4 min read

Algorithmic Bias: What’s Missing from Big Data?

Featured Image

According to The New York Times, Forbes, and McKinsey, we live in the Information Age. Our era derives its name from information technology: the use of computers to create, process, store, and retrieve electronic information—i.e., data.

Humans have conveyed information between people, places, and generations since the development of written language around 3000 B.C. Oral histories suggest that this tradition may originate even earlier—more than 10 thousand years in the past. Over time, our sharing methods have grown more sophisticated, accurate, and efficient, from clay tablets then to information technology now.

Why is information technology special? Well, from an evolutionary perspective, it’s effectively a superpower. Lifespan and brainpower place a “cap” on all species’ ability to learn. We can only know as much information as we have time to encounter in a single lifetime. We can only create, process, store, and retrieve as much information as our memory and processing systems allow. Most nonhuman species are unable to exceed this cap. Research suggests that some animals can “accumulate knowledge and improve performance” over generations. For example, pigeons have been shown to “share” information about the fastest flight routes. However, no nonhuman species comes close to our information-sharing abilities. 

The invention of written language and oral history allowed humans to create, process, store and retrieve information collectively. We can share knowledge, transcending individual limitations. But we are still subject to human limitations:

  • Sensory. Can you see ultraviolet light, hear infrasound, feel tectonic vibrations, or detect radioactivity? Specialized technology can.

  • Processing. Consider the equation “7810933232 multiplied by 32392.” Who can solve it faster, your mind or your smartphone?

  • Memory. Can you memorize the lyrics to every song in your smartphone’s music library or every word in your Kindle’s archive? Probably not.

Information technology allows us to transcend even these limitations. Today, we can create, process, store, and retrieve oceans of electronic information (data) with the push of a button. From search engine optimization to social media algorithms, information technology is ubiquitous and transformative in our daily lives.

However, it’s not all good news. Information technology relies on Big Data: modern data science that leverages huge, complex data sets to accomplish equally huge, complex objectives. Three characteristics make Big Data, well, “big” data:

  • Big Data consists of a massive volume of data gathered from a diversity of sources: smart devices, industrial equipment, research equipment, etc.

  • Big Data travels at an incredible velocity, thanks to the Internet.

  • Big Data comes in a wide variety of formats: text, audio, image, etc.

Big Data can broaden the information available to us by handling superhuman volumes, velocities, and varieties of data. It works quietly behind the scenes to augment, enhance, and optimize human work. Often, Big Data helps us by making small but significant choices for us—and therein lies a danger.

Big Data makes bias automatic. 

The information technology we rely on is vulnerable to algorithmic bias: “instances when…an algorithm compounds existing inequities,” for example, inequalities linked to socioeconomic status, gender race, or other identities.

Our data collection systems are not impartial, having inherited the biases of their human developers. These biases can easily manifest in the underrepresentation and undervaluation of data from marginalized groups. For example, there’s the “sex and gender gap” described by bestselling author Caroline Criado Pérez in her novel, Invisible Women. Pérez asserts that although women’s rights have advanced greatly in recent decades, “man” remains the implicit default. This bias privileges male data as gender-neutral. Hence, male data exists in excess and, often, instead of female data.

Algorithms tend to skew toward distinctly male averages. In 2020, the British Broadcasting Corporation (better known as the BBC) exposed bias in the design of “unisex” personal protective equipment, like masks. Even the smallest sizes were “too big for some women—who make up 77% of the [National Health Service] workforce.”

Big Data is not All Data. Biases such as the “female-shaped absent presence” that Pérez describes in Invisible Women warp our data, algorithms, and lives, which rely on the information technology that defines our era. Living in the Information Age, it is our responsibility to ask and address the question: What's Missing from Big Data?

Synaptiq is Here to Help

Synaptiq focuses on the humankind of AI; building a better world as we lean into an age of collaboration between humans and machines.

We are a full-scale data science and AI consultancy with over 60 clients in 20 sectors worldwide. Our mission centers on growth and partnership—in and between organizations, people, and the technology that makes our work better.

Learn more about us and our partnerships on our website, https://www.synaptiq.ai/

Additional Reading:

Cross-Species Communication? Researchers Say AI is the Key

Technology Could Help Us Understand Animals

Although animals don’t “speak” like us, they do communicate.  In 1973, ...

We Helped a Non-Profit Expand Road Access in Rural Rwanda

The Mission: Expanding Road Access in Rural Rwanda

In 2020, a non-profit set out to solve a problem: the lack of road...

Artificial Intelligence in the Wild: Helping Conservationists Save Species

Artificial Intelligence in the Wild

Scientists warn that we’re facing Earth's sixth “mass extinction.”

IExtinction is...