Transformed social interaction
More actions
Transformed social interaction (TSI) is a research paradigm and theoretical framework, set out in a 2004 paper by Jeremy N. Bailenson, Andrew C. Beall, Jack Loomis, Jim Blascovich and Matthew Turk, for studying how collaborative virtual environments (CVEs) can filter and modify the appearance and nonverbal behavior of the people who use them. Because a CVE tracks each user and renders their avatar separately for every other participant, the behavior that others see can be decoupled from the behavior the user actually performed. The authors wrote that, as a result, "conceptual and perceptual constraints inherent in face-to-face interaction need not apply", and that decoupling algorithms "can enhance or degrade facets of nonverbal behavior".[1]
The framework sorts possible transformations into three dimensions, called self representations, sensory capabilities and contextual situation in the 2004 paper and self representation, sensory abilities and situational context in a 2005 follow-up: how an avatar looks and behaves, extra information given to a participant, and the spatial and temporal arrangement of the conversation.[1][2] Experiments built on it include "augmented gaze", in which a presenter's avatar appears to look at several listeners at once, and "digital chameleons", in which an agent copies a listener's head movements after a delay. In both cases participants rarely or never detected the manipulation, yet it made the speaker more persuasive (for female listeners only, in the gaze study).[2][3] The paradigm has since been applied to social VR platforms, virtual classrooms and asynchronous VR meetings, and related work by other groups has explored artificially "augmenting" social behavior in multi-user VR.[4]
Definition
In the 2004 paper that set out the paradigm, published in Presence: Teleoperators and Virtual Environments, TSI "involves novel techniques that permit changing the nature of social interaction (either positively or negatively) by providing system designers with methods to enhance or degrade interpersonal communication". A videoconference sends a camera image from one person to another, but a CVE sends tracking data for eye gaze, facial gestures and body gestures, and each receiving machine renders that data onto an avatar. The rendering step is where transmitted and received signals can be separated: gaze directed from person A to person B "can be transformed without A's knowledge, such that B experiences the opposite, gaze aversion". TSI can be applied to some, all or none of the members of a CVE.[1]
The 2005 follow-up in Human Communication Research summarized the idea as follows: the appearance and behavior of avatars "can be systematically filtered in immersive CVEs idiosyncratically for other interactants, amplifying or suppressing features and nonverbal signals in real-time for strategic purposes", and these transformations should in theory affect a speaker's persuasive and instructional abilities.[2] A 2024 review by Eugy Han and Bailenson describes the same framework as positing that "the physical appearance and behavioral actions of avatars can be edited idiosyncratically for all interactants".[4]
The 2004 paper stated that the specific CVE implementation was not critical. TSI should work in projection-based CVEs, head-mounted display CVEs, CAVEs and certain types of augmented reality CVEs.[1] The authors also noted that the idea of decoupling rendered from actual behavior was not new, citing earlier work by Steve Benford and colleagues on user embodiment in CVEs (1995) and by Loomis, Blascovich and Beall (1999).[1]
The three dimensions
| Dimension | What is transformed | Examples given by Bailenson et al. |
|---|---|---|
| Self-representation | The rendered appearance or behavior of a user's avatar, which can differ for each viewer | Morphing a speaker's face toward each listener's face; mimicking listeners' gestures; non-zero-sum mutual gaze; human "cyranoid" assistants who drive a leader's nonverbal behavior |
| Sensory abilities | Extra information rendered only to a particular participant | Running gaze meters for a teacher; tallies of nods or signs of confusion; filters that hide distracting behavior; view frustums showing where others look; name labels; hidden "virtual ghost" consultants |
| Situational context | The spatial or temporal structure of the shared scene | Different seating arrangements for each participant; taking another participant's viewpoint; rewinding or speeding up a recorded meeting; letting a temporally absent member join later |
