Search

Gender bAIs

Updated: Feb 28, 2021

Once, a CEO of a tech company once said, “The accountant helping manage our business is my brother.” And this wasn’t a lie - they both have the same parents. Yet, when asked the same question, the accountant claimed not to have any brothers. Take a second to think about this.


You may or may not have come to the conclusion that the CEO is actually the accountant’s sister. We’ve been trained since childhood to think of certain careers as masculine or feminine, and these prejudices have only been reinforced in higher education. When you hear the word “pilot,” you are likely to picture a man, and when you hear the word “flight attendant”, you are likely to envision a woman. Sure, this may be just a riddle, but it highlights one of the largest problems that still persists in society today: gender stereotypes.


A recent article published by Harvard Business Review describes the differences in words used to describe job performance for men and women. While men who are decisive are considered dependable and confident, women with the same quality are described as inept, selfish, or vain. This bias is engrained in the core of our society, manifesting in every decision we make. And despite being depicted as something only women face, stereotyping goes both ways, harming men the same way. If I was asked the same riddle as above, this time with a nurse and a secretary, it would be just as hard for me to associate a man with either of those professions.

When it comes to pursuing a career, these stereotypes rule our judgement. And if an individual steps outside of these bounds, he or she is frowned upon, both consciously and implicitly.

We may have hope that as younger generations grow up with more progressive ideals, these stereotypes will naturally fade away. Unfortunately , this expectation might be mere wishful thinking. If anything, there is evidence that these boundaries will only become stronger. As technology advances, artificial intelligence (AI) is being utilized to make major decisions, such as medical diagnosis, credit risk assessment, and marketing. Many proponents of AI believe that it isn’t biased by the emotions and misconceptions of humans. It’s easy to see computers as being perfect; it’s what they were made to be. However, computer models are affected by the same prejudices that exist within our human society. In fact, this issue might even be amplified for computers because unlike humans, who are able to learn and grow, computers function on mechanical algorithms and they cannot unlearn behavior.


When a model is created, it takes in data and learns from it. The more data inputted, the more accurate the results become. This process is known as machine learning. Because this process is facilitated by humans, any innate stereotypes they hold automatically are reflected in the AI. To look at an example, let’s say that an image database is being developed. The algorithm looks at all the pictures that are clicked after the search, finds similarities, and uses this to improve the first page of the search results. In the beginning, we can assume that the model is unbiased, because it hasn’t taken in any input yet. Users search ‘engineer’ and see a page with half men and half women. Because they are accustomed to engineers being men, they choose that picture. The algorithm adapts a bit. The next person does the same.

Again and again, users reinforce this bias, and the model learns that more male pictures are being chosen. Eventually, users searching for engineers will only find men, further strengthening the bias both in their minds and in the algorithm.

We take these unconscious actions daily, and they become a barrier, preventing us from overcoming gender inequality. Of course, there’s no quick fix to this. Researchers have been fixing skewed datasets and retesting models, but this isn’t enough. Until we don’t change our own behavior, how can we expect the models to change? The first step is acknowledging our misconceptions. Next time you hear of a profession and immediately attach a gendered notion to it, try to pause and ask yourself why. It’s no secret that our thoughts affect our behaviors and actions, so the more we read, listen, and reflect, the deeper our understanding of the problem will become. In turn, the more we work on addressing our implicit biases, the quicker our actions will change. Once we change our own attitudes, it’s only a matter of time until AI does as well.