Abstract
An insurance claim decision requires claim handlers to understand the circumstances that give rise to the claim, knowledge of procedural rules and regulations, and critical reading and in-depth analysis of evidence from multiple information sources. Insurance claim processing is demanding for the handler and can be intimidating for the defendant and claimant. Many insurance companies have started to leverage the potential of AI to automate repetitive tasks and augment the cognitive capability to provide efficient and trustworthy decisions for improved customer experience.
This paper presents a methodology to capture and transform the claim handlers knowledge into the degree of belief for a set of decision-making rules. It is a transparent hybrid probabilistic expert system; a decision can be explained by the importance of the rules, weight of attributes, and the belief-degree in a decision inferred from the rules that are activated by the information of a given claim. The transparency of the decision-making system engenders trust in computer-aided decision making. Historical data may not contain rare claim circumstances. Therefore, a human expert can be leveraged to make decisions for such claims, which are then added to the training data for future machine learning from humans. This framework allows human experts and an AI system to work in partnership to enhance each other’s capabilities.
This paper presents a methodology to capture and transform the claim handlers knowledge into the degree of belief for a set of decision-making rules. It is a transparent hybrid probabilistic expert system; a decision can be explained by the importance of the rules, weight of attributes, and the belief-degree in a decision inferred from the rules that are activated by the information of a given claim. The transparency of the decision-making system engenders trust in computer-aided decision making. Historical data may not contain rare claim circumstances. Therefore, a human expert can be leveraged to make decisions for such claims, which are then added to the training data for future machine learning from humans. This framework allows human experts and an AI system to work in partnership to enhance each other’s capabilities.
Original language | English |
---|---|
Publication status | Accepted/In press - 2021 |
Event | 31st European Conference on Operational Research - University of West Attica, Athens, Greece Duration: 11 Jul 2021 → 14 Jul 2021 |
Conference
Conference | 31st European Conference on Operational Research |
---|---|
Abbreviated title | EURO 2021 |
Country/Territory | Greece |
City | Athens |
Period | 11/07/21 → 14/07/21 |