Research report ·
Co-author
Climate Scenarios to Actuarial Assumptions: A Practitioner's Framework
Society of Actuaries Research Institute. Forthcoming — expected April 2027.
follow
Selected work
Research, applications, and actuarial education
I develop practical resources at the intersection of actuarial science, risk, data science, machine learning, and climate-related uncertainty. This collection brings together my published work, interactive applications, research projects, presentations, and teaching materials.
Evidence and inquiry
Research & professional contributions
Selected research, essays, and professional contributions.
follow
Article · Author · 2026
When a Good GLM Looks Bad: A Lesson in Calibration, Exposure and Actuarial Judgment
Emerging Topics Community Newsletter ,
Society of Actuaries, October 2026.
Incorrect exposure definitions can make a sound GLM appear poorly calibrated, highlighting the importance of actuarial judgment and careful data validation.
Research report ·
Co-author · 2026
Quantifying Mortality and Morbidity Impacts from Climate Risks: A Practical Framework for Decision-Making
Carlos Arocha and Jérôme Crugnola-Humbert. Society of Actuaries Research Institute, August 2026.
Thesis project ·
Proposed
Bayesian Climate–Health Models and Liability Uncertainty
Proposed student research project · 2026–2027.
Exploring Bayesian hierarchical methods to improve the estimation of climate-related mortality and morbidity sensitivities, particularly where data are limited. The project aims to quantify how uncertainty in these estimates propagates into insurance liability projections under alternative climate scenarios.
Building on the 2026 SOA Research Institute framework, the proposed work envisages a reproducible Python prototype, model-validation analysis, and illustrative liability stress tests.
follow
Roundtable · Contributor · 2026
Virtual Roundtable: AI in Actuarial Work
Actuarial Intelligence Bulletin, Society of Actuaries Research Institute, July 2026.
Perspectives on responsible AI adoption, actuarial judgment, accountability, and the use of AI to strengthen interpretation and decision-making.
Essay · Author · 2026
A Minimum Viable Climate Integration Roadmap for Insurers
Society of Actuaries Research Institute, March 2026.
Article · Author · 2026
Navigating the AI Transformation in Actuarial Science: Opportunities, Risks and the New Professional Landscape
Career Development Community Newsletter, Society of Actuaries, January 2026.
An examination of how AI and automation are reshaping actuarial work, from emerging applications and strategic opportunities to model risk, governance, ethics, and the skills actuaries will need to remain effective in an AI-enabled profession.
Article · Co-author · 2025
Preserving and Expanding the Role of Insurance in a Fragmented Global Economy: Insights for Actuaries
With Kai-Uwe Schanz. International News, Society of Actuaries, July 2025.
An examination of how geoeconomic fragmentation is reshaping insurance risks and what this means for actuarial practice, including pricing, risk management, capital allocation, scenario analysis, and long-term forecasting.
Research report · Contributor · 2023
Guide to Domestic and International Paid Family and Medical Leave Programs
Paul Correia, FSA, MAAA. Society of Actuaries Research Institute, June 2023.
Contributed input to the research project comparing mandated paid family and medical leave programs in the United States with programs internationally.
Tools for practice
Actuarial applications
Interactive tools that translate actuarial and statistical methods into usable workflows.
Feedforward Neural Networks for Disability Income Pricing
A Columbia University project that applies feedforward neural networks to price a disability income insurance policy.
Example: reserving · pricing · capital · climate risk
Gradient Boosting for Risk Assessment
A Columbia University project that applies gradient boosting to price an auto insurance policy.
Example: simulation · forecasting · portfolio analytics
Long-form learning
Online books
Three online books designed as a unified actuarial learning series.
FutureProof P
An online learning resource designed to prepare students for the Society of Actuaries Exam P, while also providing a complete foundation for a one-semester university course in basic probability.
Carlos Arocha · Online edition · Continuously updated
FutureProof FM
An online learning resource designed to prepare students for the Society of Actuaries Exam FM, while also providing a complete foundation for a one-semester university course in basic financial mathematics.
Carlos Arocha · Online edition · Continuously updated
FutureProof MLC
An online learning resource for life contingencies designed to provide a complete foundation for a one-semester university course.
