Presenter(s): Susan Kent, FCAS, MAAA, MS; Meagan Mirkovich, FCAS; Geoffrey Werner, BS, FCAS, MAAA; Jared Smollik; Margo P. Mackenzie, FCAS

Description: Auto insurance affordability has become an increasing concern as claim costs have risen faster than household incomes. This session reviews the key drivers of personal auto insurance costs, including inflation in parts and medical expenses, vehicle technology, miles driven, legal and fraud dynamics, electric vehicles, and emerging factors such as tariffs. Speakers will also discuss forces that may help reduce losses over time, including ADAS, safety technology, telematics, and policy interventions, providing an actuarial perspective on recent trends and potential paths forward.

Presenter(s): Mason Spitz, FCAS, MAAA; Michael Henk, FCAS, MAAA; Kayla M. Robertson, FCAS MAAA

Description: Reserving surprises often hide below the surface. This session shows how systematic search techniques can help actuaries prioritize reviews, validate intuition, and uncover hidden risks. Learn how these tools fit into existing workflows, support professional standards, and sharpen focus - all without replacing the central role of actuarial judgment.

Presenter(s): Ronald T. Kozlowski, FCAS, MAAA; Roosevelt Mosley, FCAS, MAAA, CSPA; Mallika Bender, FCAS, MAAA; Tyson Mohr; Mike McKenney

Description: Actuarial pricing approaches aim to produce rates that are fair and objective. But how should our methods adapt, if at all, when concerns about systemic and algorithmic bias introduce conflicting definitions of "fairness"? Over the past five years, this question has become a central topic in regulatory, consumer, and actuarial conversations. This panel will trace the evolution of the fairness debate—from longstanding rating concerns to recent state‑level activity, consumer‑advocacy engagement, and industry response. Panelists will examine how different stakeholders have approached concerns around algorithmic bias, new data sources, and predictive models.

Presenter(s): Kristan H. McGraw, ACAS, MAAA; Paul Kutter, PhD, FCAS, MAAA, CSPA; Alec J. Fisher, PhD, CSPA

Description: This session shows how we use a large language model (LLM) to extract information from claims notes in raw text form and compares the performance of different models on real data.. The session also locates this application within the larger framework of actuarial standards and risk-based enterprise AI governance. The first segment establishes the framework; a live poll asks participants to apply the framework to different applications. The second segment gives more technical details of the application including data exploration, prompt engineering, implementing retrieval-augmented generation (RAG), and model performance evaluation. Attendees will also take part in a brief “Human vs. LLM” reasoning exercise, comparing their interpretation of a claims note to the model’s response, reinforced by a handout summarizing governance and evaluation frameworks.

Presenter(s): YiFan Zhou; Mark Ma, CPCU

Description: Building robust insurance business models requires more than advanced algorithms—it demands vigilance against hidden data pitfalls. This presentation explores common traps such as temporal infidelity, Simpson’s paradox, confounding and omitted variable bias, sample bias, and more. Drawing on real-world examples from various insurance lines, we’ll reveal how these issues often arise in automated machine learning workflows and why overlooking them can mislead decision-making and harm company performance. Attendees will gain practical insights to identify, mitigate, and prevent these biases, ensuring models deliver accurate, trustworthy results.