Empirical Research on Ensuring Ethical AI in Fraud Detection of Insurance Claims: A Field Study of Dutch Insurers

The insurance industry in the Netherlands applies artificial intelligence (AI) in different processes and acknowledges that AI should be implemented in an ethical and responsible manner. Therefore, the Dutch Association of Insurers supported the industry by publishing an ethical framework. However, the framework is a set of high-level requirements, and the question is how these requirements are translated into local practices. Our research question is how ethical requirements are applied by insurance companies when using AI systems to detect fraud in insurance claims. To answer this question, we conducted interviews with representatives of four different organizations. The study demonstrates the awareness amongst interviewees that AI needs to be applied in a responsible way. The ethical framework provides a good starting point for insurers to develop their own practical ethical guidelines. Empirical evidence confirms that accountability, safety, transparency, non-discrimination, and human agency are priorities in the process of AI implementation. The research shows that translation of the ethical framework into operational and actionable instructions is done in-house by each organization and requires a multidisciplinary approach and cooperation between teams.