Grace Frances (main photo) is a conversational copywriter at VERSA. In this guest post, Frances says the advancement in AI and robotics throws up new and unforeseen problems in terms of genders..
Gender is a hot topic within AI, and rightfully so. Voice technology is one area where this is coming to the fore, with the recent push for genderless voice assistants bringing up important questions about gender and representation. However, when thinking about representation, gender isn’t the only thing to consider.
Beyond the binary
Gender exists on a nuanced, colourful spectrum – a reality which isn’t yet represented in AI. To change this, companies have been working on genderless voices such as The Genderless Voice’s Q and Accenture’s Sam, offering alternatives to the typical female/male gender binaries represented in voice assistants.
These gender-neutral voices have been synthesised from the voices of people who identify as neither female or male, and were designed to ensure non-binary people are represented in emerging technology.
So why do we assign AI with gender in the first place, and is it necessary?
Gender trouble
AI assistants are often gendered in order to make the user feel comfortable and make the AI seem more human and less robotic. For the most part, they are assigned the female gender by default, however more recently tech companies have created male voice options.
Many studies have shown that voice assistant users prefer the sound of a female voice to a male voice, a phenomenon that could have quite Freudian undertones – harking back to associating the female voice with the nurturing qualities of a mother. It has also been argued that women’s voices are easier to understand compared to male voices due to their higher pitch, however this myth has been debunked.
Our tendency to anthropomorphise AI sees us referring to these non-human entities by the gendered pronoun of ‘she’ (or occasionally ‘he’ if there is a male voice option) rather than ‘it’. However when asked what gender they are, virtual assistants usually respond along the lines of ‘As an AI, I don’t have a gender’. Despite this, they are usually recognised as female due to their voices and female names.
Genderless voices have emerged as an alternative to the old-world gender clichés that hardly need repeating here. While there is still some work to be done to make them sound more natural and less robotic, gender-neutral voices are an important step towards diversity in gender representation.
Finding the right voice
The reality of designing voice technology is complex, and there is a risk of over-simplifying the discussion down to pure gender dynamics.
There are many factors to consider to ensure that interactions with voice assistants are as warm, empathetic and natural as possible. The complexities of voice persona, tone, pitch, and the nuance of pauses and turn-taking all need to be taken into account – all important aspects of thoughtfully crafted conversational design.
Language and accents are another topic that needs careful consideration if we are to create voice assistants that can understand users and speak their language. Amazon Alexa currently speaks eight languages, English, Spanish, French, German, Hindi, Italian, Portuguese, Japanese as well as nine dialects. However, there is still work to be done to represent more diverse languages and ensure that English isn’t always the default.
These elements become even more crucial as AI technology moves away from on-rails voice experiences and towards actual conversations. As we progress into this free-flowing conversational terrain, the need to create human-like interactions that properly understand and represent the user becomes paramount.
A future of multiple voices
Genderless voices point to a future of diverse voice applications beyond Alexa, Siri, Google Assistant and Cortana – personalised and bespoke voices designed specifically for the clients’ needs. This is being made possible by an increasing number of voice synthesis tools on the market that are making it easier for companies to create their own custom voices.
There is infinite potential for growth in this emerging area, and we can expect to see a greater range of voice applications on the horizon. So while genderless voices may not become the new default, they can certainly offer an important alternative.
Along with understanding the finer points of conversational design, greater diversity and localisation regarding language and accents; gender is absolutely an important consideration when creating voice technology. To create truly inclusive AI, diversity of representation in all aspects is key.
