A science team from Lund University is testing how humans respond to robots. Their number has recently risen sharply worldwide.
Epi, the robot, has its own page on the internet. There he is seen, wide-eyed and innocent-looking, patting the head of a torso version of himself with an artificial hand. A plush bear and cat are sitting on the table in front of him. The robot with its childlike features almost looks like a toy itself. You have to like him somehow, despite all knowledge that it is a machine. But do you trust the robot?
A team of scientists from Sweden has looked into this question and found out that language plays a decisive role: when the robot talks, people trust it more than when it doesn’t say anything. Researchers from Lund University’s Cognitive Science Robotics Groups developed the humanoid robot themselves to test its motor and social skills, and in particular its interaction with humans. Among other things, Epi is able to move his head and arms, to grasp things, his pupils can dilate or contract, the color of the iris can be changed – and he can speak.
For the experiment, the participants in the study watched videos of the robot in which it showed either error-free or incorrect behavior when handling various objects and spoke or remained silent. The more than 200 test persons were then asked to fill out questionnaires and state how much they trust the robot.
The result was clear: the robot enjoyed the greatest trust when it made no mistakes. However, the subjects also overlooked mistakes when the robot was able to speak. In fact, they trusted it almost as much as the bug-free version in this case. In the case of muteness, on the other hand, the same mistake led to a loss of trust. The researchers suspect that this behavior has something to do with the fact that people generally regard speaking as a sign of intelligence. Previous work on robots has suggested that trust in human-like machines may depend on how intelligent they are perceived to be.
The researchers from Lund support the theory that the robot’s intelligence, which is assumed to be based on its language skills, was sufficient to “mask” erroneous behavior, as study leader Amandus Krantz says. In a follow-up study, the scientists now want to investigate what happens when people interact directly with the robot and not just watch its behavior in videos. They also want to find out which other factors can influence trust in robots – such as gazes or dilated or contracted pupils.
The experiments at Lund University are much more than scientific gimmicks. Robots are not only used in research, but also in many areas of life – for example as vacuum cleaners in private households, in medicine, where they are even allowed to operate – under supervision – as fitters in the manufacture of vehicles or as warehousemen in factory buildings , to name just a few examples. The International Federation of Robotics recently reported that half a million robots will be put into operation in industry worldwide in 2021. According to the Wall Street Journal, this number exceeds the previous record from 2018 by 22 percent. According to the report, a total of around 3.5 million industrial robots are now in use worldwide.
This development is also progressing at a rapid pace in retirement and nursing homes, driven by demographic change, the lack of human nursing staff and technical innovations alike. To a far greater extent than in industry, many ethical questions arise here. Robots with human-like appearance and behavior such as Epi from Lund are also used here, more so than in other areas at present.
Not only in nursing homes, but also in your own four walls, robots will in future be able to help people who can no longer get by on their own. In order to determine what skills these robots need to have, the University of Bremen has just set up a new research laboratory equipped with a kitchen and furniture like in an apartment. Ideally, robots should help people with physical disabilities as soon as possible, for example by setting the table, making bread rolls, stocking the refrigerator, doing the dishes or preparing simple meals, according to a statement from the University of Bremen. For robots, these activities are “complex processes whose error-free completion must first be laboriously developed”. “We must not adapt the environment to the robots, as happens in factories or logistics centers,” explains Michael Beetz from the University of Bremen, one of the leading international scientists in the field of artificial intelligence: “We must develop the robots in such a way that they can find their way in the everyday human environment in order to help efficiently there.”
Based on their research results, the research team from Sweden suggests that more importance should be attached to the development of future robots with language skills and verbal interaction. It doesn’t have to be limited to humanoid robots. One thought: if the machines talk, trust can be gained to prevent them from being used again once they have not worked properly.
