Machine learning bias in google’s quick draw

Machine learning systems are increasingly influencing many aspects of everyday life, and are used by both the hardware and software products that serve people globally. As such, researchers and designers seeking to create products that are useful and accessible for everyone often face the challenge of finding data sets that reflect the variety and backgrounds of users around the world. In order to train these machine learning systems, open, global โ€” and growing โ€” datasets are needed.

Googleโ€™s latest approach to helping wide, international audiences understand how neural networks work is a fun drawing project called quick draw. In short, the quick draw system will try to guess what you are drawing within 20 seconds. While the goal of Quick, Draw! was simply to create a fun game that runs on machine learning, it has resulted in 800 million drawings from twenty million people in 100 nations, from Brazil to Japan to the U.S. to South Africa.

In an article posted by google in 2017, google shared the inherent bias in the Quick, Draw! database that they collected. One example that stands out is the shoe style example. when analyzed, 115,000+ drawings of shoes in the Quick, Draw! dataset, it was discovered that a single style of shoe, which resembles a sneaker, was overwhelmingly represented. Because it was so frequently drawn, the neural network learned to recognize only this style as a โ€œshoe.โ€


This biased is may be rooted in the Quick, Draw! user base (could it be mostly men)? or is it even more primal than that? could this bias be rooted in us as we (women and men), are thinking of a shoe? this of course is varied between cultures and context and has relates to how we know what we know – an epistemological question. 

Of course, when we build machines that aspire to know something we want them to know all the possibilities of a specific something, and for them to keep learning and be flexible to changes and a wide verity of possibilities.

This, in my perspective, is an enormous challenge, philosophically and technically. Machine learning has to adapt like humans do because change is the only constant.

Leave a Reply

Your email address will not be published. Required fields are marked *