Why Is AI And Machine Learning So Biased? The Answer Is Simple Economics Posted on : Jan 21 - 2019

As AI and machine learning have infused themselves over the last half decade into nearly every corner of our lives, there has been a growing interest in how the biases of these models may be silently impacting society. Much of this focus has been on the issues of biased training data and a homogeneous workforce that lacks sufficient diversity of experience to recognize bias. However, lost in this conversation is the far bigger driving force: the lack of economic incentive to minimize bias in the technologies that increasingly power our lives.

The digital world is an incredibly biased place. Geographically, linguistically, demographically, economically and culturally, the technological revolution has skewed heavily towards a small number of very economically privileged slices of society.

Silicon Valley churns out an endless firehose of new technologies that promise to change the world. From internet connected kitchen appliances to smartphone apps that promise to get our fast food to our doorstep ever faster, the majority of these creations target customers that are either affluent themselves or who are monetizable by others.

We focus the collective brilliance of society’s technologists not on providing clean water to those dying of droughts or disease, not halting epidemics of strange diseases that don’t immediately threaten Silicon Valley’s glitterati, not building shelters or improving literacy or anything that might make the world a better place for all. Instead, so much of our nation’s creativity is focused on frivolous pursuits that make our already extraordinary lives a bit more comfortable and their creators a whole lot richer.

This focus on commercialization potential rather than public good means our technology landscape skews heavily towards innovations that can yield the greatest profit rather than those that can realize the greatest societal good.

The web of today is a largely visual place, filled with imagery and video. Websites rush to replace text with visual metaphors, images, videos and high design. Pages of words are replaced with pages of images and videos. Even news outlets increasingly offer video-only content.

For their part, social media platforms have incentivized multimedia, prioritizing its placement and building entire platforms that exclusively distribute visual content. Why post a paragraph of text when you can post a single image that conveys all of your thoughts at once? Or a video that allows someone else to experience the world through your eyes?

Yet, as the web has rushed towards a visual world, it has invested little in ensuring that world is accessible to those with differing visual abilities.

Companies spend incredible resources hand tuning their websites to work on every conceivable mobile device. Every new phone release yields a frantic rush to ensure pixel-perfect rendition

In contrast, ensuring websites are accessible to those with different physical abilities is at best relegated to a bare minimum afterthought done only if local law requires it. View More