Virtual humans as co-workers: A novel methodology to study peer effects
Introduction
Peer effects are observed in a plethora of domains, such as financial decision making (Bursztyn et al., 2014), academic success of students (Zimmerman, 2003), or in health-related choices (Trogdon et al., 2008). One important domain where peer effects can occur is the workplace. Workers can influence each other in their work performance when they receive information about or directly observe the respective peer’s performance. How strong is the influence of a “rotten apple” on her peers’ performances? Does a top performer induce fellow workers to high effort levels? Understanding the size and nature of such peer effects allows for a more effective organization of labor by answering important questions, such as whether shifts should be composed of workers with homogeneous or heterogeneous productivities/potentials.
We analyze peer effects in an organizational framework where workers work, side by side, on identical tasks but do not interact in any way, except that one of the (two) workers can observe the other worker’s activity. In particular, workers neither complement nor substitute each other in the production process, and receive a fixed and performance independent wage. Spillovers that may occur in such situations are referred to as “pure” peer effects (Charness, Kuhn, 2011, Falk, Ichino, 2006).
In this study, we introduce a new experimental methodology for testing (pure) peer effects. We create a naturalistic work setting in an immersive virtual environment (abbrev. as IVE), where human subjects work together with peers who are computer-generated virtual humans (abbrev. as VH). Subjects perform a real-effort sorting task with their dominant hand, while at the same time the experimenter has control over the actions of the VH via an algorithm. This control over the VH eliminates the reflection problem Manski (1993): while the VH may influence the human subject, the virtual peer cannot be influenced by the human subject.1
Specifically, we address the question whether and to what extent a subject’s work performance is affected by a peer whose performance can be observed, but where the subject’s performance is unobserved by the peer. To control for heterogeneity in the performance of the specific task at hand, we design an experiment consisting of two phases. In the first phase (the initial phase), subjects perform the sorting task alone. In the second phase (the peer phase), they perform the task in the presence of a peer, working independently from each other, and a worker’s payoff is fixed. In two treatments, we model the subject’s virtual peer either as a low productive or a highly productive worker. We refer to these treatments as the SLOW and the FAST treatment, respectively.
Based on a post-experimental survey, we find that most subjects perceive the virtual peer as human-like and the IVE as naturalistic. Subjects evaluate the VH they observe in the FAST treatment indeed as the more productive one – compared to the VH they watch in the SLOW treatment. This means subjects perceive the difference in the performances of both virtual peers exactly as intended. Furthermore, a significant share of subjects, to which we refer as the “competitive” subjects, reports that they wanted to be more successful than the virtual peer, which we interpret as further evidence that subjects acknowledge and react to the presence of VHs.
Overall, we find no statistically significant peer effect between treatments. Interestingly, competitive subjects display significantly higher work performance in the presence of a high productive virtual peer, compared to when they work in the presence of a low productive peer.
Our methodology allows us to collect tracking data in the IVE, in particular, the position and orientation of a subject’s head and hand. These data allow us to define novel performance measures and refine our findings. Using tracking data, we find that competitive subjects display more careful sorting behavior than non-competitive subjects.
We make two important contributions to the literature. First, we introduce a novel experimental methodology to investigate peer effects in a naturalistic setting, but without losing any experimental control. We refine our findings using the tracking possibilities of VR technology. Second, we contribute to the literature on peer effects, by reporting a peer effect in competitive subjects.
Section snippets
Related literature on peer effects
Peer effects have been attributed to peer pressure which may cause a positive productivity effect, due to the workers’ general dislike of providing lower effort than the peer(s).2 Peer pressure, however, may also induce negative productivity effects, in particular, due to discouragement (Georganas et al., 2015). Peer effects have also been attributed to knowledge spillovers causing a positive peer effect (
Hypotheses
We investigate whether and how a worker’s performance is influenced by the physical presence of another worker, referred to as co-worker/peer, whom the worker can observe performing an identical but otherwise unrelated task. In particular, we ask whether the observable productivity of a peer affects a worker’s own work performance in a setup without any interdependency between co-worker and worker.
History and methodology of VR experiments
In this section, we provide a brief history of experiments conducted in IVEs and some introductory methodological remarks. Following some other scholars, we also refer to them as VR experiments. Behavioral research using VR experiments started in the 1990s in domains such as visual perception, spatial cognition, psychotherapy, education, and learning. For a comprehensive overview of early VR experiments in the 2000s and before, we refer to the book by Blascovich and Bailenson (2011).
The virtual environment
We conduct our experiment in the surround-projection room aixCAVE at the RWTH Aachen University (see Fig. 2). The aixCAVE provides a five-sided IVE with a size of 5.25m × 5.25m × 3.30m (width × length × height). The five projection displays, i.e., the four walls and the floor, enclose the user giving her a 360 ° field of regard in the horizontal plane in the computer-generated environment (Kuhlen, 2014). In our setting, when a human subject enters the aixCAVE, she perceives the virtual
Results: Virtual humans and peer effects
We organize our results in two sections. In Section 6, we analyze performance data similar to those collected in related experimental studies. In Section 7, we perform additional analyses based on performance measures that use tracking data that is easily generated in IVEs.
Performance analysis using tracking data
One of the advantages of conducting experiments in IVEs is that one can collect tracking data. Such data can help to identify determinants of superior performance and might measure the exerted effort more directly than a potentially noisy output measure. In our experiment, we collected data about the subject’s hand position, and about the position and the orientation of the subject’s head. In addition, we obtained data about the exact position of each cube when grabbed, and whether and how it
Conclusion and discussion
We present a novel methodology to study peer effects in the workplace. By incorporating a virtual human (VH) in an immersive virtual environment (IVE) as a peer, we avoid the reflection problem: while the VH is immune to be influenced by the human subject, it may affect the human subject. We find that human subjects rate the IVE and the VH as natural.
