60 Large Language Model Prompts for Reinsurance Actuaries (Part 1)
By Dave Ingram
Reinsurance News, August 2024
Actuaries should be good at writing prompts for getting answers out of Large Language Models (LLMs) such as ChatGPT, Gemini or Claude. We are generally good at specifics and details. But sometimes we get caught up in the illusion that an LLM is almost human. LLMs are built to usually provide a plausible sounding answer. So, unlike the computer systems we have become used to working with that often give no answer at all when we make a mistake with our instructions, LLMs will give an answer that generally sounds good. Many will say that the LLM is hallucinating when they get a bad answer. But my experience has been that the so-called hallucinations are often due to poor prompts.
We are told that the best prompts will have five elements: Clear objective, specific context, detailed requirements, desired format, and relevance. Most people will only address the third item “requirements” when asking an LLM a question. And their questions are often vague. That could be because people are used to being somewhat vague in conversations with other people, expecting the other people in the conversation to understand the unspoken parts of what they are saying. An LLM will fill in the unspoken parts, but not necessarily with the same things as another person who often will have much in common with you.
So, let's see if we can go from vague requirements to detailed requirements. (In part 2, we will tackle the other four elements of a good prompt.) A prompt that includes clear detailed requirements provides all of the information necessary to get you the type of answer that you want. So, for example, you might ask, “Evaluate the impact of recent trends in mortality?” and you would get a response. A rather general response, to a rather general question. The first question on the list below gets a good deal more specific. It asks about “the pricing and structuring of reinsurance treaties for life insurance products.” Try it yourself and see the difference in the response.
Now, before the next time you try asking a question of an LLM, take a look at the following examples. See if you can use these examples to make your next prompt more of a detailed requirement.
Here are 30 examples of more detailed prompts that are targeted to questions that a reinsurance actuary might be asked to answer. Some of them can be improved to be more detailed. See if you can identify any of those.
- “Evaluate the impact of recent trends in mortality rates on the pricing and structuring of reinsurance treaties for life insurance products.”
- “Analyze the role of predictive analytics in improving mortality risk assessments for life reinsurance portfolios.”
- “Discuss the implications of advancements in medical technology and treatments on the long-term mortality risk for reinsurance agreements in life insurance.”
- “Describe the use of stochastic modeling techniques in projecting future mortality trends and their effects on reinsurance pricing for life insurance policies.”
- “Compare and contrast different reinsurance structures, such as quota share and surplus reinsurance, in managing mortality risk for a diverse portfolio of life insurance products.”
- “Analyze the benefits and challenges of bulk reinsurance transactions for managing legacy blocks of life insurance policies.”
- “Evaluate the impact of regulatory changes on the structuring and pricing of bulk reinsurance agreements for life insurance blocks.”
- “Discuss the role of due diligence in assessing the risks and opportunities associated with bulk reinsurance of life insurance blocks.”
- “Compare and contrast different methods for valuing large blocks of life insurance policies in bulk reinsurance transactions.”
- “Describe the use of financial modeling and stress testing in determining the appropriate terms for bulk reinsurance agreements of life insurance blocks.”
- “Analyze the financial and actuarial considerations involved in structuring bulk reinsurance transactions for blocks of annuities.”
- “Discuss the impact of interest rate fluctuations on the pricing and valuation of bulk reinsurance agreements for annuity blocks.”
- “Evaluate the role of longevity risk in the assessment and pricing of bulk reinsurance deals for annuity portfolios.”
- “Compare and contrast different reinsurance structures, such as quota share and excess of loss, for managing the risks associated with bulk annuity reinsurance.”
- “Describe the use of scenario analysis and stress testing in evaluating the financial stability of bulk reinsurance transactions for annuity blocks.”
- “Compare and contrast the use of aggregate and per-risk reinsurance in managing catastrophe exposures for property insurers.”
- “Describe the impact of climate change on the frequency and severity of natural disasters and the implications for reinsurance pricing.”
- “Identify potential external data sources that could enhance the accuracy of catastrophe modeling for reinsurance purposes.”
- “Explain the effect of regulatory changes on the structure and pricing of reinsurance contracts in the U.S.”
- “Discuss the benefits and challenges of using parametric insurance solutions in the reinsurance market.”
- “Provide an example of how predictive analytics can improve the accuracy of loss projections for reinsurance treaties.”
- “Describe the application of stochastic modeling in assessing the risk of large aggregate losses in reinsurance portfolios.”
- “Discuss the implications of social inflation on the pricing and structure of liability reinsurance contracts.”
- “Evaluate the role of alternative capital, such as insurance-linked securities, in the reinsurance market.”
- “Summarize the key risk factors that should be considered in pricing cyber reinsurance policies.”
- “Explain the use of excess of loss reinsurance in managing volatility for casualty insurers.”
- “Describe the process of structuring a reinsurance treaty for a new line of business in property and casualty insurance.”
- “Discuss the challenges and opportunities of pricing reinsurance products in a soft market versus a hard market.”
