Trustworthy enough? Evaluation of an AI decision support system for healthcare professionals
Year: 2023
Type: article
Abstract: Exploring end-users' needs and involving them in designing and evaluating an AI system should be a priority.Understanding the system is essential to assess whether to trust it or not.This paper discusses a use case of a decision support system integrated into a platform for healthcare professionals (medical call operators and nurses).The system navigates them by predicting what intervention should be taken when an accident happens to a patient at home.Our use case demonstrates the importance of humancentred evaluation methods and potential struggles with mixed methods as detected by differences between qualitative and quantitative approaches.A subjective scale in combination with group interviews was used to evaluate the trust towards the system.The results showed that while users expressed a relatively high trust in the scale, the qualitative insights indicated uncertainty and the need for better explainability to trust the decision support system.In line with the results, we point out the need for better human-centred evaluation methods, as the current subjective scale needs to be complemented by qualitative methods to ensure rich insights.
Author: Kristýna Sirka Kacafírková
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: Innoviris
OpenAlex ID: W7025000635
