Abstract
This article was migrated. The article was marked as recommended. Machine learning approaches form the basis of "artificial intelligence" and have been increasingly applied in health services settings. It has been shown that such approaches may produce more accurate predictions in some contexts, compared to conventional statistical approaches, and may also reduce the costs of decision-making through automation. Nevertheless, there are both general limitations to developing and implementing machine learning approaches that must be borne in mind. To date, relatively little research has been published on the potential for machine learning to support personnel selection. Moreover, there are particular challenges and issues that need to be considered if such methods are to be used to support decision-making in medical selection scenarios. This article describes some of these potential advantages and challenges and presents an illustrative example, based on real-world data, related to the selection of medical undergraduates.
Highlights
There has been much publicity about the possibilities for ‘artificial intelligence’ to change our lives
Despite the hype surrounding artificial intelligence there are still relatively few examples of the approach being fully implemented as part of routine clinical services, though there are calls to make an understanding of the principles of artificial intelligence a core requirement of medical education, in preparation for its widespread utilisation (Wartman and Combs, 2018)
The potential of an assessment diagnostic process as a screening test is indicated by the ‘area under the curve’ (AUC) of the Receiver Operator Characteristic (ROC) curves, that ideally should be greater than 0.5 and as close to 1 as possible
Summary
There has been much publicity about the possibilities for ‘artificial intelligence’ to change our lives. The basis of artificial intelligence, occurs when a system learns from novel information presented to it in order to complete a particular task Such learning is often classified into "supervised" and "unsupervised". The ability of a system developed by DeepMind (formerly part of Google) to automate the diagnosis of eye disease from medical images (De Fauw et al, 2018) In theory, such systems only need to be as accurate as human doctors in order to justify their implementation, as they will free up medical staff time, providing cost savings. Fallibility is often considered a key part of being human, with Seneca the Younger famously quoted as stating that errare humanum est [‘to err is human’] This may be one of the reasons that, in practice, the suggestions of decision support tools are often overridden by clinicians (Roshanov et al, 2013). As such they are subject to stringent regulations in most jurisdictions and considerable resources are required in order to satisfy these so that they can be legally used in practice
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
