Abstract
Background
Over half UK Medical Schools used online interviews for 2024 entry. Artificial intelligence (AI) is widely used medicine. However, there are no publications of its use by applicants to undergraduate degrees to improve their chances of selection.
We report a pilot study to test the hypothesis that the use of AI by candidates participating in online interviews for Nottingham Medical School would enhance their performance.
Methods
Three historic scenarios and associated questions from the University of Nottingham (UoN) Medicine course selection interviews were submitted to three AI platforms as a potential candidate could during an interview.
For each scenario, the setting of the task (priming statement), the details and instructions of the scenario, and three questions were submitted in turn.
Entry and responses times and how these could be optimized were studied using two approaches for text entry, typing directly into the AI platform and recording directly into the AI platform using a second device.
The quality of the AI responses was scored independently by the authors using a ‘traffic light’ scoring system where ‘green’ was ‘the response was satisfactory/appropriate for an 18-year-old candidate’.
Results
Entry and response times: Typing into the AI platform was associated with delays before the AI response started. Direct entry using a second device had no such delays.
All three AI platforms started responding when as the text was being entered. But median times for completion of the response varied between 14sec and 41 sec.
The quality of the AI responses: Only the minority of the responses, if read verbatim, scored ‘green’.
Conclusions
AI could potentially be used by an applicant for the UoN Undergraduate Medicine course to ‘enhance’ their online interview performance, but it would be difficult in practice. The candidate would have to use a second device, use the response to the scenario details/instructions to formulate prompt and more natural answers to the questions rather than reading the AI question responses verbatim and hope there were no technical problems. We think that candidates could perform better by using AI before the interview day to generate answers that are polished, confident and natural.