Overview
From underwriting and claims processing to customer service, actuarial modeling, risk assessment, and legal compliance, Generative Artificial Intelligence (GenAI) technologies are introducing new capabilities. This report provides both a practical guide and an educational handbook in understanding Generative Artificial Intelligence (GenAI).
Key Findings
Large-language models function as "black boxes" that transform inputs to outputs through learned patterns distributed across billions of numerical weights. For actuaries, these LLM "black-boxes" create a fundamental professional challenge to meet the actuarial standards of practice and regulatory frameworks required to explain their methods, justify their assumptions, and communicate their findings. This report examines ethical and regulatory frameworks; existing standards and voluntary guidance framework; checklists for governance and ethics in AI; key concepts on AI systems and their validation, along with case studies. It provides an understanding on the necessary validation processes to ensure safety and governance within the framework.
Acknowledgements
The SOA would like to thank the members of the Ethics for GenAI Models Project Oversight Group for their support, guidance, direction, and feedback throughout the project:
- Samuel Baker, FSA, FRM
- Robert Gomez, FSA, MAAA, CERA
- Chris Lombardi, FSA, CERA, MAAA
- Michael Niemerg, FSA, MAAA
- Gunjan Saxena, ASA
- Mark Sayre, FSA, CERA
- Tina Yang, FSA, CERA, MAAA
- Noman Zafar, FSA, FSAI, IFRI
- Yifan Zhang, FSA
At the Society of Actuaries Research Institute:
- Joe Alaimo, ASA, ACIA
- Korrel Crawford, Sr. Research Administrator
FAQ
The integration of Generative AI is creating new paradigms of human-AI collaboration that demand careful consideration. Actuarial use cases involve human-led, agent support and agent-led, human support.