Whether or not humans trust robots is greatly influenced by ``whether the robot speaks words''

It has been suggested that how much humans trust humanoid robots depends on how intelligent they think they are. Research has revealed that this ``how intelligent humans think robots are'' is greatly influenced by what robots talk about.

[2208.13688] Using Speech to Reduce Loss of Trust in Humanoid Social Robots
https://arxiv.org/abs/2208.13688

Can a robot's ability to speak affect how much human users trust it?
https://techxplore.com/news/2022-09-robot-ability-affect-human-users.html

This was revealed in a study by Amandus Krant et al. of Lund University in Sweden.

Krant et al. have previously investigated how incorrect gaze movements affect the reliability of humanoid robots. We found that there was no decrease in reliability only when uttered short utterances.

Therefore, we conducted this research based on the idea that ``the robot's utterance elements make us feel intelligent, and the resulting change in the degree of trust is enough to mask the decrease in trust due to incorrect behavior.''

The 227 people who participated in the study watched videos of multiple humanoid robots, and then rated the robot's ``reliability'', ``likability'', ``animation (whether it is close to humans)'', and ``did you feel intelligence''. was evaluated.

The robots that appeared in the video had some with flawed behavior and some without flaws, and some were talking and some were non-speaking.

As a result of the evaluation, it was naturally the robot that behaved flawlessly that won the trust. On the other hand, we found that even robots with clearly flawed behavior could almost completely mitigate the loss of trust caused by mistakes. In other words, Krant et al.'s hypothesis was correct.

According to Mr. Krant, although there are similar studies on 'speaking robots', there are many studies that focus on 'robot's utterance content (mainly apology)', and there are many studies on the presence or absence of speech ability and reliability from humans. The research is said to be the first.

This result could be useful for manufacturers of consumer robots, such as robot vacuum cleaners, who want to reduce the number of robots that are not used due to operational errors.

In addition, this research was conducted online, and if a human and a humanoid robot face each other, the results may be different. Krant says he plans to study.