Sources for the table: Bailenson et al. (2004) and (2005).[1][2]
Transforming self-representation
The 2004 paper drew on the social psychology finding that similarity breeds attraction and on the authors' earlier work showing that people treat avatars that look like themselves more intimately. It proposed that a speaker could blend part of each listener's facial structure and texture into his or her own avatar, and that a CVE could show a different hybrid face to each listener at the same time, since "the leader does not need a consistent representation across interactants".[1] Behavior could be borrowed the same way. Citing Tanya Chartrand and John Bargh's 1999 "chameleon effect", in which people who are subtly mimicked report liking the mimicker more, the paper suggested that algorithms could blend a speaker's animations with the detected movements of listeners.[1]
The best-known example is what the paper called non-zero-sum mutual gaze (NZSMG). In face-to-face conversation mutual gaze is zero-sum: if A looks at B for 70 percent of the time, A cannot look at C for more than 30 percent. With avatars, A "can be made to appear to maintain mutual gaze with both B and C for a majority of the conversation".[1] The same decoupling can work in reverse, letting a person watch someone closely while their avatar appears to look elsewhere. To keep such an avatar behaving believably, the authors considered an artificial intelligence algorithm, which they judged "may still be many years off", and a more likely alternative: a team of human assistants, each supplying the nonverbal behavior shown to one listener, a many-to-many "Wizard of Oz" implementation they related to Stanley Milgram's "cyranoids".[1]
Transforming sensory abilities
Because a CVE must record every participant's movements to work at all, it can summarize those movements for selected users. The 2004 paper described a gaze meter that keeps a running total of how long an instructor has looked at each student and warns her when her attention is uneven. It also described automatic tallies of nonverbal cues, such as the share of students showing signs of confusion.[1] Filters could suppress a distracting habit such as pen tapping on the sending side, or let a listener stop rendering a speaker's hand movements on the receiving side. Other tools rendered only to one person included color-coded wire-frame view frustums showing each participant's field of view, floating name labels over other participants' heads (used in the authors' own gaze studies), and human consultants visible only to the user they advise.[1]
Transforming the situation
The third dimension alters the spatial or temporal arrangement of the shared scene. In a three-person meeting, one participant might see the group as an isosceles triangle while the other two see an equilateral one, with the system adjusting head and eye movements so that intended gaze directions stay intact. The paper said this is straightforward to design for up to four interactants.[1] Participants could also adopt "multilateral" perspectives, seeing themselves through another participant's eyes during a live conversation. Recorded interactions could be rewound and played back faster to catch up. A member who could not attend could place an autopilot avatar in her seat during the live meeting, then later enter the recording and receive the nonverbal behavior directed at her.[1] The 2005 paper gave a classroom version: in a CVE, every one of 20 students can sit directly in front of the virtual blackboard and perceive the others as sitting behind or beside them.[2]
History
The paradigm grew out of collaborative work between Bailenson, then at Stanford University's Department of Communication and its Virtual Human Interaction Lab, and Beall, Loomis and Blascovich of the psychology department and Turk of the computer science department at the University of California, Santa Barbara. The 2004 paper acknowledges support from National Science Foundation awards SBE-9873432 and ITR IIS-0205740.[1] It cites an earlier exploratory study of non-zero-sum mutual gaze by Beall, Bailenson, Loomis, Blascovich and Christopher Rex, presented at HCI International 2003, which found that interactants were not aware of the manipulated gaze and responded to it as if it were real.[1]
Turk, Bailenson, Beall, Blascovich and Rosanna Guadagno also presented the paradigm to the multimodal-interfaces community at the 6th International Conference on Multimodal Interfaces (ICMI) in October 2004. That paper described TSI as rendering "a user's visual representation ... in a way that strategically filters selected communication behaviors in order to change the nature of a social interaction", and noted that it requires technology to detect, recognize and manipulate facial expressions, gestures and eye gaze.[5]