Carlos Arocha · Online edition · Continuously updated
Talks and workshops
Presentations
Selected talks, workshops, and webinars. More entries are being added.
Talk · 2026
Online
Feedforward Neural Networks for Disability Income Pricing
Mid Atlantic Actuarial Meeting, November 2026. Forthcoming.
This session compares generalized linear models (GLMs) and feed-forward neural networks (FFNNs) from the perspective of practicing actuaries who must deliver models that are accurate, stable, explainable, and auditable. Using a pricing framework, we examine what each approach does well, where each can fail, and how to evaluate them in ways that align with actuarial governance (calibration, stability, tail behavior, and portfolio O/E). We will use a practical case study with high-cardinality categorical variables (e.g., region/occupation/employer) and climate-hazard covariates (heat and smoke proxies) to illustrate common real-world challenges: rare-event incidence, heavy-tailed severities, and drift over time. A key theme is that “better AUC” is not enough—pricing depends on probability calibration, segment adequacy, and stable performance across rolling time windows. We will also discuss hybrid deployment patterns (e.g., GLM frequency combined with FFNN severity) that often deliver the best balance of lift and governability. The goal is how to test, decide, and implement models that “actually work” under operational and regulatory constraints.
follow
Talk · 2026
Online
Quantifying Mortality and Morbidity Impacts from Climate Risks
Mid Atlantic Actuarial Meeting, November 2026. Forthcoming.
This session examines how climate change can affect mortality and morbidity and what this means for actuarial work in life and health insurance. Drawing on recent Society of Actuaries Research Institute research, a forward-looking, scenario-based framework linking climate hazards, population exposure, and vulnerability to actuarial outcomes is presented.
follow
Keynote Talk · 2026
Zurich, Switzerland
Impact of Climate Change on Life & Health
With Jérôme Crugnola-Humbert. General Assembly, Swiss Association of Actuaries, 04 September 2026.
The session explores why historical data alone may be insufficient in a non-stationary climate environment and shows how climate-related variables can be incorporated into mortality and morbidity modelling. A climate-adjusted Lee-Carter framework is introduced, together with a case study assessing heat-related mortality impacts in France, Italy, and Spain under a high-emissions scenario.
Talk · 2026
Segovia, Spain
Translating Physical Climate Hazards into Life & Health Assumptions
Climate Change and Insurance Conference. 02 September 2026.
This session explores how physical climate hazards can be translated into actuarial assumptions for life and health insurance. It focuses on the practical challenge that mortality and morbidity projections already contain implicit assumptions about future climate exposure, population vulnerability, and adaptation—and that these assumptions need to be made explicit if they are to be challenged, governed, and monitored.
Podcast · 2026
Quantifying Mortality and Morbidity Impacts from Climate Risks: A Practical Framework for Decision-Making
With Dale Hall and Jérôme Crugnola-Humbert. Research Insights Podcast, Society of Actuaries. 01 September 2026.
How can actuaries assess mortality and morbidity risk when historical experience may no longer fully reflect a changing climate environment?
In this episode of the Research Insights Podcast, Dale Hall, Managing Director of Research at the Society of Actuaries Research Institute,
welcomes Carlos Arocha, FSA, SCR, and Jérôme Crugnola-Humbert, SAV, IA, authors of the report Quantifying Mortality and Morbidity Impacts
from Climate Risks: A Practical Framework for Decision-Making.
Carlos and Jérôme discuss how climate hazards, population exposure and
vulnerability can translate into life and health outcomes and ultimately affect actuarial decision-making. They explore why historical
mortality and morbidity data alone may be insufficient in a changing risk environment, how climate-related variables can be incorporated into
mortality modeling, and what a case study examining France, Italy and Spain can teach actuaries about interpreting forward-looking scenarios.
The conversation also examines the “streetlight effect” and the “quantitative fallacy,” along with practical considerations for life and
health insurers seeking to better understand and prepare for climate-related risks.
Listen to the episode and explore the full report,
accompanying actuarial toolkit, and additional research from the Society of Actuaries Research Institute by visiting SOA.org and navigating
to the Catastrophe & Climate research area.
