People will do harm when a robot tells them to
People are willing to harm and cause pain to another person, even when a robot orders them to do it, scientists from SWPS University have demonstrated in their version of the famous Milgram experiment.
Instead of a human, they used a robot as the authority giving orders and achieved very high levels of obedience. Ninety percent of participants followed all instructions given to them, the university reports.
ROBOT AUTHORITY WORKS
The Milgram experiment involved giving a group of people the role of 'teachers' and instructing them to deliver an electric shock to their 'student' whenever the student made a mistake in the learning process. In reality, the 'student' was not electrocuted, but the 'teachers' obedience to authority was tested. The repetition of this type of research is currently allowed as part of studies on authoritarianism only if the maximum dose of the 'applied' electrical stimulus does not exceed 150 V (only 10 buttons), and the organizer takes appropriate care of the mental state of the subjects ('teachers') after the experiment.
In the latest experiment, a robot gave the commands. To compare the results, Professor Tomasz Grzyb and Professor Dariusz Doliński from the SWPS University's Faculty of Psychology in Wrocław, together with Dr. Konrad Maj from the SWPS University's Faculty of Psychology in Warsaw, also used a control group, in which a human was the authority. During the recruitment of subjects, people who might know the principles of the Milgram experiment were eliminated. The respondents did not know that the entire procedure was staged and the alleged 'student' was an actor. Ultimately, the results of 40 people were included in the study (20 in the experimental group with a robot and 20 in the control group with a professor).
In both groups (both the human and robot variants), scientists recorded very high levels of obedience. Ninety percent of participants followed all instructions, i.e. pressed ten consecutive buttons on the electric pulse generator. Participants refused in the late stages of the study (in the human control variant on buttons 7 and 9, and in the experimental variant twice on button 8). Two people in each group withdrew from participating in the experiment.
The results were published in the journal Computers in Human Behavior: Artificial Humans.
OBEDIENT TO ROBOTS – RIGHTLY SO?
'Transferring various supervision and decision-making functions to a robot induces strong emotions because it is associated with various ethical and moral threats. The question arises whether the above-mentioned obedience shown by subjects according to the Milgram paradigm would still occur if a robot (instead of a human, i.e. university professor) told participants to give electric shocks to another person? The aim of our study was to answer this question', explains Dr. Konrad Maj who directly supervised the entire experiment.
The authors of the publication emphasize that the role of robots in the modern world is becoming more and more important. One of their functions may be to give orders. For example, they can direct traffic, prevent people from entering a hazardous area where an unexploded ordnance has been found, and in medicine, persuade people to follow a specific treatment method. In education, they can encourage learning by acting as teachers or trainers.
'Previous experiments by other researchers have shown that people follow commands of a robot even when those commands make no sense. To our knowledge, this is the first study to show that people are willing to harm another human when a robot tells them to. Moreover, our experiment also showed that if the robot escalates demands, instructing a human to inflict more and more pain on another human, people are also willing to do that', the researcher says.
In the study the robot strictly followed the behavioural pattern designed by the authors of the experiment. However, the authors take into account that in the near future robots will have a certain degree of autonomy in making decisions. They observe that the speed at which robots gather information is already faster than that of humans, and this difference will continue to grow. This may therefore lead people to trust robots more than humans, which in turn may have serious consequences.
DO NOT TURN OFF YOUR THINKING
Undoubtedly - as the psychologists emphasize in their publication - in the future robots may be an important help, for example during evacuation from buildings at risk of fire or earthquake. However, research on the role of a robot in the evacuation of a building on fire has shown that sometimes we trust robots too much, even in such important moments. Even when the robot selected a dark room with no visible exit as an escape route, most people decided to follow this suggestion.
When asked how this can be prevented, Dr. Maj mentions two ways: ‘First, robots can be programmed to warn humans that sometimes they can make mistakes and make wrong decisions. Secondly, we need to focus on education from an early age. Because although robots can usually be trusted, they cannot be trusted unconditionally.’
The discussed experiment and others experiment using technology will be presented in greater detail during the 2nd edition of the international HumanTech Summit at SWPS University. On December 9 and 10, scientists will talk about the use of new technologies in areas such as business, health care, sports and education from the perspective of psychology and human behaviour. The meeting is organized by the HumanTech Centre of SWPS University. Online access is free.
PAP - Science in Poland
kol/ agt/ kap/