Increasing workplace diversity with unbiased AI tools

David Leithead 09.11.2018

Companies are striving to eliminate discrimination in their recruitment processes by implementing ‘unbiased’ Artificial Intelligence tools.

Removing unconscious biases to increase theory

Diversity is an integral aspect of recruitment today. This includes the need to source talent from a diverse pool, but also ensuring that talent flow is not staunched by conscious or unconscious interventions from people with preconceptions and preferences for a certain type of candidate. In theory, recruitment tools that match or screen CVs have the potential to reduce bias because they make decisions solely based on data, and are not influenced by context. Artificial Intelligence developers and those working with Augmented Intelligence claim their tools can improve diversity because they are designing them in such a way that biases play no role when it comes to hiring decision making.

This is an attractive prospect for companies, their hiring managers and for recruitment agencies alike, both for the reasons of finding more diverse talent and eliminating discrimination in the hiring process.

Read the first article in our AI series

How will these tools work?

How can we eliminate automatic, mental shortcuts that are ingrained in every human and so often drive decision making? Artificial Intelligence tools have the potential to do this as they assess large quantities of data on an objective basis, free from unconscious bias. When analysing CVs, all demographic information (including race, sex, sexual orientation, national origin, religion, disability and age) is ignored when it comes to selecting the most appropriate candidates, but also the less obvious biases like, educational background, implied social status and so on.

For example, MeVitae uses natural language processing (NLP) to remove information that could be cause for discrimination, namely gender and ethnicity, before forwarding the CV to employers. Their technology aims to help companies avoid unconscious bias throughout the hiring processes.

Decision making uninfluenced by past experiences

People’s behaviours and attitudes are highly context specific, changing on a daily and even hourly basis, especially when there are time constraints - these attitudes inevitably influence decision making. Humans have biased views, whether we recognise them or not - and this is increasingly argued to be, in principle, a major flaw in human CV reviewing. But are we right to just trust that AI tools will be running totally bias-free?

Despite the claim of an unbiased tool, the nature of AI is to learn by perceiving patterns from past decisions and if there are hidden biases (which there likely are), the tool will inevitably pick them up.

Historical biases that are reflected in current workforce patterns may teach machines bad habits. Where hiring companies give the AI recruitment tools access to their talent management platforms, things could get even uglier because the AI tool can learn from an historic bias and simply perpetuate it, because it doesn’t know otherwise.

Increasing efficiency and improving processes

Say a company is trying to improve the diversity of its workforce, and is using Artificial Intelligence tools to screen CVs, where the tools have learned what CVs have most success in the hiring processes in the past. Humans can intervene to audit/assess the tools for potential biases by asking the tools to rank and grade applicant CVs, then assessing the demographic breakdown after the grading is complete. If there are glaring similarities in the results to the current status quo, it can be assumed that the machine has learned or inherited the biases active in the company’s processes.

This could turn out to be positive as the tool will expose this as a hidden bias in the company’s recruiting history and gives the opportunity for it to be acted upon - but it’s obviously a case of humans working alongside AI to improve recruitment processes.

There’s definitely something fundamentally appealing about removing human bias from CV screening, but there’s also clear fundamental risk in trusting the machine alone. Identifying diverse talent is a complex puzzle, it’s often about breaking patterns, not following them or making them - and for the moment at least, teamwork gets a better result.

Get in touch

If you have any questions or opinions about AI in recruitment, click here and fill out the form to get in touch today.

Other Artificial Intelligence in recruitment articles

Keep an eye out for our other articles in the series...

David Leithead's picture
Chief Operations Officer UK