Bailenson and Beall expanded the framework in "Transformed Social Interaction: Exploring the Digital Plasticity of Avatars", the opening chapter of Avatars at Work and Play (Springer, 2006). The chapter set three goals for the research program: to implement the strategies in CVEs, to measure which TSI tools people actually use, and to examine TSI's effect on the effectiveness of interaction and on the specific goals of particular interactants.[6]
Research
Detecting transformations
The 2004 paper reported a pilot "nonverbal Turing test". Participants wearing head-mounted displays sat with two avatars: one driven in real time by another person, the other replaying the participant's own head movements with a delay of 1, 2, 4 or 8 seconds. Although the 41 undergraduates were told that one avatar was mimicking them, they averaged only 66 percent correct (chance was 50 percent), more than a quarter were not reliably better than chance, and accuracy dropped as the delay grew.[1] A later two-experiment study in Computers in Human Behavior found that participants were significantly worse than chance at identifying the other human when the computer agent mimicked them. Mimicry was easier to spot when an agent reproduced a participant's head movements exactly on three axes (pitch, yaw and roll) than when it reversed them on the same axis or mapped them onto another rotational axis.[7]
Key studies
| Year | Study | Transformation | Reported result |
|---|---|---|---|
| 2004 | Bailenson, Beall, Loomis, Blascovich and Turk, Presence[1] | Self-mimicking avatar in a nonverbal Turing test | 66 percent average accuracy at picking the human; detection harder at longer delays |
| 2005 | Bailenson et al., Human Communication Research[2] | Augmented gaze (presenter's head turned toward each listener 100 percent of the time) | Women agreed more with the message under augmented gaze; no participant explicitly detected the manipulation |
| 2005 | Bailenson and Yee, Psychological Science[3] | Agent mimicking a listener's head movements at a 4-second delay | Mimicking agents were more persuasive and rated more positively, though participants did not detect the mimicry |
| 2006 | Bailenson and Yee, Presence[8] | Nonverbal mimicry and facial similarity over 15 sessions | Making avatars more facially similar sometimes improved task performance |
| 2008 | Bailenson, Yee, Blascovich, Beall, Lundblad and Jin, Journal of the Learning Sciences[9] | Augmented teacher perception, transformed seating, virtual co-learners | Teachers spread gaze more evenly; students learned more in the center of the teacher's view and closer to the teacher |
| 2016 | Oh, Bailenson, Krämer and Li, PLOS ONE[10] | Slightly enhanced avatar smile during live conversation | More positive affect and social presence; over 90 percent did not detect the change |
| 2018 | Roth et al., IEEE VR[11] | Visual augmentations for eye contact, joint attention and grouping | Significantly higher social presence; behavior suggested more eye contact |
| 2023 | Han et al., Journal of Computer-Mediated Communication[12] | Self-resembling versus uniform avatars, 192 environments, 8 weeks | Self-avatars raised nonverbal synchrony; larger visible space raised synchrony and presence |
Gaze and mimicry
In the 2005 augmented gaze experiment, 72 participants heard a confederate presenter read a persuasive passage in an immersive CVE under natural, augmented or reduced gaze. In the augmented and reduced conditions the presenter's real head movements were scaled down by a factor of 20 and recentered so that the avatar looked into each listener's eyes (augmented) or down at the presenter's computer screen (reduced); the slight remaining movement kept the avatar from appearing frozen. Women agreed with the message more under augmented gaze than under the other conditions, men recalled more of the passage than women, and coders found zero cases in which a participant explicitly noticed that the gaze had been altered.[2] The HMD used in the study had a 50 by 38 degree field of view per eye with full binocular overlap.[2]
In the 2005 "digital chameleons" study, an embodied agent in immersive VR delivered a persuasive argument while either mimicking the listener's head movements at a 4-second delay or replaying another participant's recorded movements. The authors described the result as the first demonstration of social influence from "a nonhuman, nonverbal mimicker".[3] A later summary of the study reported that participants detected their own gestures in only about 5 percent of instances.[7] Daniel Roth and colleagues later built a "Mimicry Injector" that added artificial mimicry to live two-person VR negotiations. Most participants did not detect the change, but it did not significantly affect how they perceived the conversation.[13]
