Robot Surprise

Studying Perceptions of Robots In Group Activities

Overview

An experiment being run by the Human-Robot Interaction Lab at New Mexico State University, researchers were tasked with creating a scenario in which participants would bring 2 friends expecting to partake in a group decision-making exercise, only to be surprised to learn that a robot would also be assisting them in the task. The participant's overall perception to the robot after being placed into this condition would be measured by their initial reactions to both meeting and performing the task with the robot, as well as by their responses to the survey taken after the task had completed.

My primary responsibilities as a volunteer on this experiment included survey creation and task development (utilizing Qualtrics' javascript API), as well as running test sessions for both individual and groups of participants for the study. This study is currently ongoing. As such, the following information may be incomplete until the study has officially concluded.

Lost On The Moon Task

Upon meeting the robot and, depending on the condition, being given the choice whether or not to work with the robot, participants were brought into a separate room where they were asked to perform a decision-making task. The task was an adaptation of the "Lost On the Moon" task, where participants were presented with a list of items that they would need to prioritize for survival after crash landing on the moon. Participants were presented with item pairings, needing to decide which of the two was more important to bring on their journey, and then provide their reasoning. Participants would then get the opportunity to ask the robot's opinions before submitting their final decision.

Snippets from the Lost on the Moon Task.

Survey & Task Creation

The experiment required the creation of 3 separate surveys, all using the Qualtrics survey creation and analysis software:

A "main" survey to measure participant responses and perceptions upon completing the task.
A survey to help measure the "neutrality" of statements that the robot would be responding with when agreeing or disagreeing with a participant's decision.
The task itself, "Lost On the Moon", which would be a Qualtrics survey masked as a game.

The main survey was designed by Dr. Marlena Fraune, who utilized previous research in the area of robotics and group participation to use reliable metrics of measuring the attitudes and perceptions of participants towards the robots involved in the study. Constructing the survey served as a valuable exercise in gaining familiarity with the Qualtrics software suite, as it required a large variety of question and response styles with which to construct the survey document. This would ultimately include standard survey questions, including demographics as well as a debriefing, that each participant would need to fill out at the conclusion of their exercise session.

The "neutrality" survey was formulated during the construction of the task itself. While participants may believe they are getting the robot's opinion on their decision, in actuality, the robot's responses are pre-written. We constructed a small survey to gather data to ensure that the robot's responses to decisions are not overly partial towards one item or another, analyzing responses and adjusting the responses accordingly.

Lastly, the Lost on the Moon task would be the "game" the participants would be engaging with during test sessions. While the foundation of the task was built as another Qualtrics survey, it required a lot of custom JavaScript and CSS code to help dress it up as a game that would be played with the robot. This included enabling and disabling key elements to ensure that participants would not accidentally follow steps out of order, as well as simulating an exchange with the robot, highlighting it's choice of the item and displaying text responses as to it's reasoning for that choice.

Testing Sessions

All researchers involved in the study would run in-person test sessions in the HRI Lab, recruiting students from all across the New Mexico State University campus. Sessions were run both in groups and individually, following a script written by fellow researcher Krista Apodaca.

Each session would involve ensuring that participants would read and sign an informed consent document. Afterwards, they would be explained the activity they would be participating in, shortly before being introduced to the robot. We would alternate between giving participants the choice to work with the robot or simply introduce the robot as "their teammate". If any participants said no when given the decision, we would respond to them as if we had made a mistake, and that they would have to work with the robot anyway for the task. We would then record the resulting post-task survey and search for trends between these factors.

A group of participants performing the "Lost on the Moon" task.

Results

Upon gathering enough responses from participants, the research team analyzed the results tracked by Qualtrics and formulated a hypothesis. Due to the small sample size we gathered during the semester, we had to strip out many data points that were considered statistically insignificant. We centered our hypothesis on group interactions with the robot vs individual, highlighting the fact that individual participants seemed to experience more positive emotions toward the robot than groups. This hypothesis will be presented along with supporting data on a poster in NMSU's Research & Creativity event (forthcoming).

Screenshots of survey results tracked by Qualtrics.

Page updated

Google Sites

Report abuse