Robots Can Command Compliance, But Less Effectively Than Humans

A study conducted by Polish researchers and published in Cognition, Technology & Work examined robots' performance as managers in the workplace. The research found that while robots are capable of commanding compliance, they are less effective than humans in this role. 

Konrad Maj, Ph.D., a psychologist at SWPS University and director of the HumanTech Center for Social and Technological Innovation, emphasizes the importance of considering the psychological impact of integrating robots into the workplace. This includes factors such as the perceived authority of robots, trust in them, and potential resistance to following their commands.

Robot as an Authority Figure?

With advancements in robotics, robots are increasingly being placed in positions of authority, such as in law enforcement, healthcare, and education. Researchers were interested in how society would respond to robots in these roles. The study found that, while the level of obedience to humanoid robots was somewhat lower than that to humans (63 % vs. 75 %), people still showed a significant degree of compliance with robots as authority figures.

However, the study also found that individuals may be less motivated when overseen by machines. The experiment showed that people performed tasks more slowly and less efficiently when a robot was supervising them. Maj notes that if automation is not thoughtfully planned from a psychological standpoint, it may not always lead to higher productivity.

Course of the Study

The study, conducted by scientists from SWPS University, including Professor Dariusz Doliński, Magda Franjo, Tomasz Grzyb, Ph.D., and Konrad Maj, Ph.D., took place in the university's laboratory. After being welcomed to the lab, participants were randomly assigned to two groups: one supervised by the Pepper robot and the other by a human experimenter.

The task involved modifying computer file extensions. The robot or the human experimenter provided verbal encouragement whenever participants showed signs of reluctance, such as pausing for more than 10 seconds.

Under human supervision, it took participants an average of 23 seconds to modify the extension of a single file. However, in the robot-supervised group, it took 82 seconds on average. In the human-supervised group, participants altered 355 files on average, while in the robot-supervised group, this number decreased by over 37 %, with only 224 files modified.

Human-Robot Relations

The experiments demonstrate the complexity of human-robot interactions and the expanding role of robots in society. Research shows that robots with anthropomorphic features tend to be trusted and obeyed more, but too much anthropomorphization can result in the "uncanny valley" effect, reducing trust and comfort.

Maj explains that if a machine has human-like traits but also shows imperfections, it causes cognitive conflict, leaving people uncertain about how to treat it. Additionally, there is an emotional conflict: fascination and affection combined with fear and disappointment. Evolutionary theory suggests humans are instinctively programmed to avoid potential threats, and robots that mimic humans but aren’t perfect may evoke this sense of danger, as they can seem abnormal or unwell.

While giving robots human-like features can make them easier to work with, as people are accustomed to interacting with humans, this may create unintended consequences. If robots become too similar to humans, the boundaries between them may blur. People might form emotional bonds, demand rights for robots, or even marry them in the future. This could lead to a division among people and increase misunderstandings and aversions.

As Maj points out, robots designed to be personalized, constantly available, and empathetic in communication might not align well with human social dynamics.

Journal Reference:

Maj, K., et. al. (2025) Comparing obedience and efficiency in tedious task performance under human and humanoid robot supervision. Cognition, Technology & Work. doi.org/10.1007/s10111-024-00787-1