- “Discuss the use of geospatial data in enhancing the accuracy of catastrophe models for reinsurance purposes.”
- “Evaluate the effectiveness of multi-year reinsurance contracts in managing long-term risk for insurers.”
Here are 30 more prompts about other actuarial topics that would be of interest to a reinsurance actuary, categorized by general topics:
Pricing and Underwriting
- “Compare and contrast the use of generalized linear models (GLM) and machine learning algorithms in pricing auto insurance policies.”
- “Describe how catastrophe models can be integrated into the pricing of homeowner’s insurance in hurricane-prone areas.”
- “Identify potential external data sources that could enhance the accuracy of pricing models for commercial property insurance.”
- “Discuss the implications of social inflation on the pricing of liability insurance products.”
- “Evaluate the role of reinsurance in managing volatility and stabilizing pricing for property casualty insurers.”
- “Summarize the key risk factors that should be considered in pricing cyber insurance policies.”
- “Analyze the impact of climate change on the long-term pricing of flood insurance in the US.”
- “Discuss the benefits and challenges of using telematics data for pricing auto insurance.”
- “Describe the application of experience rating and schedule rating in commercial liability insurance pricing.”
Risk Management
- “Compare and contrast the use of Value at Risk (VaR) and Conditional Value at Risk (CVaR) in assessing the financial risks of an insurance portfolio.”
- “Describe the impact of regulatory capital requirements such as Solvency II on the risk management strategies of U.S. insurance companies.”
- “Identify potential external data sources that could enhance the accuracy of operational risk models for a large multi-line insurer.”
- “Explain the effect of emerging cybersecurity threats on the risk management practices of life insurers.”
- “Discuss the benefits and challenges of implementing an enterprise risk management (ERM) framework in a mid-sized property and casualty insurance company.”
- “Analyze the impact of climate change on the risk management strategies for coastal property insurance.”
- “Describe the application of scenario analysis in assessing the impact of macroeconomic changes on the reserve adequacy for a reinsurer.”
- “Evaluate the role of diversification in managing investment risks within an insurer’s portfolio.”
- “Summarize the key risk factors that should be considered in developing a pandemic risk management plan for a life insurer.”
- “Explain the use of economic capital models in quantifying and managing risks for a global insurance conglomerate.”
Reserving
- “Compare and contrast the use of chain-ladder and Bornhuetter-Ferguson methods in reserving for auto insurance claims in a highly competitive market.”
- “Describe the impact of changes in tort law on the accuracy of reserves for general liability insurance.”
- “Identify potential external data sources that could enhance the accuracy of reserve estimates for workers’ compensation insurance.”
- “Discuss the implications of social inflation on the adequacy of reserves for casualty insurance products.”
- “Evaluate the role of reinsurance in managing reserve volatility for catastrophic event exposures in property insurance.”
- “Summarize the key factors that should be considered in setting reserves for cyber liability insurance.”
- “Explain the use of discounting techniques in the reserving process for workers’ compensation insurance.”
- “Discuss the challenges and opportunities of reserving for P&C insurance products in a soft versus a hard market cycle.”
- “Analyze the impact of economic inflation on the development patterns of claims reserves for commercial auto insurance.”
Big Data and Advanced Analytics
- “Explain the impact of big data and advanced analytics on the future of pricing in the property casualty insurance industry.”
- “Evaluate the role of reinsurance in managing mortality risk and stabilizing pricing for life insurers.”
And don’t worry if you do not get it right the first time. Iterating by asking a series of questions that get you closer and closer to the answer you are looking for is a valid and useful technique.
Crafting effective prompts for LLMs like ChatGPT will become a crucial skill for reinsurance actuaries. By moving from vague to detailed requirements, actuaries will find that they get more precise and valuable answers from LLMs. This shift not only maximizes the utility of these powerful tools but also aligns with the actuarial profession’s emphasis on specificity and detail. The 60 prompts show examples of how detailed prompts can transform the quality of responses, ultimately enhancing the decision-making process in reinsurance. As you explore and refine your prompts, remember that the clarity and specificity of your questions directly impact the relevance and accuracy of the answers you receive. This approach will be further expanded in Part 2, where we will delve into the other elements of effective prompting.
In conclusion, it's important to note the caveats associated with prompt engineering. Effective prompting requires detailed knowledge of the subject matter to avoid misleading responses. Using zero-shot prompting, where no prior information is provided, can result in significant hallucination if the model lacks sufficient training data on the topic. Given the potential financial risks of inaccurate information, it's crucial to approach LLM outputs with caution. A safer method is multi-shot prompting, where relevant documents are incorporated into the prompt, thereby reducing hallucination risk and improving accuracy. This careful approach ensures that the use of LLMs remains a valuable tool without compromising the reliability of the information generated.
Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries, the newsletter editors, or the respective authors’ employers.
Dave Ingram, FSA, CERA, is an ERM advisor and a Society of Actuaries Board Member (2021 – 2024). Dave can be contacted at daveingram@optonline.net.