Category: DATA SCIENCE / AI - actuview - the international streaming platform for actuaries

Panel Discussion on AI Update – New Use Cases for Actuaries

Thu, 02 Oct 2025 08:29:16 +0000

AI is expanding the role of actuaries beyond traditional modeling into areas like automation, real-time analytics, and strategic insight. But as these tools grow more powerful, the human element becomes just as important. This panel will explore not only the emerging use cases for AI, but also how behavioral science and human decision-making influence its adoption. Could human nature itself limit how far AI can reshape the actuarial profession?

Multiple Yield Curve Modeling and Forecasting using Deep Learning

Tue, 30 Sep 2025 13:08:15 +0000

We introduce some deep learning models that simultaneously describe the dynamics of several yield curves. We aim to learn the dependence structure among the different yield curves induced by the globalization of financial markets and exploit it to produce more accurate forecasts. By combining the self-attention mechanism and nonparametric quantile regression, our model generates both point and interval forecasts of future yields. The architecture is designed to avoid quantile crossing issues affecting multiple quantile regression models. Numerical experiments conducted on two different datasets confirm the effectiveness of our approach. Finally, we explore potential extensions and enhancements by incorporating deep ensemble methods and transfer learning mechanisms.Read the paper here.

Fair Enough? Building Trustworthy and Equitable AI in Healthcare

Thu, 25 Sep 2025 15:54:20 +0000

AI is transforming healthcare, but hidden biases in data and algorithms can undermine trust and lead to unequal outcomes. This session explores fairness in medical AI, using real-world case studies and the FAIR-MED/XFAIR-MED frameworks to show how bias can be detected, explained, and addressed. Attendees will gain practical insights into building AI systems that are not perfect, but “fair enough” to support safe and equitable decision-making.

AFDP 2024: Property & Casualty I

Wed, 10 Sep 2025 10:34:38 +0000

In this session, participants will receive an introduction to the fundamentals of Property and Casualty (P&C) insurance. The module begins with an overview of the global insurance industry, highlighting the significance and market dynamics of P&C insurance. Key components such as underwriting, pricing, claims, and reserving will be explained, emphasizing their importance in effective insurance management. Basic definitions and methodologies for pricing and reserving will be introduced. The session concludes with an exploration of the unique challenges and growth opportunities in developing countries, illustrated with relevant case studies.

Better Healthcare and Lower Cost? Real Examples of Using AI to Improve the Delivery of Healthcare

Wed, 27 Aug 2025 17:30:43 +0000

As healthcare systems strive to deliver better outcomes at lower costs, artificial intelligence (AI) is proving to be a powerful enabler of transformation. This webinar, tailored for actuaries, showcases real-world applications of AI improving delivery, enhancing outcomes, and reducing costs. We will hear from Tanya Dillard Stark, SVP at CirrusMD, a pioneer in chat-first virtual care, and Stephen Bonner, CEO of SkinIO, whose AI platform enables early skin cancer detection via smartphone-based exams. Attendees will explore how AI is reshaping risk assessment, cost forecasting, and care navigation, with practical insights into the actuarial implications of these advancements. Join us to learn how actuaries can play a pivotal role in leveraging AI to build a more efficient, equitable healthcare system.

Incident-Specific Cyber Insurance

Fri, 22 Aug 2025 13:14:51 +0000

In this webinar presentation, our speakers will explore how cyber insurance products are structured to cover different types of incidents such as data breaches and ransomware attacks, each with their own limits and deductibles. They will also discuss how real-world data can help design these products to better serve both insurers and policyholders.

Generative AI for Actuaries

Mon, 18 Aug 2025 07:19:43 +0000

Explore how actuaries can harness the power of generative AI to boost efficiency, enhance insights, and drive innovation. This exclusive video offers a glimpse into the tools and trends shaping the profession’s future.

Find the report here.
This is just a taste of the expert-led content available through PD Edge Plus. Learn more here.

