Paper Reading #3: Ripple Effects of an Embedded Social Agent: A Field Study of a Social Robot in the Workplace

Intro:

Title: Ripple Effects of an Embedded Social Agent: A Field Study of a Social Robot in the Workplace

Reference Information: Work from the Human-Computer Interaction Institute and the Robotics Institute at Carnegie Mellon

Author bios:

Min Kyung Lee is a PhD candidate at Carnegie Mellon with the Human-Computer Interaction Institute; she received her master's from Carnegie Mellon and her bachelor's at Korea Advanced Institute of Science and Technology.

Sarah Keisler is a Hillman Professor of Computer Science and Human-Computer Interaction at Carnegie Mellon. She is working on scientific work in teams, collaborative work, and human-robot interaction. She is a recipient of the 2009 CHI Lifetime Achievement Award.

Jodi Forlizzi is a professor in Design and Human Computer Interaction. She works with interaction design like peripheral displays and robots.

Paul Rybski is a Senior Systems Scientist in the Robotics Institute at Carnegie Mellon. He has worked on a large number of robotics projects.

Snackbot

Summary: For five months in 2011, a snack delivery robot worked in a university office; users could order snacks for delivery on Mondays, Wednesdays, and Fridays, and the robot would deliver them in the afternoon. The robot was programmed to exchange small talk with the users as it delivered the snacks, and for half of the participants, exchanges were personalized to include previous snack orders, the last time the robot saw the participant, and such. The robot was also programmed to acknowledge and apologize for breakdowns or mistakes after the fact and handle such instances with conversational statements in the moment (i.e. "I can't see you. Can you please stand in front of me?") The majority of interactions were audio recorded or video recorded for analysis. According to the paper, the longevity of the study and the robot's work was the most novel contribution of the study.

Related Works:

An approach for a social robot to understand human relationships: Friendship estimation through interaction with robots - Kanada, Ishiguro

A robot intended to identify friendships was studied for two weeks in a Japanese school; it developed friendly relationships with students and was able to identify some friendships through observation. Also a study focusing on a social robot.

Anthropromorphism and the social robot - Duffy

A study of social robots with varying degrees of anthropomorphism and its effects on their interactions with people. Snackbot was somewhat anthropomorphic.

 End-user programming of a social robot using dialog - Gorostiza, Salichs

A social robot can be programmed by the end-user using dialogs; users can combine facial expressions and these combinations can be stored for use later. Investigates interaction with social robots.

A survey of socially interactive robots - Fong, Nourbakhsh, Dautenhan

Provides an overview of various implementations and forms of social robots.

Coordination of verbal and non-verbal actions in human–robot interaction at museums and exhibitions - Yamazaki, Yamazaki, Burdelski, Kuno, Fukushima

A study of human-robot verbal and non-verbal interaction at museums where robots could serve as guides or explain features of the exhibits. Also discusses how people react to and interact with a robot in a service role.

"It doesn’t matter what you are!" Explaining social effects of agents and avatars - von der Pütten, Kramer, Gratch, Kang

A study evaluating social interaction with avatars and virtual people with different levels of interactivity on the virtual side. Shows that humans can interact socially with artificial agents.

Relating conversational expressiveness to social presence and acceptance of an assistive social robot - Heenrick, Krose, Evers, Weilinga
A social robot with various levels of sociability was introduced in an elder-care institution. People were more expressive and interactive with a more sociable robot, but accepted it whether it was social or not. Very similar to the workplace study.

User Localization During Human-Robot Interaction - Alonso Martin, Gorostiza, Malfaz, Salichs
A social robot uses user localization using visual and sound information to facilitate more natural communication. Similar to Snackbot's interactions.

Integration of a Voice Recognition System in a Social Robot - Alonso-Martin, Salichs
A social robot is given a speech recognition system; very important in social communication between robots and humans.

Exploratory studies on social spaces between humans and a mechanical-looking robot - Walters, Dautenhahn, Woods, Koay, Te Boekhorst, Lee
A study of participants' preferences about how close to be to a mechanical-looking robot; some preferred a close space like they would with friends, while others got closer, seeing it not as a social entity but only a robot. Useful for determining interaction space with social robots.

Based on the related works found, Snackbot itself is not a novel concept, but the long-term study of a robot in the workplace is.

Evaluation: The author conducted a qualitative, subjective evaluation of the robot by interviewing participants about their opinions, the robot's mistakes, and whether they enjoyed the robot. This seems to be the only appropriate and accurate way to gain complete insight into how this social experiment worked. Overall, participants reacted positively, happy to have constant, pleasant visits from a robot; those with personalized interaction felt more emotionally connected to and attached to the robot. On the other hand, five out of the 21 participants found it slightly awkward, knowing that the robot would not actually miss them when it said "I'll miss you!" or feeling uncomfortable chatting with a service robot. Breakdowns "broke" the illusion of humanity for many users. Eavesdropping was also observed, and observation of other users affected behavior towards the robot. Interestingly, the users came to view the robot as a social entity, feeling jealous when it interacted with other users and using the robot's visit as common ground and an excuse to socialize.

Discussion: Based on the conversation examples provided, I think I would classify the robot as "cute" and I would have enjoyed interacting with it in my work environment. I think workplace robots could prove useful and helpful in a number of situations, especially when people are not available to fill certain roles. I find it fascinating that most of the participants considered Snackbot to be a social entity, almost human-like, and I think that knowledge will prove extremely useful if robots begin to fill more service roles. I think the evaluation was performed perfectly; there is no other way to accurately learn about the users' opinions and evaluate whether the project was a success. Additionally, this robot's presence affected workplace dynamics even when it wasn't present. Because of the common ground it provided, many people socialized more, and they even became protective of the robot as one would become protective of a friend.