Robot Language Adaptation: Greater AI Exploitation Risk

A group of researchers in Germany conducted a study to investigate the impact of language on perceived trustworthiness of talking robots. They wanted to determine whether people would trust a robot more if it spoke in a standard language or in a local dialect. The study involved 120 German native speakers who were shown videos of a small robot speaking in both a Berlin dialect and standard German. Participants were then asked to indicate which voice seemed more trustworthy. Surprisingly, the results showed that people had mixed feelings about robots speaking in a dialect, with some preferring it and others favoring standard language. This led the researchers to consider the role of personal preferences in shaping trust in robots’ language capabilities.

Previous research has shown that speaking to people in their own language helps to convey the intended message effectively. The German team noted that “accentism” can influence human communication. Listeners may perceive the standard dialect as more prestigious or competent, leading them to prefer it in certain situations. Conversely, they may find comfort in a local dialect that they can relate to and therefore prefer it.

The researchers concluded that robots that spoke in standard language were generally perceived as more trustworthy or likeable, possibly due to an in-group bias and accentism. In contrast, other studies have shown that participants preferred robots with a dialect. This discrepancy suggests that trust in a robot’s language is influenced by more than just the choice of dialect; it also depends on the individuals involved.

Interestingly, the researchers also found that the type of device being used affected participants’ preferences. Mobile device users, such as those using phones or tablets, tended to trust standard German speakers more. In contrast, participants using PCs showed a preference for the dialectic speaker. The researchers propose that users on the go or with smaller screens may have higher cognitive loads, making them more likely to be mistrustful of a colloquial voice.

This study demonstrates that people’s level of trust in a robot can be influenced by the dialect it speaks. Depending on individuals’ preferences and the specific circumstances, certain users may be more or less inclined to trust a robot solely based on its choice of language. This finding highlights the importance of considering language preferences and user context when designing and developing human-robot interactions.