The Power of Tradition in the Age of AI

Mon, 11 Aug 2025 13:15:49 +0000

This short interview introduces ActuaryGPT, an AI-powered tool built to assist actuaries in their work. It outlines how automation can support routine processes, enhance accuracy, and give actuaries more time to apply their expertise in areas such as strategy, risk and communication. The conversation highlights how traditional actuarial skills like judgment, ethics and interpretation remain essential, even as the profession adapts to advances in data science and machine learning. It also recognises the importance of diverse perspectives in navigating complex problems and making better decisions in a fast-changing environment. A real-world example is included, where ActuaryGPT uses Monte Carlo simulation to estimate the ASX 200 index. While the result was highly accurate, the key message is that innovation is most effective when guided by human insight. The video reflects how the actuarial profession can evolve by combining long-standing principles with emerging tools and new ways of thinking.

IAAHS JoCo 2025 – Episode 7: E-Mergency Room: Predictive Modeling of Supplementary Healthcare Expenses using Machine Learning and Deep Learning Techniques

Fri, 08 Aug 2025 11:33:24 +0000

With the outbreak of Covid-19 pandemic, the Brazilian supplementary healthcare sector became a conducive environment for using complex data analysis and modeling tools. In this study, we apply different Machine and Deep Learning techniques (SVM, XGBoost and RNN) to predict healthcare expenses and evaluate if these techniques would present better performance in comparison to traditional ones, such as time series and regressions. Prediction scenarios were generated upon expense official databases between 2015-2022, considering two panoramas: (i) real, and; (ii) counterfactual, in which we assume the non-existence of the pandemic data for 2020. Using RMSE as the performance indicator, we find out that XGBoost model presented the best performance for the real panorama, with better fit in 32.2% of the scenarios. For the counterfactual panorama, we observe that RNN and SVM models obtained better fit in 22.3% of the cases. It is noteworthy that, until now, no studies were identified that address the use of predictive Machine and Deep Learning models into the Brazilian healthcare expenses. We also expect that this study offers insights for decisions made by the several players in this sector, such as operators and regulators, especially when it comes to pricing and development of healthcare products.

Time to know your insurance contracts

Fri, 08 Aug 2025 09:29:19 +0000

PoliSee is an AI-powered assistant designed to help individuals better understand, manage, and optimize their insurance coverage. In a world where most people juggle 3 to 5 insurance policies without fully grasping their rights, exclusions, or deductible. PoliSee brings clarity, personalization, and peace of mind. By analyzing all your contracts from life, auto, home, and travel to retirement plans PoliSee summarizes your coverage, detects unnecessary overlaps, and recommends adjustments tailored to your lifestyle, hobbies, family, and future plans. More than just a comparison tool, PoliSee supports users during claims, helps them choose the right policies, and anticipates their needs through smart recommendations. It also benefits insurers by reducing adverse selection and moral hazard through better education and risk awareness. Through real-life use cases from handling a broken windshield to planning international travel this video demonstrates how PoliSee transforms the insurance experience into something understandable, proactive, and user centric. PoliSee stands for transparency, trust, and better decision making. It reconnects policyholders and insurers around a shared goal: smarter, clearer protection.

Man or Machine?

Fri, 08 Aug 2025 08:59:32 +0000

This video explores the growing influence of artificial intelligence in decisions that affect people’s lives — from loans and insurance pricing to healthcare and education. Through real-world examples, it highlights how automated models can reproduce and amplify human biases, often without transparency or accountability.At the heart of the message is a challenge to the actuarial profession: when models become faster, more complex, and harder to question, who ensures that they remain fair?The narrative traces the deep roots of actuarial science — a tradition built on clarity, auditability, and ethical responsibility — and asks how those values can guide us through today’s evolving risk landscape.Rather than stopping at critique, it proposes a forward-looking idea: the development of Ethical AI Scorecards — frameworks that bring measurable standards of fairness, explainability, and transparency to AI-based risk decisions.Because in the end, while machines can calculate the risk… only humans can choose to carry it.

The Application of Actuary in Data-Driven World

Fri, 08 Aug 2025 08:47:51 +0000

The video reveals how big data is changing actuarial science. First, we go deeper — by truly understanding each individual.Second, we go broader —by covering risks in insurance which we thought uninsurable.The actuary of tomorrow is not a product assembler—but a designer of protection for unique lives. The world is beautiful because of its diversity, and our protection becomes truly meaningful when it is precise.