Read the Report: https://www.soa.org/resources/research-reports/2026/climate-mortality-morbidity-framework/
Talk · 2026
Online
Actuarial Perspectives on Climate Risk
With Jérome Crugnola-Humbert. Actuarial Education & Research Monthly Call. 30 June 2026
This session provides an actuarial perspective on climate risk, with a particular focus on mortality and morbidity in life and health insurance. It examines why climate change should increasingly be viewed not only as a sustainability concern, but also as a socio-economic and financial risk that can affect human health, insurance portfolios, and actuarial assumptions.
Workshop · 2026
Hannover, Germany
Practical Integration of Feed-Forward Neural Networks into Core Actuarial Workflows
Insurance Data Science Conference. 09 June 2026
This session examines how feed-forward neural networks (FFNNs) can be integrated into core actuarial workflows without abandoning the transparency, stability, and governance strengths of traditional GLMs. Using a synthetic dataset for an insurance pricing case study, the presentation compares GLM and FFNN approaches within a frequency–severity framework, with a focus on practical model adoption rather than model replacement.
Talk · 2026
Online
Reportes de riesgo climático para aseguradoras
INTECAP Guatemala. 2 June 2026
Esta sesión examina cómo los informes de riesgo climático para aseguradoras están evolucionando desde un ejercicio de divulgación hacia una herramienta de resiliencia y gestión. El punto de partida es una pregunta sencilla pero exigente: ¿entendemos el riesgo climático y podemos demostrar que lo estamos gestionando? La presentación aborda cómo ese riesgo afecta la suscripción, los siniestros, las inversiones, la solvencia, las operaciones y la conducta de mercado.
Talk · 2026
Online
Defensible AI: Actuarial Model Governance and Validation.
With Amit Jiwani. AI Insights for Actuaries. 14 May 2026
This session examines what it means for actuarial AI to be defensible in practice. It begins from the premise that stronger predictive performance is not enough: insurers increasingly need evidence of appropriate governance, validation, documentation, human oversight, and operational resilience. The session places this in the context of the EU AI Act, DORA, and the growing supervisory focus on how AI-supported insurance decisions are controlled.
Talk · 2026
Online
A Day in the Life of an AI-Assisted Actuarial Team
National Institute of Standards and Technology (NIST) AI Consortium.
28 May 2026.
This session offers a practical look at how a small actuarial team can use AI throughout a typical working day without losing control of the underlying actuarial work. Framed around a pricing-model refresh, the presentation follows a team developing a GLM, testing a feed-forward neural network, automating Python tasks, and preparing stakeholder communications. The central question is not whether AI should be allowed, but how it can be used while preserving professional judgment, transparency, and accountability.
Talk · 2025
Seattle, WA
Data-Driven Leadership: Navigating Climate Risk in Actuarial Work
With Brian Hinkle and Daniel Pribe. Society of Actuaries ImpACT Conference. 03 November 2025.
This session explores how actuaries can incorporate climate risk into practical decision-making when data are incomplete, uncertain, and highly scenario-dependent. It combines qualitative risk assessment with quantitative approaches, illustrating how climate variables can be integrated into actuarial models for mortality, morbidity, pricing, reserving, ORSA, and enterprise risk management. The presentation also examines a range of modelling techniques—including GAMs, Bayesian networks, survival models, and gradient boosting—and emphasizes the importance of interpretability, calibration, scenario analysis, and governance when translating climate signals into actionable actuarial insights.
Talk · 2025
Online
AI Under Regulation: The EU Sets the Standard
Society of Actuaries Emerging Topics AI Subgroup Call. 04 September 2025.
This session explores two practical dimensions of AI adoption for actuaries: understanding the emerging regulatory framework and using large language models effectively in model-building workflows. It summarizes key prohibited practices under the EU AI Act and then demonstrates how LLMs can support the development of auditable Excel models through clear problem definition, structured prompting, iterative refinement, and user validation. A case study on economic scenario generation illustrates how actuaries can use LLMs as collaborative tools to improve efficiency, documentation, and understanding while retaining control over model logic and outputs.
Learning resources
Teaching materials
Course notes, notebooks, exercises, and other reusable learning resources. More entries are being added.
Selected teaching materials will be added here as they are prepared for public release.
Collaboration
Interested in this work?
I welcome conversations about actuarial innovation, data science, climate risk, applied research, and quantitative education.