Virtual assistants, from Siri to Cortina, are gendered female-by-name and voice (although, after outrage, consumers were allowed to choose a male-voiced Siri). Along with the gender come the stereotypes: Siri was designed as a slightly sassy, demure female assistant. Technically, Siri, Alexa, and Cortana identify as genderless, but researchers have found that these virtual assistants spend much of their time fending off sexual harassment. Gender matters.

Humans tend to gender machines (because, in Western cultures, gender is a primary social category). Designers say that if they don’t embed a gender, users will. What to do?

State of the art machine translation systems, such as Google Translate or Systran, often default to masculine pronouns: “he” said. How can we create algorithms that determine the gender of each person in a text—a deep “fix” that enhances the quality of translation? When trained on historical data (as Google Translate is), the system inherits bias (including gender bias). Importantly, Google Translate will carry over to shape future technologies: our devices, programs, and processes shape human attitudes, behaviors, and culture.

Possible fixes include: Creating filters to filter out bias and developing mathematically rigorous definitions of fairness that guarantee fairness.

As the world’s population ages, robust new markets for assistive technologies are emerging. Elderly men and women may have different needs. Designing with these distinctive needs in mind helps engineers develop technologies that may yield broader markets.