AI Bias Is Real, And We Created It
Updated: May 25, 2020
Goldman Sachs Accused of Gender Bias This week Goldman Sachs was accused of gender bias when applicants for the Apple card noticed that women received much less credit than men with the same qualifications. Goldman Sachs put out a statement that says gender and marital status are not considered in its computerized approval process.
Even so, the story reminded me that artificial intelligence has bias, and we created that bias.
Who’s Awesome? 🙋 Despite what Google thinks, I am. About a year ago, I wanted to email a cute meme to a bunch of girlfriends that said, “I’m awesome!” The immediate results on Google were disheartening. I couldn’t find a woman. Are we just not awesome? Likely, we don’t say how awesome we are as much as men do.
A Study of Googled Memes On November 11, 2019, I decided to take a deeper dive into these search results to better understand who Google thinks is “awesome” or who frequently states, “I’m awesome.” To remove my own bias, I searched “im awesome memes” incognito. (see screenshots of results at the bottom of this article)
Aside from a bunch of white men with mustaches and neckties, these were the results:
Result #7: First Latino Man
Result #14: First Black Man
Result #27: First White Woman
For this study, I categorized the memes as 1) male or female (I did not include trans individuals, and, unsurprisingly, there were no examples explicitly representative or inclusive of trans individuals), as 2) white people or people of color, and 3) other or not enough information to tell.
Can we do better? You may be thinking, “But, Google has an algorithm.” Yes, and it’s based on the content we, as a society, have created and regularly consume. I’m not the expert on AI, but you and I are training AI with our daily behavior. Our record for bias isn’t great, even when we become aware of it.
We have a responsibility to consciously go against our gut. We must make choices to lift the voices of those who have been silenced from saying, “I’m awesome!” in the past. AI is our future, and AI is listening. Can we teach it to do better than we did? Click below to scroll through the first 50 or so results in Google.
The ideas in this article are my own and not that of my employer or any program I am a part of. I am not a scientist or DEI expert; I am someone who cares. For more information on AI and bias, see the resources below: