We aim to support innovation across the insurance market players and contribute to actuarial science by co-constructing cutting-edge projects with partners, delivering training programmes, and sharing our knowledge through our publications and open-source tools.
Senior Expert
Expert
Expert
How can you reconcile the predictive performance of your pricing algorithms with growing fairness requirements and strict regulatory constraints (AI Act, GDPR), particularly when these require the elimination of algorithmic discrimination?
Rather than correcting biases once the model has been built, we integrate fairness constraints directly into the loss function during algorithm training. Our two tree-based “in-processing” learning methods enable neural network models to simultaneously optimize actuarial accuracy while complying with non-discrimination criteria.
Secure your AI deployments against regulatory scrutiny without compromising technical profitability. Benefit from an inherently ethical, transparent and auditable model, turning a compliance requirement into a strong competitive advantage in the insurance market.
How can you assess and quantify the accumulation risk arising from interconnected cyber losses — such as a global cyberattack or systemic failure — that could simultaneously affect a large number of companies within your portfolio and threaten the insurer’s solvency?
We develop advanced stochastic models capable of mapping how shocks propagate across networks of interconnected companies. This approach makes it possible to simulate realistic catastrophe scenarios while overcoming the lack of historical data through structural modelling of dependencies.
Assess the resilience of your portfolio and optimize your reinsurance programmes with a robust, objective and auditable view of accumulation risks.
How can you calibrate rigorous stress scenarios in response to the extreme concentration of risks among a small number of major cloud service providers (AWS, Microsoft Azure, Google Cloud), while accurately measuring the actual diversification effect within your portfolio?
We have developed a granular calibration methodology (link to “Cloud Failure and Cyber Insurance: Calibration of Stress Scenarios and Diversification”) to simulate the cascading impact of a major cloud infrastructure outage. Our approach combines the topology of technological failures with quantitative exposure analysis to accurately assess the limits of diversification.
Anticipate the domino effect of a major technological outage and fine-tune your underwriting strategy by identifying hidden areas of overexposure resulting from policyholders sharing the same IT dependencies.
How can you move beyond overly homogeneous national mortality tables to refine the pricing and reserving of life insurance portfolios in response to significant regional disparities in longevity?
We use penalized B-splines (P-splines) to model and continuously smooth mortality rates at the sub-national level. This robust statistical approach simultaneously captures age dynamics and geographical variations while overcoming the statistical noise caused by limited data availability for smaller populations.
Gain a more granular understanding of longevity risk and optimize your capital requirements by aligning your pricing and reserving models more closely with the regional demographic realities of your policyholders.
How can you overcome the limitations of traditional aggregate methods such as Chain Ladder — which fail to account for the individual characteristics of each claim and changes in claims management practices — to accurately assess ultimate claim costs and one-year risk within a strict regulatory framework (Solvency II, IFRS 17)?
We support actuarial teams in the strategic transition from aggregate triangles to individual loss reserving by modelling the claims lifecycle through semi-Markov multi-state processes.
This modelling framework captures the dynamic trajectory of each claim — opening, partial payments, reopening and closure — while accounting for the time already spent in each state.
Combined with an innovative matrix-based framework, our methodology provides exact analytical calculations — without relying on computationally intensive simulations — of both ultimate claim costs and the volatility of the Claims Development Result (CDR) over a one-year horizon.
Turn initial reserving into a precision-driven process from the moment a claim is opened. By leveraging the richness of individual claims data, you can eliminate the masking effects inherent in triangle-based methods and replace average assumptions with tailored estimates.
The model immediately identifies severe claims and reduces the subjectivity associated with claims handlers’ assessments, ensuring that initial reserves are accurately calibrated, predictive and consistent across your entire portfolio.
UCLouvain (BE)
ULB (BE)
UCLouvain (BE)
ENSAE (FR)
ISFA (FR)
