Ryan Siri, Alexa, Cortana, and Google Assistant — what is the common thread among these AI voice assistants? A striking similarity is that they are predominantly female.
Though some might argue that modern voice assistants now offer male voice options, the default settings often feature female voices, and even their names frequently carry a feminine association. This trend in AI technology reflects a broader societal pattern, and it has implications for how gender roles are perceived in everyday interactions with technology.
Voice assistants are now an integral part of our daily routines, with nearly 3 billion people using them to set alarms, check the weather, or find a nearby restaurant. As this technology continues to grow in usage, however, the decision to default to female voices has sparked criticism, with many suggesting that it reinforces traditional stereotypes of women as passive, compliant, and ready to serve.
But why do companies continue to rely on female voices for AI? In this article, we’ll explore the reasons behind this trend, the challenges of developing male voice assistants, and what can be done to reduce gender bias in AI.
Why Are Voice Assistants Predominantly Female?
A Preference for Female Voices
Research shows that humans tend to prefer female voices. Some psychologists even speculate that this preference starts in the womb, where a fetus is calmed by the sound of its mother’s voice. Other studies suggest that female voices are perceived as clearer and easier to understand, particularly in professional environments.
Historically, female voices have been preferred in various communication systems. For instance, during World War II, female voices were used in airplane cockpits because they could be distinguished more easily from the lower-pitched male voices of the pilots.
However, this preference is not without controversy. Studies challenging the assumption that female voices are inherently easier to hear or understand in all environments have debunked many long-standing myths. In fact, female voices are often subject to negative scrutiny. A quick Google search for “women’s voices are…” frequently yields the suggestion “annoying,” underscoring how societal biases shape perceptions.
The Lack of Data for Male Voices
One of the key obstacles to creating male-voiced AI assistants is the lack of existing data. Text-to-speech systems have historically relied on female voice samples, making it easier for companies to design female-based AI. This abundance of female data stems, in part, from the history of female telephone operators, which began in the late 19th century and established a precedent for women’s voices in communication technology.
Since female voices have been the industry standard for over a century, developers have a wealth of female voice recordings to use for training AI. In contrast, creating a male-voiced assistant would require substantial time and financial investment, with no guarantee of user acceptance. Thus, many companies have opted to stick with female voices for practical reasons.
The Challenges of Male Voice Assistants
Creating male voice automation has proven more difficult than anticipated, as seen in Google’s experience with its Assistant product. When Google introduced the gender-neutral “Assistant” in 2016, the goal was to offer both male and female voices. However, the speech recognition systems used to train the assistant performed significantly better with female voices due to the existing data biases.
Google’s team realized that adding a male voice would require significant adjustments to their system, which had been optimized for female voices. The process was deemed too resource-intensive, and the company ultimately decided to delay creating a male voice until they could ensure the same level of quality.
Reducing Gender Bias in Voice AI
While gender bias in voice automation is rooted in historical data and long-held social perceptions, there are concrete steps that the industry can take to create more gender-inclusive AI.
1. Investing in Advanced Machine Learning
Recent advancements in machine learning have made it possible to create more naturalistic male and female voices. Google, in collaboration with AI research lab DeepMind, developed a groundbreaking text-to-speech algorithm known as WaveNet, which can generate realistic human voices using fewer recorded samples. This technology allows companies to create both male and female voices with more accuracy and efficiency.
Today, users of Google Assistant in the U.S. can choose from 11 different voices, offering greater diversity in both gender and accent. Furthermore, new users are assigned a voice at random, without a default setting based on gender.
2. Establishing AI Standards
As AI becomes more pervasive, with an estimated market value of $267 billion by 2027, the industry must establish clear standards for how AI systems portray gender. Currently, most voice assistants are female by default, which can perpetuate outdated gender roles. By creating industry-wide guidelines, companies can develop more inclusive and balanced AI that reflects diverse gender identities.
These standards should involve input from diverse stakeholders across the AI community, ensuring that gender representation in AI reflects the complexity of human identity.
3. Data Transparency
To combat gender bias, AI companies must be more transparent about their data collection methods. This includes sharing the demographic makeup of their development teams, user preferences for different voice options, and any research into creating gender-neutral voices. Greater transparency will help uncover the sources of bias and foster collaboration toward more inclusive AI design.
4. Promoting Inclusivity in AI Careers
Increasing gender diversity within the AI workforce is essential to addressing gender bias. Currently, only 26% of data and AI roles worldwide are held by women, and the numbers are even lower for transgender and non-binary individuals. By expanding educational opportunities in AI for people of all genders and backgrounds, the industry can build a more diverse talent pool capable of tackling complex gender issues in AI development.
This can be achieved by encouraging underrepresented groups to pursue AI careers from an early age and by featuring diverse role models in AI education materials.
Final Thoughts
As voice assistants continue to integrate into our lives, it’s crucial to address the gender biases embedded in this technology. By investing in advanced machine learning, establishing inclusive standards, promoting transparency, and fostering diversity within the AI industry, we can work toward creating AI systems that are more equitable and representative of all people.
