Research Description
This is a rough version of the original project idea. I have kept it in its original form in the interest of historical preservation. It amazes me to look back at this and see how much it changed over the course of the summer.
When Lewis the Robot was out taking pictures in the field, something odd was noticed, not in Lewis’s behavior, but in the behavior of those that interacted with him. Lewis is a large red robot, standing about five feet tall, and he bears a striking resemblance to a gigantic red garbage can. In a crowded room, one would assume that he would be somewhat noticeable. However, what was observed was quite the opposite – after a few moments of interacting with Lewis, most people tended to ignore him. This fire-engine red cylinder somehow managed to fade into the background. How did this happen? More specifically, what sorts of behaviors was Lewis doing that allowed him to become “invisible”? This experiment will try to determine these behaviors by having human participants complete a task that requires them to move from room to room through a hallway, and watching their responses to Lewis as he rolls down the hallway past them.
Our experiment will be set up as follows: each participant will be told that they are helping with an inventory experiment. A robot will be asked to complete an inventory task, and its time will be compared with an average human time. The participant is assisting with determining the average human time. In reality, this experiment will be measuring the human’s time completing the task without interacting with anyone in comparison with sharing a hallway with another human and sharing a hallway with Lewis.
To begin with, each participant will be asked to fill out a questionnaire. We want to determine what each participant’s familiarity with robots is before beginning the experiment, to see if there is any correlation between familiarity and reaction to Lewis. We want to determine –
After having filled out the questionnaire, the participant will be led to the start room and asked to complete a simple task that will help the participant learn the layout of the experimenting area. Right now, this task will probably be visiting each of the three storage rooms, counting the number of boxes in each, and entering that number into a laptop in the start room. The laptop will be programmed to only display one task at a time – “Go to room A, count the number of boxes in there, and enter that number here”, for example – and will not go on to the next task until it has received input from the user. Each task will require the participant to return to the start room. In this way, the time it takes to complete each task can be measured with a decent degree of precision. Once these tasks are completed, the real experiment begins.
Now the laptop will instruct the participant to complete a series of tasks that will take the participant from room to room. For example, one such task might be “Go get box #4 from room A and bring it here,” the next might be “Take this box to room B and return here,” and so on. The order of rooms to visit will repeat three times (for example, ABCABCABC, or CABCABCAB). During the first cycle, a participant would pass either Lewis, another human, or no one. During the second, the participant would pass one of the two not passed during the first cycle. During the last, the participant would pass the last entity. For example, during the first cycle, a participant will pass no one in the halls; during the second, he will pass a human, and during the third, he will pass Lewis. The participants will be divided into six groups to cover all the different orders of entity passing.
Both the human and Lewis will follow the same path through the hallway, passing the participant at hopefully the same point. If Lewis is truly “invisible,” then any delay in completing the tasks should be no different than the delay found when passing a human. Interaction between Lewis and the participant will be kept to a minimum; we are trying to find out if Lewis can be completely unnoticeable, and it is difficult to avoid noticing something that is taking your picture or asking how the weather is or something. The tasks the participants will be asked to complete can be as simple as go here and bring this box back, now take it to this other room and come back, and so forth. If it turns out that there is not enough time being allowed for the robot to be reset for another interactive pass, then one of the tasks could be to open the box (in the start room) and count the contents. The robot doesn’t necessarily have to pass by the participant every single time a room change is made; two or three passes would probably be enough. We may put some article of clothing on Lewis, like a bow tie, that would be unobtrusive, but would us to determine how well the participant noticed the robot.
After the experiment concludes, each participant will be asked to complete an exit questionnaire. We will want to ask questions such as:
And other such awareness questions. Perhaps a poster will hang in one of the rooms and a question could be asked about what was on the poster, to determine the participant’s overall level of awareness.
With any luck, this experiment can provide a baseline for other types of experiments in the future. Depending on the results we obtain, Lewis’s behavior can be changed to make him less noticeable. However, as we do not know what kinds of findings we will gather, it makes sense to start with Lewis just rolling down the hallway past the participant.
Questions to think about:
Will telling the person about the robot skew the findings in any way? Could the experiment be done without mentioning the robot? Is there any other way to do that and still get the answers needed on the entry questionnaire? Is the entry questionnaire needed? (Looking for correlations between perception of smartness and noticability.) How feasible would it be to run the experiment with twelve groups, half of them meeting the robot beforehand and half going in blind? What other questions can test for awareness? What sorts of tasks would work well for the participant to do? Is there anything connected with that that we should be careful of? Will having something to notice on the robot (like a tie or a sign) make a difference? How important is the order of the questions on the questionnaires? How anonymous should we make the questionnaire? What about a human accompanying the robot? (But do we even want to bother with that?) Should the subjects be told that their task will be timed? Should the starting task be timed for a baseline?