# AI, insurance, discrimination and unfair differentiation. An overview and research agenda

**Authors:** Marvin S. L. van Bekkum, Frederik J. Zuiderveen Borgesius  
**arXiv:** [2401.11892](https://arxiv.org/abs/2401.11892)  
**Date:** 2024-01-22  
**Summary page:** https://www.summarizepaper.com/en/paper/2401.11892v1/

## Key points

- Growing use of artificial intelligence (AI) technology in the insurance field
- Two specific situations where insurers employ AI: data-intensive underwriting and behavior-based insurance
- Data-intensive underwriting involves using AI for data analysis to assess risks and adjust insurance premiums accordingly
- Behavior-based insurance utilizes AI to monitor consumers' real-time behavior and offers discounts based on adherence to certain criteria
- Advancements in AI offer benefits for insurers and policyholders but raise concerns about potential discriminatory effects
- Focus of the paper is examining discrimination-related effects from data-intensive underwriting and behavior-based insurance
- Two types of discrimination-related effects considered: discrimination itself and unfair differentiation
- Four factors introduced to assess fairness of insurance practices: transparency, accuracy, relevance, and proportionality
- Paper draws upon literature from various disciplines to develop a comprehensive understanding of implications associated with AI implementation in the insurance industry
- Research aims to contribute towards shaping policies that promote fairness and equality in AI-driven insurance practices.

## Summary

In the field of insurance, there is a growing use of artificial intelligence (AI) technology. This paper delves into two specific situations where insurers employ AI: data-intensive underwriting and behavior-based insurance. Data-intensive underwriting involves utilizing AI for data analysis to assess risks and adjust insurance premiums accordingly. On the other hand, behavior-based insurance utilizes AI to monitor consumers' real-time behavior, such as safe driving habits, and offers discounts based on their adherence to certain criteria. While these advancements in AI offer numerous benefits for insurers and policyholders alike, they also raise concerns regarding potential discriminatory effects. This paper focuses on examining the discrimination-related effects that may arise from the use of data-intensive underwriting and behavior-based insurance. There are two types of discrimination-related effects considered: discrimination itself and unfair differentiation. Discrimination refers to situations where certain groups protected by non-discrimination laws, such as individuals belonging to specific ethnicities, are harmed by the practices employed by insurers. Unfair differentiation, on the other hand, does not necessarily violate non-discrimination laws but is perceived as unjust or inequitable. To assess the fairness of insurance practices in relation to discrimination and unfair differentiation, four factors are introduced in this research: transparency, accuracy, relevance, and proportionality. These factors provide a framework for evaluating whether the use of AI in insurance leads to fair outcomes for all individuals involved. The paper draws upon literature from various disciplines including law, philosophy, and computer science to develop a comprehensive understanding of the implications associated with AI implementation in the insurance industry. By addressing these issues surrounding discrimination and unfair differentiation within AI-driven insurance practices,this research aims to contribute towards shaping policies that promote fairness and equality in this evolving field.