Unconscious Bias in AI?


How does a computer discriminate? You add a formula into a computer and it will provide a result. Unless there is a deliberate amendment in the algorithm, a computer can’t discriminate right?


As we have been seeing over the last few months, bias and lack of diversity are now affecting products where decisions should be automated. Earlier this year, there was the case of the sensor hand dryer. Black man puts his hand next to the sensor, nothing. A white man gives it a go and magic, its working ok. There was no attempt from the manufacturer to discriminate, the issue was that they had no black product testers. Your teams should be representative of wider society, and a properly implemented diversity policy would have identified the well before it left the factory.

The latest case of bias in AI relates to the Apple Card, or not the apple card but the black box automation in the application process. If you stand back, the algorithm responsible for the credit decisions seems to be giving women lower credit than equally qualified men.  The most high profile case relates to Devid Heinemeir Hansson , where he and his wife both file joint tax returns. The algorithm sitting in the black box decided that David deserved a credit limit of 20 times higher than his wife.

When they tried to query it with Apple card customer support, the customer service staff advised how there could be no discrimination as all the decisions are  based on a algorithm. Other high profile names also such as Steve Wozniak (yes, co-founder of apple) and film maker Lexi Alexander have also raised concerns about the gender bias in the process.

While the algorithm should not develop any form of bias, there is the potential for bias of people developing the algorithm to feed in. A clear example can be seen with the Apple card. Treating everyone as individuals on an application would benefit men as we still live in a society where we have not reached equality in the workplace.

At Ohcul jobs, we promote diversity within the workplace through removing the opportunity for people’s unconscious bias to reflect in the recruitment process.

Employee diversity provides a number of benefits to any organisation (large and small). Ethnically diverse companies perform 35% better than their industry median and gender diverse companies are 15% more likely to haver returns above the industry median. With the national customer base becoming more diverse, it is important that a company is representative to fully understand its clients’ needs and, according to the latest research, 67% of employees cite a diverse workforce is a key factor when looking at a role.

At Ohcul, we remove the opportunity for hidden bias from a CV review stage. Our unique tool removes identifying data from a clients CV allowing a recruiter to progress the best candidates.

If you would like some more information you can contact us on info@ohcul.com or visit the web site www.ohcul.com