Using artificial intelligence to reduce bias in recruitment
The results are in and clear – diverse organisations do better. Using audited AI in the process can help avoid workplace bias.
Article information and share options
Money is not everything in life. Past a certain threshold, money will not make people happier. Below that threshold, money can make a huge difference. Too often, those below that threshold have been overrepresented by societal groups defined by gender, ethnicity, race, sexuality or neuro-diversity. In the search and allocation of good quality jobs, we are failing to be representative.
The question for recruiters is why this failure is taking place. Recruitment measures are outdated. An understanding of psychology is rudimentary; and processes are led by a 2D review of resumes, often screened artificial intelligence with no ability to contextualise.
A 3D approach to job applicants takes into accounts soft skills. These range from decision making to numerical agility, from risk tolerance to generosity. Not only do soft skills signal inherent potential; they are equally distributed and unbiased. Around 92% of hiring managers weight soft skills at least as highly as resume qualifications; but 75% of companies have difficulty assessing the soft skills of graduate applicants. Human beings are modular. There is no one-size-fits-all approach.
pymetrics provides three ways of overcoming the soft skills gap. Scientific innovation provides a better basis of measuring soft skills; tests can be administered with the support of artificial intelligence; and the methodology is fully audited and transparent to remove ethnic and gender bias.
There is a danger with AI that it replicates bias found in the wider world. This can be seen by how it accounts for women's colleges against co-eds. Some of the largest companies in the world have had to shut down AI hiring tools because they found them biased. This does not mean we should give up on AI, quite the opposite. Human-based recruitment processes are also biased – but unlike AI, humans cannot be audited, explained, or reprogrammed. Algorithms should be designed so that they can continuously and dynamically checked for bias.
The AI used by Pymetrics has increased job offers made to women by firms using the service by 74%; and has doubled the number of job offers to ethnic minority candidates.
And why is diversity important? A 2019 study by Gartner suggests that gender-inclusive and diverse teams can outperform their non-diverse counterparts by up to 50%. The difference between employee performance in diverse and non-diverse organisations is estimated at 12%.
Transparent and audited AI is one of the ways to achieve that diversity.
Summary based on Swiss Re Institute event: Algorithms for hope.
See also other summaries from this event
- Kate Darling - The new breed: What our history with animals reveals about our future with machines
- David Dao - Using artificial intelligence to help restore the natural world
- Patrick Schwab - Using AI to develop medicines with a higher probability of success
- Cathy O'Neil - Algorithmic accountability will lead to a better world