Teaching robots to speak local dialects could make it easier for AI to exploit us

A team of researchers in Germany found mixed results when attempting to determine whether study participants would find an artificial intelligence system that spoke a standard language more or less trustworthy than one using a regional dialect.
A team of researchers in Germany found mixed results when attempting to determine whether study participants would find an artificial intelligence system that spoke a standard language more or less trustworthy than one using a regional dialect.

A team of researchers in Germany recently published a study examining the relationship between language and trustworthiness when it comes to human views on talking robots.

The question at hand was whether humans would find a robot more trustworthy if it spoke in a standard language or if they’d prefer it speak in a local dialect. 

A group of 120 German native speakers were recruited to find out. Study participants were shown videos of a small robot speaking in both a Berlin dialect and the more general standard German dialect. Respondents were then asked which seemed more trustworthy.

As it turns out, there wasn’t a single one-size-fits-all answer to the question. Per an article in the Frontiers scientific journal, the study’s lead author, Katharina Kühne, said people’s preferences were a bit more nuanced than the team expected:

“Surprisingly, people have mixed feelings about robots speaking in a dialect—some like it, while others prefer standard language. This made us think: maybe it’s not just the robot, but also the people involved that shape these preferences.”

Existing research shows, to paraphrase Nelson Mandella, that speaking to people in a language they understand gets the point across, but communicating in their own language speaks to their heart.

However, as the German team writes in their paper, so-called “accentism” can come into play during human communications. This means that, in some instances, listeners may find the general dialect more prestigious or competent and thus prefer it, while others may find a local dialect (that they share) to be comforting and thus preferable.

Per the paper:

“In summary, standard language-speaking robots were perceived as more trustworthy or likeable presumably due to the in-group bias and accentism, while according to other studies, participants preferred robots that spoke with a dialect.”

The discrepancy between viewpoints was attributed to more than just whether a person spoke a particular dialect. It extended to what particular device a given subject was using.

Those using mobile devices such as phones or tablets showed a tendency toward standard German speakers, whereas PC users who participated in the study showed some prevalence toward the dialectic speaker.

Related: Beyond iPhone: Experts grapple with Apple’s future in our AI-powered world

 The researchers suggest that users on the go or with smaller screens could have higher cognitive loads and, thus, exhibit a greater likelihood to mistrust the more colloquial voice.

Ultimately, this research demonstrates that under the right circumstances, certain users could be more or less inclined to trust a machine based on nothing more than the dialect it speaks in.