The Only Insurance Your Career Needs

Fri, 08 Aug 2025 08:32:47 +0000

What if your skills had an insurance policy? In a rapidly evolving AI-driven world, obsolescence isn’t fiction—it’s a financial risk. This TED-style talk introduces Skillset Insurance, a revolutionary actuarial concept that quantifies the risk of human skill becoming outdated, and creates a safety net to protect careers before it’s too late. Using formulas like Skill Decay Rate, Probability of Displacement, and Adaptive Upskilling, we calculate an individual’s Human Capital at Risk (HCaR)—a technical yet practical way to assess career vulnerability. The talk explains how this model can be used to design personalized insurance products that fund retraining, reduce turnover, and create a resilient workforce. Through storytelling, humor, and real actuarial math, we explore how this policy could become the future of employment benefits, transforming how companies value their employees—and how individuals take charge of their relevance. Whether you’re an actuary, a decision-maker, or simply someone worried about being replaced by a machine this session will not only inform, but inspire action.

Actuary vs Data Scientist: An Employers Dilemma

Fri, 08 Aug 2025 08:07:30 +0000

A young actuary turned data scientist explores the employer's dilemma: who to hire for risk modeling - actuary or data scientist? Same math, different mindsets.

Recovery Ready

Thu, 07 Aug 2025 14:43:02 +0000

Recover Ready is an innovative new project designed to help patients better understand the post-operative recovery process. It provides essential knowledge about what to expect after surgery and how to support their own recovery effectively. Patients can take short quizzes to reinforce their understanding and ensure they are prepared for each stage of recovery. By increasing awareness and engagement, Recover Ready empowers patients to take an active role in their healing journey—leading to better outcomes and a smoother recovery experience.

Operational AI Alignment

Thu, 07 Aug 2025 14:33:14 +0000

Today, businesses are already deploying AI systems that optimise the wrong things. There's a temptation to focus on the metrics that are easy to track, but completely misaligned with real goals. My submission is about "Operational AI Alignment", a key part of AI Governance which is about the coordination of teams, systems and resources within an organisation to ensure alignment of AI with business objectives. My video goes through a framework for approaching the problem and a range of useful tools that are practical and immediately applicable to current AI implementations. In my role as a consultant, I've worked with organisations implementing AI solutions across various functions and seen many of the same mistakes and oversights. Many of these include issues with bridging the gap between high-level strategy/business understanding and proper technical grasp of capability, risks and limitations of the technology... a gap Actuaries are used to working within!

PNL AI and quantum computing in actuarial sciences

Thu, 07 Aug 2025 14:15:13 +0000

Have you ever thought that it’s possible to identify personality traits based on the language someone uses? Well, there’s plenty of scientific evidence and research backing this up! But… what if we used it to predict risks? In this video, we explore how the intersection of Natural Language Processing (NLP), Artificial Intelligence, and Quantum Computing could redefine risk analysis in the insurance industry. We address four key questions:

How can we apply NLP in the insurance sector?
What tools should we develop?
What implementation risks should we anticipate?
And what could be a viable way forward?

We don’t just raise questions—we also propose solutions. From implementation strategies to risk management, we suggest a roadmap for building effective governance in the face of emerging technologies.

GPU’s for Actuarial Modeling, Hype or Hyper-drive?

Mon, 04 Aug 2025 11:14:08 +0000

In this presentation, attendees will investigate if GPU technology is all hype or if it will be bringing the actuarial modeling space into hyperdrive. Learn about the history of actuarial computing, why GPUs make sense for actuarial computations, what are the challenges and benefits of GPU’s and if GPUs truly are the answer, what are the options available for insurers.

DENK LAUT - der Podcast: FIT4AI, Episode X.1: Wir erklären KI-Begriffe

Tue, 05 Aug 2025 13:34:42 +0000

In dieser ersten Sonderausgabe der Reihe FIT4AI widmen wir uns der Klärung einiger grundlegender Begriffe des Themenbereichs Künstliche Intelligenz - und das auf ganz spezielle Art und Weise.Wir drehen das Glücksrad und erläutern was ist:

Künstliche Intelligenz,
Maschinelles Lernen,
Overfitting, Underfitting & Regularisierung,
Bias & Diskriminierung in KI-Systemen,
No-Free-Lunch Theorem,
Hyperparameter & Feature Engineering.

