The first robotic voice generator was the Voder. It was more like a musical instrument than an automated voice, though — the operator could change the pitch of the voice, add vibrato, and even make it sing.
The Voder was the basis of human ideas about robot voices for many years, if we judge by how robots sounded in movies.
That’s a pretty good way to judge, actually, since few of us heard robots speaking in daily life until quite recently.
But nowadays plenty of robots speak, so makers of automated machinery have to make decisions about how robot voices should sound.
One study found that people were able to decide how a robot looked based on a voice. The voices were male and female and neutral, with more or less mechanical vs.natural sounds, and local or foreign accents.
Not only did people match photos of robots with voices with confidence, but they were pretty consistent, as well. Robots with facial features were matched with natural human voices more often, while round robots were more likely to be assigned a female voice.
The subjects all accepted the voices — all human voices, though some were altered to give a more mechanical sound — as robot voices. They used the information in the voices to make decisions about the robots before seeing any pictures, and had no trouble matching the voices to the “faces” of robots.
This suggests that robot voices matter in human reactions to robots.
Racism and sexism
The decision to make Siri’s and Alexa’s voices female, along with all the other automatic personal assistants, was made through the use of focus groups. Essentially, the companies found that people like having a nice, competent woman as a personal assistant. So that’s what they made.
In other words, the makers reinforced the gender stereotypes of their customers.
No research has been done on why no voice assistant has a recognizably Black or Hispanic sounding voice. However, research has shown that Alexa is not as good at recognizing African-American speakers’ words as white words. This probably just reflects lack of experience.
The solution? Makers should use more diverse data to train their voice recognition software.
The uncanny valley
A possibly bigger issue is the question of whether robot voices should sound like human voices. Is it deceptive to use a very natural sounding human voice for a robot, or does it just make the robots more comfortable for the humans?
At this point, there are questions about the legal implications of having a human interact with a robot without alerting them to the fact that it is a robot. It’s not a problem with Pepper or Alexa, but Google is working on helpers that make phone calls. Is it ethical to allow someone to believe they’re having a conversation with a human when they’re not?
These concerns have caused many observers to say that a robot should always sound like a robot.
On the other hand, research suggests that people generally prefer humanlike voices in medical robots.
Do you want your industrial robots to speak in a particular kind of voice? Probably not. Unlike social robots, industrial robots are mostly valued just for getting their work done. If your Rexroth motion control system is not getting that work done the way you want it to, contact us. We specialize in Rexroth drive and control, and we specialize in factory repair and reman.