Appearance
Early facial-similarity studies used two-dimensional images. In a 2006 Political Psychology study, half of the subjects saw a candidate's photograph morphed to a blend of 60 percent of the unfamiliar candidate and 40 percent of the subject's own face; male subjects rated the morphed candidate more favorably than the unaltered photograph, while female subjects rated him more negatively.[14] A 2008 Public Opinion Quarterly paper reported three experiments with national samples of voters: facial similarity influenced preferences mainly for unfamiliar candidates, and participants showed no conscious awareness of the manipulation.[15] Bailenson and Beall's 2006 chapter also described a "team face", a morph containing 25 percent of each of four team members. Only three of 32 participants noticed their own face in it, and people scrutinized arguments more closely when their own team face delivered them. The authors noted that these studies used two-dimensional avatars on non-immersive displays.[6]
In a longitudinal study, three groups of three participants met in a CVE for 15 sessions over ten weeks, about 45 minutes per session; transforming avatars to increase facial similarity sometimes improved task performance.[8] In 2016, Soo Youn Oh and colleagues tracked participants' facial expressions in real-time dyadic conversations and rendered each avatar's smile either veridically or slightly enhanced; the enhanced-smile pairs used more positive words, reported more positive affect and felt stronger social presence.[10]
The paradigm also shaped the Proteus effect experiments. Nick Yee and Bailenson used TSI's decoupling so that a participant saw his or her own avatar as more or less attractive, or as taller or shorter, while the confederate, blind to condition, saw an untextured face or an avatar of the confederate's own height. This separated the Proteus effect, a change in the user's own behavior, from behavioral confirmation by others.[16]
Social VR and later work
Han and colleagues used the TSI paradigm in a study of 81 participants who met in groups over eight weeks in the social VR platform ENGAGE, wearing either self-resembling or uniform avatars. A second study of 137 participants cycled through 192 virtual environments. Self-avatars raised nonverbal synchrony and self-presence but lowered enjoyment compared with uniform avatars. Environments with more visible space raised synchrony, presence and enjoyment.[12] Han and Bailenson's 2024 review also describes a 2023 study by Meng Ting Shih and colleagues in Horizon Workrooms: a confederate whose avatar looked moderately similar to the participant was more persuasive than one that looked identical or dissimilar.[4]
Portia Wang, Mark Miller and Bailenson's 2023 "belated guest" work returned to the situational dimension. It transforms recorded proxemics and head orientations of avatars so that a person who missed a VR meeting can enter the recording later and be nonverbally included in the scene.[17]
Work outside Stanford has developed related "social augmentation" techniques. At IEEE VR 2018, Roth, Constantin Kleinbeck, Tobias Feigl, Christopher Mutschler and Marc Erich Latoschik presented visual transformations for eye contact, joint attention and grouping. In a between-subjects study of 125 participants in groups of five, exploring a virtual museum in a large multi-user tracking space, the augmentations significantly increased social presence, and the authors reported that they also seemed to produce more eye contact and more focus on avatars and objects in the scene.[11] A companion ISMAR 2018 paper compared natural, hybrid, synthesized and random gaze transmission between avatars. Ratings of virtual rapport, trust and interpersonal attraction showed a linear trend, increasing with the naturalness and social adequacy of the transmitted gaze.[18] In 2019 the same group reviewed augmentation approaches for VR, MR and AR, proposed a software architecture based on four data layers, and presented a prototype, injectX, that tracks body motion, eye gaze and lower-face expressions, then analyzes, augments and blends behaviors in immersive interactions. Their stated motivations included synthesizing behavior when sensor input is missing and supporting inclusion and training for people with social communication disorders.[19]