DENK LAUT – der Podcast: FIT4AI, Episode 4 – Generative KI: Wie sie funktioniert – und was sie (nicht) kann

Wed, 30 Jul 2025 15:33:27 +0000

Was steckt hinter Tools wie ChatGPT & Co? In dieser Folge sprechen wir über die technischen Grundlagen generativer KI: von Tokens und Transformern bis zu neuronalen Netzen.

Unsere Gäste, die Professoren Alexander Brandt und Thomas Kessel, bringen dabei nicht nur viel Fachwissen, sondern auch Erfahrungen aus der Hochschullehre mit. Wir diskutieren, wo generative KI sinnvoll eingesetzt werden kann, wo sie halluziniert – und wie man mit gutem Prompting mehr aus ihr herausholt.

Philosophisch wurde es auch: Zwischen Immanuel Kant und R2D2 war alles dabei.

Und hier gibt's auch den Link zum angesprochenen Fachbuch von Thomas Kessel, Alexander Brandt und Jonas Offtermatt "ChatGPT und Large Language Models? Frag doch einfach!": https://www.utb.de/doi/book/10.36198/9783838562766.

Time-penalised tree, un algorithme temporel interprétable appliqué aux risques climatiques

Wed, 30 Jul 2025 13:49:21 +0000

Cette présentation introduit un algorithme d'arbre de décision innovant conçu pour prendre en compte des variables temporelles dans des environnements dynamiques. Les méthodes traditionnelles échouent souvent lorsque les variables explicatives varient avec le temps, entraînant des prévisions inexactes. Notre solution, l'algorithme de Time-penalised Tree (TpT), utilise un critère de division pénalisé par le temps, permettant la partition récursive conjointe de l'espace des covariables et du temps. Cette approche intègre les tendances historiques dans la construction du modèle, offrant un cadre clair et interprétable.
Nous présentons la structure et le fonctionnement de l'algorithme TpT, en soulignant ses avantages par rapport aux méthodes existantes. Un cas d'étude sur la prédiction des risques climatiques illustre l'application pratique de TpT, montrant comment il améliore l'interprétabilité des prévisions en utilisant des données climatiques historiques. Puisque celles-ci forment des séries temporelles, dans l’espace et dans le temps, l’algorithme TpT aide également à la sélection des variables climatiques les plus prédictives.
Nous abordons également les propriétés théoriques de TpT, son efficacité et son potentiel d'application dans divers domaines comme la santé, la finance et l'assurance. Les directions futures de recherche sont discutées, notamment la validation et la comparaison de TpT avec d'autres algorithmes sur des ensembles de données actuariels divers.

Machine Learning en assurance vie, application à la valeur client

Wed, 30 Jul 2025 13:46:16 +0000

Dans un monde où les données sont devenues le nouvel or, l’intégration du machine learning dans le secteur de l’assurance vie ouvre des perspectives inédites. Cet atelier, destiné aux actuaires et data scientists, propose une exploration approfondie des techniques de machine learning appliquées à l’assurance vie, en particulier sur l’amélioration de la valeur client.
La valeur client est un indicateur important mesurant la rentabilité de chaque client. Elle est calculée comme la somme des résultats futurs probables réalisés par l’assureur sur chaque contrat des clients. Elle peut ainsi être mesurée par la VIF (Value In Force) liée aux encours en stock.
Le machine learning peut intervenir tout d’abord dans le processus d’individualisation de la VIF. En effet, des méthodes de clustering peuvent être utilisées pour rassembler les contrats en groupes homogènes en termes de Valeur Client, qui sera la maille utilisée dans les modèles ALM.
La valeur client est utilisée depuis plusieurs années chez Abeille Assurances, qui souhaitait toutefois intégrer une dimension plus prospective dans son calcul. Le Machine Learning permet ici une prédiction des flux futurs (rachats, arbitrages, reversements, etc.), pouvant ainsi être intégrés dans la valeur client.