We find that competitive subjects perform well in our setting, in particular in the FAST treatment. Whether and to what extent this is due to
Acknowledgments
We thank Jonathan Wendt for his assistance in conducting the experiment, and Lucas Braun for preparing the raw data for the analysis. We also thank Almut Balleer, Robert Böhm, Bernd Irlenbusch, the participants of the seminars in Bamberg, Erfurt, Göttingen, Karlsruhe; the conference participants at the 2015 Economic Science Association Meetings in Sydney, Heidelberg, and Dallas; and the participants of the 15th TIBER Symposium in Tilburg, for their valuable comments. We acknowledge the
References (56)
- et al.
Peer effects in the workplace
Am. Econ. Rev.
(2017) - et al.
Clean evidence on peer effects
J. Labor Econ.
(2006) Z-tree: zurich toolbox for ready-made economic experiments
Exp. Econ.
(2007)- et al.
Because of You i Did not Give up – How Peers Affect Perseverance
(2017) - et al.
Teaching students and teaching each other: the importance of peer learning for teachers
Am. Econ. J.
(2009) Exploring Virtuality: Virtualität im Interdisziplinaren Diskurs
(2014)- et al.
Peers at work
Am. Econ. Rev.
(2009) - et al.
Peers at work: evidence from the lab
PLoS One
(2018) Kinect For Windows
(2011)3ds Max
(2014)
Superstar extinction
Q. J. Econ.
Interpersonal distance in immersive virtual environments
Personal. Soc. Psychol. Bull.
The scientific research potential of virtual worlds
Science
Social incentives in the workplace
Rev. Econ. Stud.
Virtual reality as a communication system
Commun. Age Virtual Real.
Infinite reality: avatars, eternal life, new worlds, and the dawn of the virtual revolution
TARGET ARTICLE: immersive virtual environment technology as a methodological tool for social psychology
Psychol. Inq.
Studying social interactions through immersive virtual environment technology: virtues, pitfalls, and future challenges
Front. Psychol.
Social vr: howpersonal space is affected by virtual agents’ emotions.
Virtual Reality (VR), 2018 IEEE
Peers at work: economic real-effort experiments in the presence of virtual co-workers
Virtual Reality (VR), 2017 IEEE
The uncanny valley: does it exist
Proceedings of Conference of Human Computer Interaction, Workshop on Human Animated Character Interaction
Understanding mechanisms underlying peer effects: evidence from a field experiment on financial decisions
Econometrica
Lab labor: what can labor economists learn from the lab?
Handbook of Labor Economics
Surround-screen projection-based virtual reality: the design and implementation of the CAVE
SIGGRAPH ’93: Proceedings of the 20th annual conference on Computer graphics and interactive techniques
How immersive is enough? a meta-analysis of the effect of immersive technology on user presence
Media Psychol.
Virtual prejudice
J. Exp. Soc. Psychol.
Virtual experiments and environmental policy - frontiers of environmental and resource economics
J. Environ. Econ. Manag.
Cited by (15)
Exposure to immersive virtual environments decreases present bias
2024, Journal of Behavioral and Experimental EconomicsMe, myself, and I: Image concerns and honesty in immersive VR
2023, Computers in Human BehaviorLearning with simulated virtual classmates: Effects of social-related configurations on students’ visual attention and learning experiences in an immersive virtual reality classroom
2022, Computers in Human BehaviorCitation Excerpt :The authors manipulated virtual classmates' attention-related behavior (i.e., peer learners being attentive or distracting during instruction; see Experiment 4) and found positive effects of more attentive virtual classmates on students' performance after the IVR lesson. Similarly, a study in the field of economics manipulated virtual co-workers’ productivity and found a positive relation with the performance of participants doing the same task as the virtual co-workers in an IVR environment (Bönsch et al., 2017; Gürerk, Bönsch, Kittsteiner, & Staffeldt, 2019). In sum, based on existing IVR studies, it can be assumed that classmates' performance affects students' learning in IVR settings; however, it is still unclear how exactly peers' (performance-related) behavior in an IVR classroom needs to be configured in order to be recognized by K-12 students in an IVR classroom scenario.
Virtual reality and sustainable behavior in business
2021, Cleaner and Responsible ConsumptionCitation Excerpt :The management environment is experienced from behind a computer screen and interaction with the virtual environment is based on an assumingly functional role in the decision-making process, for instance as a virtual tourist navigating through a location or as a consumer making buying decisions in a fixed setting. Management and busines studies using high-immersive VR are the exception focusing mainly on behavioral aspects in business (Gürerk et al., 2019). While VR as a research tool to represent an existing, or near-future, reality has been accepted research, VR experiments in management research to explore what does not yet exist in reality or cannot be researched in reality (black boxes), are still in their infancy.
Competing on the Holodeck - The effect of virtual peers and heterogeneity in dynamic tournaments
2021, Journal of Behavioral and Experimental EconomicsCitation Excerpt :They did not find any significant difference between laboratory experiments and field studies when it comes to that spillover effect. Gürerk et al. (2019) conduct an experiment based on a virtual setting similar to the one used in this paper. They embed a virtual human as a co-worker of a human subject into an immersive virtual environment, and observe that low-productivity human subjects increase their work performance more when they observe a low-productivity avatar than when they observe a high-productivity avatar.
Applications and research trends of digital human models in the manufacturing industry
2019, Virtual Reality and Intelligent Hardware