A general talent gap and shortage are challenging today’s medicine and healthcare. As the Association of American Medical Colleges’ Center (AAMC) Workforce Studies found, we may be facing 45,000 few primary care physicians and 46,000 few surgeons and other medical specialists in the next 10 years. On the other hand, J.P Morgan’s 2018 Healthcare Industry Outlook revealed that 92% of senior healthcare executives are worried about finding qualified candidates to fill respective roles.
As such, healthcare organizations may find themselves in fierce competitions for acquiring new talents. On average, it takes about 14 months to fill for a physician but the rivals among recruiters may prolong the process. Since artificial intelligence (AI) is gradually exercising its influence in the industry, it’s not at all surprised to find traces of it within the medical recruitment domain, to automate and streamline the rather time-consuming and not so straightforward hiring process.
What can AI do?
Traditionally, recruiters will have to generate job descriptions, analyze the current workforce data, fact-check information and references provided by applicants before deciding on the candidates they will like to proceed with. Machine learning may be handy in the resume screening process especially for positions with high-volume of applicants or high-turnover rates. While predictive analytics may take into consideration one’ previous experiences, tune, and turnover rates as well as indicators from aptitude, personality or skill tests to make a more informed decision on whether the candidate is truly suitable for the role and the corporate culture.
Survey results showed that up to 86% of physicians are passive job-seekers, which means they are not actively looking for new jobs yet most of them do not mind finding out more about appealing job opportunities that are posed to them. Knowing that, AI may widen the number of channels a hiring manager can get hold onto; from conventional job platforms where most recruitment procedures take place, to others like professional communities; personal websites; portfolio portals and social media networks. AI is able to review many of these candidate sources based on a few data points on the ideal candidate one is searching for; this is hard to achieve manually.
Can AI do the job?
Earlier on, Amazon abandoned its automated system that ranks job seekers with one to five stars as it begins to show a preference towards male candidates for technical roles. This AI was trained using 10 years’ worth of resumes the company had received. Since technology is a relatively male-dominated line, the system was unintentionally trained into a biased machine, discriminating resumes containing words like “women” or “female college”. Although Amazon tried to savage it by neutralizing certain terms, they have to eventually lost confidence in what it can achieve.
Recently, the Electronic Privacy Information Center (EPIC) filed a complaint with the Federal Trade Commission (FTC); requesting it to investigate HireVue, an AI tool that is assisting many companies in rating job applications based on their video interviews. The software analyzes details like interviewees’ choice of words and facial expressions to generate a so-called “employability score”. The problem lies in whether facial expressions will determine a person’s success in a job.
Besides, some worry that the HireVue algorithm may be bias towards White, male candidates or someone whose first language is English and/or carry certain accents. It’s also possible that candidates may reply or game the interview in the way they know HireVue will prefer. It’s challenging to determine if an AI-driven hiring tool is unfair because they are not regulated at the moment, so companies have no obligation to openly share their training data or prove that their systems are impartial. Like facial recognition, even if AI can do the job, it doesn’t mean it can do a good job as they come with too many uncertainties.
Can AI help us hire the best doctors?
We will probably have to define what is “best” in the first place. For a healthcare institution, the “best” doctor maybe someone who has a brilliant track record or expertise in a particular area. For patients, the “best” doctor tends to be someone who goes an extra mile to care for them. “Best” is relative in this sense and unless data used to train the AI can capture this, we may not be able to use it to hire “the best doctors”.
Nevertheless, as mentioned earlier, AI does have its benefits in the hiring process. What we can do, at the minimum, is to inform candidates that they will be screened by an algorithm and for the institutions to weigh the pros and cons of possible prejudice.