How may I assist you?
Did you know that 3D virtual assistants are brown-haired woman in their thirties?
A year ago, I reviewed potential Hubware competitors in the customer service and AI space. After five similar examples of what a virtual assistant should be, a singular pattern clearly emerged (and it was kinda scary).
Many early AI companies, some that have been around since the early 2000’s, chose to represent their assistants as 3D avatars.
Their common founding characteristic is they all use Natural Language Understanding (NLU), or the capacity to “understand” sentences. The NLU technique — what understand means is still not clearly defined today — was used as a proxy to deal with complex customer conversations. In fact, it turns out that this is not an effective proxy, because understanding is not even half the job of talking with someone (you also need to formulate an answer and reply).
Because the conversational technology, back in the day and still today, doesn’t really work, I guess that people thought having a realistic image representing a human might help the conversation flow a bit better.
But, as you can see, another common shared trait amongst 3D avatar companies is that they all have huge esteem for brown-haired women. These Gynoids — aka THE assistant — all look alike. That’s scary n°1.
To be honest, these Gynoids aren’t too different from the average Customer Service Representative that you might meet doing business today in France. A thirty-something woman who is generally overqualified and underpaid.
But the fact that a whole industry chose to represent their virtual assistant as a brown-haired woman is hard to explain. The obvious issue is that this bias creates:
a perception that some think might be having a negative effect on society in training everyone to think of women as assistants
— Clayton Moore, Digital Trends (link). Scary n°2.
All of this could be just coincidence or anthropomorphism, since virtual assistants, artificial intelligences, or robots, are not men nor women, they are just zeros and ones.
But, virtual assistants talk, so you have to consider the choice of a voice. In a study :
In 1980 the U.S. Department of Transportation reported that several surveys among airplane pilots indicated a “strong preference” for automated warning systems to have female voices.
Even though empirical data showed there was no significant difference in how pilots responded to either female or male voices.
Several of the pilots said they preferred a female voice because it would be distinct from most of the other voices in the cockpit.
— From Adrienne Lafrance, TheAtlantic (link)
The conclusion is probably that few men pilots have a falsetto register or that there are few women pilots because they are all assistants. Scary n°3 on my scorecard.
There are many implications behind the example of biased 3D avatar that suffer from sexism or a lack of diversity due to practical valid reasons, but in the end a choice has to be made, at least for the default voice or other characteristics of AI that will be integrated in software. So, what voices would you choose for your 3D avatars?
What choices will we collectively make for our emerging AI creations? This is a nice open question that materializes the challenges that our society is facing with new generation tech, especially AI.
In my opinion, these type of questions — tone of voice, traits of character such as humor or empathy, biases in learning — should not be treated by engineers or business people. Biases in tech should be treated by a new kind of team, the one that would think about social acceptance and social utility of products.
Because I don’t want kids to believe women are here to be given orders* (How millions of kids are being shaped by know-it-all voice assistants).
*This is clearly scary n°4.
