@inproceedings{10.1145/3630106.3658971,
author = {Schor, Bianca Giulia Sarah and Kallina, Emma and Singh, Jatinder and Blackwell, Alan},
title = {Meaningful Transparency for Clinicians: Operationalising HCXAI Research with Gynaecologists},
year = {2024},
isbn = {9798400704505},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3630106.3658971},
doi = {10.1145/3630106.3658971},
abstract = {AI systems can bring great benefits to our healthcare systems, e.g. by improving patient outcomes. Yet implementing them into clinical practice remains challenging. To bridge the gap between academic research and design implementation, we argue clinicians need transparency about such systems that is meaningful—i.e. contextually appropriate—to them. Towards this, we explore recent HCXAI recommendations for building transparent AI systems for users in a specific domain: gynaecology. By better understanding clinicians’ perspectives on meaningful transparency, our aim is to complement and help operationalise such recommendations. We conduct a co-design workshop and interviews with n=15 gynaecologists in the UK and the Netherlands. We show that HCXAI must better account for clinical teams with different types of gynaecologist users, and that the timeliness and relevance of the information provided about the AI-based tool throughout its design lifecycle—in particular before a tool is implemented into clinical practice—is crucial for transparency to become meaningful. Our contributions include: i) testing recommendations from the latest HCXAI literature with a prospective, real-life AI application in a relatively less-studied clinical domain; ii) describing and visualising gynaecologists’ understanding of meaningful transparency for clinicians; iii) outlining four design recommendations towards realising meaningful transparency for clinicians and opportunities for research; and iv) expanding HCI and AI research in women’s health by directly engaging with gynaecologists as users and co-designers. Exploring such issues is key to facilitate the implementation of AI systems that meet clinicians’ information needs and that they can trust.},
booktitle = {Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency},
pages = {1268–1281},
numpages = {14},
keywords = {clinical Artificial Intelligence, clinicians, co-design, explainable AI, gynaecologists, transparency},
location = {Rio de Janeiro, Brazil},
series = {FAccT '24}
}