subcommittee, he has played a leading role in Definity’s efforts to operationalize bias testing for predictive models. Craig also actively supports initiatives to advance the actuarial profession, in roles such as Volunteer Chair of the CAS Canada Race and Insurance Pricing Task Force, and as Vice Chair of the CIA Property & Casualty Practice Committee.

Additional Information, Liz Bellefleur-MacCaul et Craig Sloss

• Liz & Craig have co-presented on the topic of “Bias, Fairness, and the Modeling Lifecycle” at the CAS Annual Meeting in 2023, which was also summarized in an Actuarial Review article. Their presentation at the Workshop on Fairness and Discrimination in Insurance will be a sequel to that presentation, discussing the trends that have emerged since that presentation.

• Craig & Liz, along with the other members of the Bias & Fairness subcommittee, received a Definity CEO award in 2025, the company’s highest level of recognition.

• Outside of their core roles, both Craig & Liz are deeply involved in Definity’s Inclusion, Diversity, Equity and Accessibility (IDEA) initiatives. Craig is the Vice Chair of the LGBTQ+ employee group, and Liz is the founding Chair of the Disability & Accessibility Advocacy employee group.

About the Talks

Things to consider when adjusting for discrimination and fairness in insurance, Mathias Millberg Lindholm

I will start by going through different notions of fairness in relation to proxy-discrimination. This includes both classical definitions from the algorithmic fairness literature, but also more recent contributions more directly targeting the insurance pricing setting. In order for this to be practically implementable we need to think about e.g. estimation, access to data, and how to measure discrimination. The talk will be non-technical.

Beyond Coverage: Utility, Disparate Impact, and Substantive Fairness in Conformal Prediction: Mouloud Belbahri

Conformal prediction has emerged as a principled framework for uncertainty quantification, offering statistical guarantees through calibrated prediction sets. Recent work has shown that these prediction sets can substantially improve human decision making by communicating uncertainty and providing plausible alternatives. However, improving decision quality is only part of the story. When prediction sets are used in human-AI decision pipelines, their benefits may not be distributed equally across demographic groups. Our empirical studies reveal that standard fairness objectives for conformal prediction, such as equalized coverage, can paradoxically increase disparities in downstream outcomes. These findings motivate a broader perspective on fairness: moving beyond procedural properties of prediction sets toward the substantive fairness of the decisions they enable. In this talk, I will present a research journey spanning controlled human experiments, theoretical analysis, and large-scale LLM-based evaluation. The central message is that fairness in uncertainty quantification should be assessed through its impact on downstream decision making, and that equalizing prediction-set size may be a more effective route to equitable outcomes than equalizing coverage.

Title and abstract forthcoming, Liz Bellefleur-MacCaul and Craig Sloss

Title and abstract forthcoming, Arthur Charpentier

The trilemma of fairness in insurance pricing, Olivier Côté

Abstract Forthcoming.

Assessing Counterfactual Fairness of Insurance Pricing via Sequential Transport, Agathe Fernandes Machado

Algorithmic fairness refers to a set of principles and techniques aimed at ensuring that the decisions produced by an algorithm are fair and non-discriminatory toward all users, regardless of personal characteristics such as gender, ethnicity, or other so-called sensitive attributes. Its assessment can be conducted at the individual level by focusing on a specific individual from a minority group and asking counterfactual questions such as: “What would this woman’s premium be if she were a man?” To evaluate the unfairness of a machine learning model, we adopt the notion of Counterfactual Fairness proposed by Kusner et al. (2017). We introduce a distributional framework for causal mediation analysis based on optimal transport (OT) and its sequential extension along a mediator Directed Acyclic Graph (DAG), in which the sensitive attribute corresponds to the treatment variable. Rather than relying on cross-world structural counterfactuals, we construct mediator counterfactuals in a mutatis mutandis sense: mediators are modified only as necessary to align an individual with the distribution under the alternative treatment, while respecting the causal dependencies among mediators. Sequential transport (ST) builds these counterfactuals by applying univariate or conditional OT maps following a topological order of the mediator DAG, and naturally extends to categorical mediators through adapted transport techniques on the probability simplex. We apply this method to assess the fairness of an insurance pricing algorithm on a real-world dataset.

FairGGPR: A multi-criteria fair Gaussian regressor for insurance pricing, Charlotte Jamotton

We study how multiple notions of fairness can be incorporated into a single Bayesian non-parametric regression framework for insurance pricing, with a focus on claim frequency modelling under a log-link. We consider a Generalized Gaussian Process Regression (GGPR) model for count data with risk exposure and introduce fairness interventions in its architecture. Specifically, we address notions of individual fairness by altering the kernel structure to control the similarity between policies (e.g., to mitigate omitted variable bias). We also address group-level fairness by enforcing demographic parity through constraints affecting the posterior. This modified GGPR architecture allows us to jointly enforce multiple fairness notions within a single probabilistic model. We empirically explore trade-offs with actuarial fairness, and how different fairness criteria interact when combined. The results highlight the importance of adopting a multi-criteria, context-aware approach to fairness in insurance pricing.

Balance and Fairness Through Multicalibration in Nonlife Insurance Pricing, Marie Michaelides

Autocalibration is known to be an important requirement for insurance premiums since it guarantees that premium income balances corresponding claims, on average, not only at portfolio level but also inside each group paying similar premiums. Also, fairness has become a major concern because unfair treatment may expose insurers to lawsuits or reputational damage. Translating fairness into conditional mean independence allows actuaries to combine autocalibration and fairness into the multicalibration concept. This paper studies the properties of multicalibration in an insurance context and proposes practical ways to implement it, through local regression or bias correction within groups including credibility adjustments. A case study based on motor insurance data illustrates the relevance of multicalibration in insurance pricing.

Discrimination-insensitive pricing, Kathleen Miao

Rendering fair prices for financial, credit, and insurance products is of ethical and regulatory interest. In many jurisdictions, discriminatory covariates, such as gender and ethnicity, are prohibited from use in pricing such instruments. In this work, we propose a discrimination-insensitive pricing framework, where we require the pricing principle to be insensitive to the (exogenously determined) protected covariates, that is the sensitivity of the pricing principle to the protected covariate is zero. We formulate and solve the optimisation problem that finds the nearest (in Kullback-Leibler (KL) divergence) "pricing'' measure to the real world probability, such that under this pricing measure the principle is discrimination-insensitive. We call the solution the discrimination-insensitive measure and provide conditions for its existence and uniqueness. In situations when there are more than one protected covariates, the discrimination-insensitive pricing measure might not exist, and we propose a two-step procedure. First, for each protected covariate separately, we find the measure under which the pricing principle becomes insensitivity to that covariate. Second we reconcile these measures through a constrained barycentre model. We provide a close-form solution to this problem and give conditions for existence and uniqueness of the constrained barycentre pricing measure. As an intermediary result, we prove the representation, existence, and uniqueness of the KL barycentre of general probability measures, which may be of independent interest. Finally, in a numerical illustration, we compare our discrimination-insensitive premia and the constrained barycentre pricing measure with recently proposed fair premia from the actuarial literature.