The Reality of AI Bias

Unfortunately, bias is a massive part of human history. Bias is a prejudice that unfairly influences one result more than another due to internal or societal factors. This can be present in all areas of life, from unfair laws set by a biased judge to someone assuming everyone else thinks like them and making a wrong business decision.

Marketers especially have to be aware of bias in their decisions. Since they are trying to create content and build a brand for a specific target audience, all decisions should be made with that target audience in mind. Sadly, this is not how it always happens. A certain advertising campaign might seem like a good idea to a marketer who has a specific interest relating to it, while the target market might not care about the campaign at all. This bias can lead to huge losses on campaigns and should be avoided at all costs. Luckily, the introduction of data has helped prevent much of this bias, but ultimately the decision-maker will still be biased in situations without data to help.

This article from NPR describes how AI hiring software can be biased against people with disabilities. Using online “tests” designed to measure different aspects of an applicant's ability can cause flawed decisions if not supervised by a human. An example would be an applicant that has severe arthritis who might score poorly on a typing test. A hiring manager would excuse this if the candidate was qualified enough, but a hiring AI wouldn’t even show the applicant to the hiring manager.?

Humans are inherently biased, and it only makes sense that the things we create are biased too. In the distant future, it might be possible through machine learning and multiple human inputs to eliminate bias in AI. For now, we must learn to mitigate that bias instead. Using AI in our hiring systems was supposed to be a way to filter out human bias, but it was just replaced by AI bias. Using additional people to check an application after the AI reviews it would help lessen bias in this area. Unfortunately, that would require more money and a slower evaluation time. This negative bias can have both legal and economical repercussions so it is best to invest in company-wide anti-bias measures.