Ah AI, Our Old Friend

By now you know what AI is. Not just that it stands for artificial intelligence but that it takes in information and uses it to make predictions and decisions. AI powers the social media feed that suggests the next post, it informs the commercials that show up on your streaming service, it drives the software that just auto-completed this sentence. In more serious situations it can determine if you qualify for bail or a home loan. This means that AI-enabled systems both make decisions about you and also make decisions for you. The effects of this can be dire — unethical, even.

When AI decides what you should see or who you ‘are’ it’s also deciding what you shouldn’t see and who you won’t ‘be’.

Where are the humans?

All around us, you can find AI making choices for us. But it isn’t simply a computer acting on its own. AI is designed by us, humans, and that has a huge impact on everything AI does. It’s us humans that tell the computer what to consider and how highly it should rank certain data points. The expectation with AI was that a computer would come to a conclusion using objective information and some fancy coding. The reality is that AI is just as unfair as the creators that trained it.

Here are some examples. Take credit scores — algorithms trained in a biased way may calculate a lower score for a woman compared to a man, which could in turn affect what credit cards they could open or mortgages they could apply for. Another example is job hiring — recruiters who rely too heavily on software may inadvertently reinforce race and gender-related blindspots. There’s even car insurance pricing — when opaque algorithms determine who should pay less or more for similar coverage, it offers the opportunity for unfair pricing.

Cansu Canca walks us through this problem. Watch the video below to learn more about where and how human bias is introduced into AI-enabled systems.