The benefit of being physically present: A survey of experimental works comparing copresent robots, telepresent robots and virtual agents$
Jamy Li n
Department of Communication, Stanford University, 450 Serra Mall, Building 120, Stanford, CA 94305, USA a r t i c l e i n f o
Received 10 April 2014
Received in revised form 3 January 2015
Accepted 7 January 2015
Communicated by C. Bartneck
Available online 14 January 2015
Virtual a b s t r a c t
The effects of physical embodiment and physical presence were explored through a survey of 33 experimental works comparing how people interacted with physical robots and virtual agents. A qualitative assessment of the direction of quantitative effects demonstrated that robots were more persuasive and perceived more positively when physically present in a user's environment than when digitally-displayed on a screen either as a video feed of the same robot or as a virtual character analog; robots also led to better user performance when they were collocated as opposed to shown via video on a screen. However, participants did not respond differently to physical robots and virtual agents when both were displayed digitally on a screen – suggesting that physical presence, rather than physical embodiment, characterizes people's responses to social robots. Implications for understanding psychological response to physical and virtual agents and for methodological design are discussed. & 2015 Elsevier Ltd. All rights reserved. 1. Introduction
Intelligent robotic agents are being designed for everyday life (Goodrich and Schultz, 2007) while virtual agents are already prevalent in online and gaming domains (Blascovich and Bailenson, 2011).
Physical robots and virtual agents are both defined and distinguished by the nature of their embodiment (Dautenhahn, 1999): a robot's body is made of metal and plastic parts while a virtual agent's body is digitally generated using computer algorithms. This demarcation in embodiment lends robots and virtual agents unique qualities. While a robot in the real world is uniquely able to touch and be touched by people, a virtual agent in a simulated world can perform activities not possible in real life. For example, a healthcare robot can pick up and deliver a patient's medication, gently tap on a patient's shoulder, and respond to a patient's touch. Conversely, a virtual healthcare assistant can alter its gender, age, height and other visual characteristics on-thefly in order to suit a patient's preferences.
Today the choice of whether to implement a robotic or virtual agent is largely governed by the requirements of the tasks to be performed. A robot may be used in situations where physical objects such as beverages or hospital equipment need to be delivered, while a virtual agent may be employed on a website to aid in teaching an online learning course. As robotic technologies are developed that focus on general social interactionwith people, however, the choice of what type of agent to use becomes less clear. Consider, for example, a virtual agent displayed on a computer screen that acts as a receptionist and a similar-looking robot designed for the same role: is there a difference in their ability to engage patrons? Given the enormous flexibility and prevalence of digitally-rendered agents that can dynamically change their appearance, are readily transportable across distances and live on relatively inexpensive digital screens, such questions are of particular importance to roboticists interested in exploring the consequences of introducing socially-competent robots as well as to researchers interested in interactions with physical and virtual media.
Previous experimental work that has compared social robots with virtual agents has claimed that the physicality of the robot is beneficial to user interaction (e.g., Wainer et al., 2006; Kiesler et al., 2008). However, many such studies conflate two different dimensions of physicality: the physical embodiment of the robot, and the fact that it is physically present in front of the user. Is a collocated robot advantageous because it has a physical body or because it is physically collocated in a person's space? Is the recognition that an agent has a physical body sufficient or must the physical body be present in front of a user? If both the embodiment and presence of an agent influences people equally, it may mean that interaction design of social agents is more complex than previously thought.
Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/ijhcs
Int. J. Human-Computer Studies http://dx.doi.org/10.1016/j.ijhcs.2015.01.001 1071-5819/& 2015 Elsevier Ltd. All rights reserved. ☆This paper has been recommended for acceptance by C. Bartneck. n Tel.: þ1 323 474 1084.
E-mail address: firstname.lastname@example.org
Int. J. Human-Computer Studies 77 (2015) 23–37
Conversely, if one factor has a greater impact on user response than the other, development of social agents may be simplified. While these questions are not new (e.g., Kidd and Breazeal, 2004; Wainer et al., 2006; Kiesler et al., 2008 and other authors have addressed them to varying degrees), a systematic survey provides an original perspective on these questions by capturing the corpus of past work, introducing and applying a common framework to its analysis and identifying themes across studies.
The current paper investigates whether people respond differently to physical vs. virtual agents by presenting a survey of past experimental work. Experimental stimuli are classified into three categories to independently investigate the effect of physical embodiment and presence on human response (see Fig. 1): “copresent robots” that are physically embodied as well as physically present in a user's space (i.e., a physical robot that is in front of a person); “telepresent robots” that are physically embodied but displayed on a screen (i.e., a live or recorded video feed of a physical robot); and “virtual agents” that may have similar appearance and behavior to a robot but are digitally embodied (i.e., a computer graphics model of a robot). Its main contributions are an explication of embodiment and presence from a theoretical standpoint to support the development of structured research questions (Section 2); a comprehensive survey identifying 33 key works comparing physical and virtual agents (Section 3); a searchable table of past work to aid the research community (Section 3); the finding that physical presence, not embodiment, governs psychological response to agents across a variety of dependent measures (Section 4); and a discussion of implications for experimentation and suggestions for future investigation (Section 5). 2. Interactions with physical and virtual agents