Design

Why do we keep gendering our AI assistants?

4 min read
Will Fanguy  •  Jan 26, 2018
Copied To Clipboard

Apple recently opened pre-orders for their new HomePod, a speaker and smart home device that brings their virtual assistant Siri into your home in the same way Amazon’s Echo devices and Google’s Home devices have allowed Alexa and Google’s assistant to help control your lights, play music, and add items to your shopping lists.

This new device gives us another opportunity to explore one of the major issues in AI design today: assigning gender roles to the devices and AI personalities we utilize every day. This is an overarching issue that is prevalent in design and often reinforced by the little choices we make every day as designers.

Image courtesy of #WOCinTech Chat || wocintechchat.com

By assigning gender to these AI personalities, we may be saying something about the roles we expect them to play. Are technology companies catering to our desire for robotic assistants with personality, or are they reinforcing our biases about gender, and the roles that women play? What are our expectations for an AI with a female voice and feminine personality?

Related: What every designer can learn from Alexa

The problem arrises because our technological culture consists of typically male-dominated engineering groups (the Silicon Valley standard). These groups routinely give women’s names to the things we issue commands to with little to no user feedback or input from a forward-thinking, socially-minded design team. It’s a problem because we’re letting a minority stakeholder (possibly affluent white men) make design decisions that will affect millions of users. How can we solve this problem?

One way is to remove gender from our creations altogether. According to an interview with Quartz, “[Dror Oren, the chief product officer and co-founder of Kasisto, makers of a banking chatbot that can solve about 80% of online customer interactions] believes gendered bots are giving way to ‘robot-specific’ identities without such clear gender lines. ‘It never pretends to be a human, and the lines are never blurry,’ he says.

Image courtesy of #WOCinTech Chat || wocintechchat.com

“Kasisto says it’s found its bots are most effective at answering requests and positive customer interactions when their personalities stick to their nature as an artificial intelligence, rather than mimicking human conventions.“ When designing Kai (the banking chatbot), Oren says that he “wanted it to be personable, but not a person — I wanted the bot to express itself as a bot.”

Therein lies the challenge for designers. We’ve spent the majority of our lives interacting with humans rather than robots. Our task is now two-fold: we have to create an entirely new way of interacting and communicating while remain aware of (and trying to prevent the furtherance of) the stereotypes and bias present in our daily lives. We’ve compiled a short list of tips to keep in mind when you’re (hopefully) asked to weigh in on the design of a new AI-powered tool.

“By assigning gender to these AI personalities, we may be saying something about the roles we expect them to play.”

Tips for avoiding AI gender pitfalls

Remember that AI assistants and chatbots do not inherently have a gender.

If we strip these tools of the names and voices added by their human creators, there's nothing there that requires them to be "he" or "she," other than our own assumptions. These creations are tools that help us solve problems and accomplish tasks. Nothing more and nothing less. Keep this in mind when you’re scripting interactions and responses. What’s the quickest way to be the most helpful?

Include multiple points-of-view on your design team.

“This is fundamentally a data problem. Algorithms learn by being fed certain images, often chosen by engineers, and the system builds a model of the world based on those images,” says Kate Crawford, a leading researcher, academic and author who has spent the last decade studying the social implications of data systems, machine learning and artificial intelligence. We need to ensure that our tools and assistants are being designed by diverse groups of people from varying backgrounds and walks of life.

Make your new AI more robotic and less human.

If all else fails, think in binary. Have your bot express itself as a bot. Your creation can be authoritative and professional (heck, even friendly) while remaining firmly in the artificial intelligence camp. Stick to the facts, be helpful without being condescending, and keep it businesslike.

As we move into a new age of automation, technology will continue to assist us in many ways as we move through the world. It’s important to remember that that's a role that could be filled by anyone—a man or a woman—or even a digitally-enhanced, and genderless, AI.