Retour d’expérience sur la migration d’un modèle de projection de cash-flows dans un langage open source

Wed, 30 Jul 2025 13:44:53 +0000

Cette présentation est un retour d'expérience sur la migration d'un modèle de projection de cash flows depuis un progiciel vers R. Cette transition constitue l’occasion d’optimiser les coûts financiers et opérationnels liés à l’utilisation des modèles tout en améliorant la gouvernance et gagnant sur la maîtrise de la maintenance.
Sur les plans financier et opérationnel, la migration permet la réduction significative des coûts. Cela comprend les coûts de licence et de cloud en comparaison aux logiciels propriétaires de la place. La flexibilité et l'efficacité du langage R (refactorisation du code) facilitent sa modularité, sa maintenance et son développement au plus près des caractéristiques des produits ainsi que des évolutions de marché et de règlementation. Les performances calculatoires et la parallélisation permettent d’optimiser l’utilisation des ressources matérielles et humaines.
Sur le plan de la maintenance, l’utilisation de R ouvre vers une communauté de développeurs au-delà du périmètre de l’entreprise et des éditeurs usuels. L’entreprise bénéficie des dernières avancées techniques (librairies) et de l’accès facilité à la connaissance. Cela est renforcé lorsque l’entreprise oriente sa stratégie IT vers les infrastructures adaptées pour R.
Enfin, la gouvernance est facilitée lors de l’utilisation d’environnements tels que GitLab qui assurent la traçabilité et l’auditabilité des modèles sous réserve de la définition de politiques de modèles en interne.

Les Large Language Models pour comprendre et driver les pratiques en matière de durabilité

Wed, 30 Jul 2025 13:40:32 +0000

Cet atelier vise à présenter un retour d’expérience de l’application d’une solution d’intelligence artificielle fondée sur les Large Langage Models (LLM) à un dispositif réglementaire de durabilité (le dispositif « article 29 LEC »), sur différents cas d’usages : accélération de la recherche sémantique au sein des rapports, comparaison des acteurs, appréciation des évolutions.
Opéré par l’ADEME, le Climate Transparency Hub (CTH) est la plateforme règlementaire centralisée de dépôt des rapports réglementaires « article 29 LEC » des institutions financières, qui décrivent leurs pratiques en matière de durabilité : stratégies climats et biodiversité, expositions à la taxonomie européenne et aux énergies fossiles, … Ces rapports font l’objet d’une analyse annuelle par l’ADEME (cf. dernier rapport sur les remises 2023) qui combine résultats statistiques et mise en avant des pratiques.
L’information collectée est riche et hétérogène. Elle constitue donc un champ d’application potentiel particulièrement pertinent pour les outils d’intelligence artificielle, alors même que les attentes en matière de durabilité qui pèsent sur les acteurs financiers ne cessent de croître.

Exploitation de LLM pour l'extraction de graphes de connaissances causaux en assurance

Wed, 30 Jul 2025 13:28:34 +0000

L'identification de causes dans les documents sinistres est essentielle pour améliorer l'évaluation des risques et son atténuation. Cependant, la nature textuelle et la complexité des déclarations posent un défi car elles ne s'intègrent pas de manière directe dans les méthodes conventionnelles.
L’objectif de ces travaux est de concevoir des méthodes utilisant les grands modèles de langage (LLM) afin d’extraire des graphes connaissances causaux à partir des corpus documentaires liés à la gestion des sinistres.
D’abord, nous introduirons la méthodologie :
• Le processus qui vise à extraire des informations cruciales des documents, conduisant à la création d'un graphe de de connaissance causal.
• La méthode d'itération sur le graphe causal, évaluant la stabilité des résultats, et agrégeant les graphes de documents individuels en un méta graphe bayésien.
Nous présenterons ensuite deux cas d’usage, l'un consacré aux documents en assurance RC médicale et l'autre aux dossiers sinistres en assurance énergie. Nous reviendrons sur la préparation des données, la modélisation de la structure causale, l’évaluation et l'implémentation du système.
Enfin, nous comparerons les résultats obtenus aux techniques existantes. Nous mettrons en évidence des limites liées à la qualité, à la stabilité, au temps de traitement et à l'empreinte carbone de ces approches. Nous discuterons finalement de la manière dont ces méthodes peuvent contribuer aux sciences actuarielles.