Purpose
By surveying the current perceptions of clinicians, it is possible to better tailor education syllabuses to ensure that potential gaps or concerns are addressed as early as possible. The results of this study could help computer scientists to consider how machine learning systems are seen by new users, to make sure that potential issues are addressed in new implementations that can only come from outsider perspectives. This study explores participant’s thoughts on both existing technologies, and ideas they may have for potential future technology they...
Methods and materials
The practice of studying the user of healthcare technologies is a well established aspect of design in the field of Human Computer Interaction1. Mistakes due to overly complex or outdated systems can cause catastophic failures and severe patient harm2 3. Similarly, there is a growing call to educate both the public and professional on the nature and internal workings of AI to mitigate anxiety around its ubiquitous nature and possible misapplications4. These factors demonstrate the need to survey clinicians for their perspectives on machine learning....
Results
When asking participants about the effectiveness of different technologies, most participants rated each technology as roughly 6/10 [Fig 1].There were individual scores that placed quite highly but every prompt averaged out to a mean between 6 and 6.5. This implies that, in general, participants gave a middle-of-the-road score and appeared hesitant to display any strong feelings on the topic [Fig 2].
Regarding examples of machine learning and automation that participants experiences regularly, the key response across the cohort was that digital assistants such as Alexa...
Conclusion
We identified three key themes of perception on ML and AI; an awareness of the ubiquity of these systems in the medical field but a lack of distinct knowledge regarding the origin and nature of the data they use to “learn”, a general lack of trust in ML-based tools to operate as effectively as a human co-worker, and a cohesive group response to many prompts through discussion before, during and after formulation of an idea.
Whilst there appears to be optimism regarding the presence of...
Personal information and conflict of interest
R. S. Clark:
Nothing to disclose
M. Porcheron:
Nothing to disclose
M. Jones:
Nothing to disclose
P. Wardle:
Nothing to disclose
V. E. Whitchurch:
Author: National Imaging Academy of Wales
Advisory Board: National Imaging Academy of Wales
References
1) Mark Hartswood, Rob Procter, Mark Rouncefield, and Roger Slack. 2002. Performance Management in Breast Screening: A Case Study of ProfessionalVision. Cognition, Technology and Work 4 (01 2002), 91–100.
2) Arun Rai. 2019. Explainable AI: from black box to glass box. Journal of the Academy of Marketing Science 48, 1 (2019), 137–141
3) Nancy G. Leveson. 2017. The Therac-25: 30 Years Later. Computer 50, 11 (2017), 8–11.
4) Elisabeth Sulmont, Elizabeth Patitsas, and Jeremy R. Cooperstock. 2019. Can You Teach Me To Machine Learn?. In...