ABSTRACT
To test our hypotheses, we conducted a 3 x 3 laboratory experiment. The experiment was a between-subject design, manipulating human likeness (human, human-like robot, machine-like robot) and status (subordinate, peer, supervisor) with the human condition as the baseline. Each participant was asked to collaborate on a task with a confederate who reflected one of the nine cells in the design. The confederate used the same script for all conditions and was unaware of the status manipulation. In the robot conditions, we used a Wizard of Oz approach in which the robot was teleoperated, appearing to be operating autonomously. The same man teleoperated and spoke for the robot in the two robot conditions, and he acted as the human confederate. The experiment was videotaped with cameras suspended from the ceiling of the experimental lab. 3.1. Participants
Participants were 292 students recruited on a university campus, randomly assigned to condition, and paid for their participation. The mean age of participants was 20.51 years old. Fifty-nine percent of the participants were women. Because we thought it was important that participants believed that our robots operated autonomously, the last question we asked those in the two robot conditions was how they thought the robot worked, from a technical standpoint. Forty-two (21.5%) of the participants who worked with one of the robots expressed suspicion about whether or not the robot was autonomous. Suspicious
participants were approximately equally spread between the human-like and machine-like robot conditions. When these cases were removed from the analysis, there was no effect on the pattern of results, so the analyses we present here include data from all participants. 3.2. Tasks and Procedures
In the experiment, we asked participants to work with a partner in a parts depot for a company that develops innovative remote-control devices. They were told that their job was to collect the parts required to assemble various objects that would be assembled by another team of workers. The task entailed working with the confederate to jointly collect objects from a list, put them into bins, and take the bins to a table near the door. Participants were told that the confederate was familiar with the location of the parts and could carry the bins on its tray, but did not know what was needed and would not be able to collect parts. The division of labor helped establish interdependence between the participant and the confederate, and created a plausible story for why the confederate was not able to open drawers and pick up items. The task was also designed to capitalize on the unique capabilities of a robot (e.g., carrying materials, moving around a room, and remembering detailed information about the location of objects), although still making sense for a human confederate. Finally, the task was one that could be credibly conducted in ways consistent with each of the possible status conditions without any modification of the script.On arriving at the lab, participants were given a packet of instructions. They completed a brief demographic survey and then were provided detailed instructions on the task. After reading the instructions, participants were given four pages, each containing a list of items that were to be collected during the task; the items on each page were to be collected into a single bin, one bin per page. Then, they were introduced to the confederate (the human, the human like robot, or the machine-like robot) with the experimenter saying, “I’d like to introduce Chip, who will work with you on this study.” In all cases, the confederate entered the room ready to begin the task and, after acknowledging the participant’s greeting, started the task by asking, “What’s first on the list?” After that, the pace of the task was determined by the participant reading the parts lists. The participant read out the items from the list and the confederate identified which cabinet and which drawer the parts were in. The participant collected the parts from the drawers and put them into a bin on the tray that the confederate was holding. The confederate was prepared to answer questions about ambiguous items, if asked. The task took about 20 min to complete.
