The study of the interaction of humans and robots on the example of robotic garbage cans

The most interesting thing about simple robots is that they encourage humans to project their needs and desires onto them, allowing us to do most of the hard work of human-robot interaction (HRI). In a report presented at the HRI 2023 conference, researchers from Cornell University studied what happens when random strangers interact with a pair of robotic trash cans in New York.

These are standard 32-gallon tanks that can be found on the streets of the city, fixed on redesigned hoverboards in which the standard board was replaced with ODrive V3.6, to which Raspberry Pi 4 was connected. The roles of robots were delimited using a standard municipal color scheme: blue - for recycling, gray - for landfill.

It's interesting how much human-machine interaction is going on around these robots, which, in fact, do not have explicit HRI functions, since they are literally garbage cans on wheels. The video notes that they are controlled remotely by people, so most of the expression they demonstrate based on movement probably comes from a person, regardless of whether it is intentional or not. These remote-controlled robots move quite differently from autonomous mobile robots that perform slow movements along predictable trajectories.

One of the features that the researchers found is that people do not seem to trust autonomy very much, associating it with poor navigation and social mistakes. In other words, people were more likely to think that a robot was controlled by a computer if they saw it getting stuck, bumping into obstacles, or ignoring people's attempts to attract its attention to itself.

Initially, they encountered this opinion when a less experienced robot driver experimented with control, actively moving the robot along strange trajectories. An observer nearby claimed that the robot "must be autonomous. He's behaving too strangely to be controlled by a human."

Because of the uneven surface of the sidewalk, robots sometimes got stuck. People willingly helped robots when they got into trouble. Some observers actively moved chairs and obstacles to clear the way for robots. Moreover, people interpreted the swaying back and forth movements as if the robots were nodding and agreeing with them, even if such movements were caused simply by surface irregularities.

Another interesting point is that people think that robots want to be "fed" with garbage and feel in some way obliged to give them something. When the robot passed by and stopped near the same woman for the second time, she said: "He probably knows that I've been sitting here long enough, and I have to give him something." Some people even found an excuse to generate trash to "satisfy" the trash can by rummaging through a bag or picking up trash from the sidewalk.

In one of the previous papers on the topic of HRI, it is described in a little more detail what this leads to: it turns out that people naturally attribute internal motivation (or the desire to satisfy some need) to the behavior of the robot, and this mental model encourages them to interact with the robot socially, "feeding" it or expecting a response of social gratitude. Interestingly, the role assigned to the robot by outside observers resembles the role of a beggar, when the robot asks for alms and expects gratitude for donations. This is in sharp contrast to human counterparts, where they offer help, and the passerby receiving it is expected to be grateful.

Report "Trash Barrel Robots in the City" presented this week at HRI 2023 in Stockholm, Sweden.