Not-So-Natural Language
U.S. adults and children adapt their speech when talking to a real person or a smart speaker, according to new research at the University of California-Davis.
The study found that participants produced more effortful speech when talking to a device (longer duration and higher pitch).
These differences varied by age, with children producing even higher pitch when speaking to technology, suggesting a stronger expectation to be misunderstood by the system. It’s not unfounded, as other research found that children are misunderstood by automatic speech recognition (ASR) systems at a higher rate than adults. For example, a voice assistant responded correctly to only half of queries produced by children ages 5 to 10. In another study, speech produced by children around age 5 was accurately transcribed only 18 percent of the time by the best-performing ASR systems.
When talking to technology, adults often make their speech louder and slower, perceiving it as less communicatively competent than humans. It also mirrored the changes they often make when speaking in challenging listening conditions; in the presence of background noise, speakers produce louder, slower, and higher-pitched speech.
The research postulated that when people detect a sense of humanity in a technological system, they automatically transfer human social rules and norms to the device in a form of anthropomorphism, whereby people attribute human qualities to nonhuman entities.
On the other hand, routinized interaction theories propose that people develop scripts for how to interact with technology that differ from how they engage with another person.
The adjustments appear to be in an effort to improve intelligibility through communicative barriers, such as the presence of background noise.
In the study, the researchers found that speakers also increased duration and pitch when repairing an error; when communication went smoothly, they decreased both of these features.
Further highlighting the theory of anthropomorphism, unrelated research from Barna Group found that most U.S. adults try to be polite when interacting with AI tools like Amazon’s Alexa and Apple’s Siri or chatbots like OpenAI’s ChatGPT. This might be linked to a belief that respectful interactions with AI can lead to more positive experiences, reflecting the social dynamics they apply in human interactions.
Among those who use generative AI at least monthly, the following trends were observed:
- Three in four Millennials and Gen Xers try to be polite when asking a question or giving a prompt, along with 65 percent of Boomers.
- Two in five Millennials say “thank you” when AI answers questions.
- One in three people feel they are doing something wrong if they use an unkind tone of voice.
- Among U.S. adults who use digital assistants, one in five feels these tools might be capable of having emotions.
- Millennials tend to show the highest levels of respect and dignity, while Boomers find that less important.
These trends point to a deeper, more complex reality: Whether consciously or not, people are inclined to humanize AI.
“As AI models evolve, there’s an opportunity to shape them in a way that mirrors how we pass down values to our children—instilling the values, morals, and mannerisms we hold dear,” says Steele Billings, head of Gloo AI. “Of course, AI is not human and can’t be ‘discipled’ like a person, but these models are still in early development and are in ‘training.’ Although there’s little research on the impact of being polite to AI, it’s important to recognize that every interaction helps guide how these systems develop and respond.”
A similar study found that 45 percent of U.S. adults say they regularly use technology like smart home devices, and 42 percent use in-device assistants like Siri.
“As AI becomes a more common component in the rhythms of daily life, many Americans don’t even realize they’re engaging with it,” says David Kinnaman, CEO of Barna. “From personalized movie recommendations to predictive text that finishes our thoughts, AI is becoming a companion in the background of everyday tasks.”