AI technology is increasingly being used in recruitment for jobs. Recruiters can use AI programs that analyse the applications for a particular post, selecting the candidates that meet the criteria for the next round of recruitment. They may also use programs that can analyse applicants' responses to questions in an interview or other screening stage. The makers of AI recruitment tools say that their software can speed up recruitment and hire the right person as successfully as human recruiters. It is important to consider when humans should have input into the process, and how to check that the AI is being fair. 

​

One area of concern is the lack of transparency that comes with AI software, so the decisions on particular candidates cannot be explained in detail by the program. This could make it very difficult for a rejected candidate to complain about the process, or even to get feedback on why they were rejected. 

​

The system could also be unfair. If the AI is trained on data from the history of recruitment at the company, it could inherit unfair policies for gender, race and other protected elements. 

​

This unfair bias happened at the tech giant Amazon, when it introduced an AI recruitment program. The program inherited a bias for recruiting males for technical roles, as the company (and the tech sector in general) had a historic gender imbalance. So the AI found that bias in the training data, and used it, because the people who wrote the AI program didn’t realise the bias would be there.

​

When developing and using AI recruitment tools, we should be aware of the larger implications of using AI where we aren’t aware of the details in the training datasets on which it was trained.