Paper Number

1540

Paper Type

Completed

Description

Machine Learning (ML) systems have an enormous potential to improve medical care, but skepticism about their use persists. Their inscrutability is a major concern which can lead to negative attitudes reducing end users trust and resulting in rejection. Consequently, many ML systems in healthcare suffer from a lack of user-centricity. To overcome these challenges, we designed a user-centered, trustworthy ML system by applying design science research. The design includes meta-requirements and design principles instantiated by mockups. The design is grounded on our kernel theory, the Trustworthy Artificial Intelligence principles. In three design cycles, we refined the design through focus group discussions (N1=8), evaluation of existing applications, and an online survey (N2=40). Finally, an effectiveness test was conducted with end users (N3=80) to assess the perceived trustworthiness of our design. The results demonstrated that the end users did indeed perceive our design as more trustworthy.

Comments

13-DataAnalytics

Share

COinS
 
Dec 11th, 12:00 AM

The Power of Trust: Designing Trustworthy Machine Learning Systems in Healthcare

Machine Learning (ML) systems have an enormous potential to improve medical care, but skepticism about their use persists. Their inscrutability is a major concern which can lead to negative attitudes reducing end users trust and resulting in rejection. Consequently, many ML systems in healthcare suffer from a lack of user-centricity. To overcome these challenges, we designed a user-centered, trustworthy ML system by applying design science research. The design includes meta-requirements and design principles instantiated by mockups. The design is grounded on our kernel theory, the Trustworthy Artificial Intelligence principles. In three design cycles, we refined the design through focus group discussions (N1=8), evaluation of existing applications, and an online survey (N2=40). Finally, an effectiveness test was conducted with end users (N3=80) to assess the perceived trustworthiness of our design. The results demonstrated that the end users did indeed perceive our design as more trustworthy.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.