How Unconscious Bias is stunting the development of Artificial Intelligence and Machine Learning in Recruitment


Artificial Intelligence is revolutionising industries such as finance, travel and healthcare. Computers are more accurate, much faster and more reliable than humans are; and Deep Learning is allowing computers to talk to us, drive our cars and diagnose diseases earlier than we can.

Yet, it has failed to make much of an impact on the recruitment industry as a whole. The majority of recruitment and talent acquisition professionals that I have spoken to are still to use Artificial Intelligence in the recruitment process; and of those that have, a lot have been unimpressed with the technology.

This is because it hasn’t been developed enough to make appropriate recommendations and hiring decisions. There’s no doubt that Artificial Intelligence and Machine Learning can have a huge benefit on the recruitment industry, but there are several factors that are slowing the technologies development. Unconscious bias is one of these.

As humans we can’t help being unconsciously biased towards certain individuals. It may be because they have similar personalities, look nice or share certain abilities with us.

Unconscious bias can lead to us making bad hires or choosing one individual over another because you have a ‘good feeling’ about them or you relate to them more; resulting in a better fitted candidate being overlooked for a particular role.

People can make decisions based on unconscious biases such as;

Similarity Bias – preferring an individual because they are of similar age, ethnicity, gender, etc.

Affinity Bias – preferring an individual over another because they share similar qualities/skills to you.

Beauty Bias – preferring an individual because of their appearance by associating it with their personality

Conformity Bias – when your views are swayed by those people around you and to seek acceptance from these, we are prone to choosing someone that has attributes that other people around you will like.

Halo Effect – when you focus on a particular feature of someone that you like, you tend to view everything about that person in a positive light.

Each of these unconscious biases can result in us choosing a candidate over another because of something other than their ability or experience; and this is making it harder for Machine Learning systems to teach itself to recommend the best candidates or make good hiring decisions.

Reinforcement Machine Learning uses algorithms that are able to learn from experience to shape its decisions. It does this through finding patterns in the data it is fed. However when bias affects data, it can skew the results and make important patterns not as clear, decreasing the end results success. Therefore, we need to make sure that the data fed into the system isn’t affected by bias.

Conscious bias can be easy to spot but unconscious bias isn’t as obvious. Unconscious biases such as Similarity and Affinity Bias can’t be noticed without analysing the hiring manager as well as the candidate, and both Conformity Bias and the Halo effect would be almost impossible to identify where it has occurred.

Because of this, Unconscious Bias could result in lots of skewed data being fed to the Machine Learning system and the system’s judgement being poor.

To overcome this, lots of time and resources will need to be put into carefully vetting the data; which means it may take several years to build a system that can recommend and make hiring decisions as well as humans can.

Of course, Unconscious Bias isn’t the only factor affecting the development of Artificial Intelligent technology in the recruitment industry, but I believe it is one of the main reasons why software already out there isn’t performing as well as we would hope in sourcing & screening candidates.

Why do you think Artificial Intelligence hasn’t had a big impact on recruitment?

Author: Luke Amber – Business Development Consultant

(Click to Connect with him on LinkedIn)