It is difficult to think of a recruitment industry trend that has gained more interest and attention than Artificial Intelligence (AI). New software developers and small start-ups offer new and innovative ways of sifting, assessing and selecting the right candidates through the use of AI, Natural Language Processing (NLP) and Machine Learning (ML) algorithms. The surge in interest in these tools is clear: as our understanding of our own cognitive biases (cognitive 'shortcuts' or heuristics that underpin our thought processes) continues to grow we become increasingly aware of how frail and inaccurate our subjective judgements can be from an assessment perspective. Rightly, recruiters are now exploring new ways of working that will help mitigate bias and ensure that the right candidate is selected for the role and equality of opportunity is maintained.

Some tech companies offer totally "unbiased data collection" and others promote a "bias-free" method of assessment. However, to what extent are these marketing claims true when these new and exciting tools are scruitinised?

An investigation by Bavarian Broadcasting (BR) journalists demonstrated that popular AI recruitment tools can easily be swayed by appearances.

"Less prejudice, more objectivity: An application process that is not influenced by the personal preferences of a recruiter. That is the promise of many AI companies entering the market worldwide..." BR notes.

However, they found in controlled tests that one's choice of clothes, glasses, background content, lighting and filtering when participating in a video interview scored with AI technology achieved completely different results. Even when the content of the candidate response to the interview was the same.

Professor Uwe Kanning, is concerned that such software tools can only replicate subjective feelings that an assessor or recruiter would otherwise possess - leading to the reinforcement of stereotypes.

These findings indicate that the interest and excitement generated around this formative technology may currently outstrip its utility and its accuracy, until further research can be conducted.