Applications in VR and AR
Several TSI experiments have used classroom settings. In the 2008 Journal of the Learning Sciences paper, 40 participants taught a room of nine virtual students. Those given "augmented social perception", in which students faded toward translucency the longer they stayed outside the teacher's field of view, spread their gaze more evenly. Further experiments placed each learner in the center of the teacher's view or closer to the teacher, which is possible for every student at once in a CVE, and found better learning there. A fourth experiment showed that learners conformed to inserted virtual co-learners who either paid attention or were distracting.[9]
Meetings and remote collaboration are another recurring application. The 2004 paper proposed gaze-preserving seating arrangements, behavior summaries and asynchronous playback for meetings, and the 2023 belated-guest work developed a method for the asynchronous case.[1][17] Han and Bailenson note that such transformations may be unintentional, "a by-product of imperfect technological systems", or intentional, made by developers and designers to shape how people interact.[4]
Ethical concerns
The originators raised ethical problems from the start. The 2004 paper said the authors did "not advocate the unconstrained use of TSI" but considered decoupling of rendered from actual behavior "inevitable" as CVEs spread. It noted that the people most likely to profit may be those who enter interactions with goals such as changing others' attitudes.[1] It also warned that the threat of TSI "may be the very downfall of CVE interaction": if participants cannot trust that gestures are genuine, that avatars correspond one to one with people, or that others are present in real time, they may have little reason to use CVEs at all. As one remedy, the authors suggested TSI detectors, either algorithms or human observers.[1]
The 2005 paper said augmented gaze "may turn out to be an attractive strategy for advertisers, salespeople, politicians, and others who seek to gain influence".[2] The 2006 chapter called the "Orwellian themes" of the program "quite apparent". It asked whether TSI is fundamentally different from "nose jobs, teeth-whitening, self-help books and white lies", and observed that very few mobile phone users mind or even notice that their voices are already digitally processed. It added that the ethical concerns "largely vanish" if all participants know that everyone can use such transformations.[6]
See also
References
- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 1.10 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.20 Jeremy N. Bailenson, Andrew C. Beall, Jack Loomis, Jim Blascovich, Matthew Turk (2004-08). "Transformed Social Interaction: Decoupling Representation from Behavior and Form in Collaborative Virtual Environments". Presence: Teleoperators and Virtual Environments, vol. 13, no. 4, pp. 428-441. MIT Press (author copy, Stanford Virtual Human Interaction Lab). doi:10.1162/1054746041944803. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/bailenson-tsi.pdf. Retrieved 2026-10-06.
- ↑ 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 Jeremy N. Bailenson, Andrew C. Beall, Jack Loomis, Jim Blascovich, Matthew Turk (2005-10). "Transformed Social Interaction, Augmented Gaze, and Social Influence in Immersive Virtual Environments". Human Communication Research, vol. 31, no. 4, pp. 511-537. International Communication Association (author copy, Stanford Virtual Human Interaction Lab). doi:10.1111/j.1468-2958.2005.tb00881.x. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/bailenson-augmented-gaze.pdf. Retrieved 2026-10-06.
- ↑ 3.0 3.1 3.2 Jeremy N. Bailenson, Nick Yee (2005-10). "Digital chameleons: automatic assimilation of nonverbal gestures in immersive virtual environments". Psychological Science, vol. 16, no. 10, pp. 814-819. PMID 16181445. https://doi.org/10.1111/j.1467-9280.2005.01619.x. Retrieved 2026-10-06.
- ↑ 4.0 4.1 4.2 4.3 Eugy Han, Jeremy N. Bailenson (2024-05-22). "Social Interaction in VR". Oxford Research Encyclopedia of Communication. Oxford University Press. doi:10.1093/acrefore/9780190228613.013.1489. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/han-bailenson-2024_0.pdf. Retrieved 2026-10-06.
- ↑ Matthew Turk, Jeremy Bailenson, Andrew Beall, Jim Blascovich, Rosanna Guadagno (2004-10-13). "Multimodal transformed social interaction". Proceedings of the 6th International Conference on Multimodal Interfaces (ICMI '04). ACM. pp. 46-52. doi:10.1145/1027933.1027942. https://doi.org/10.1145/1027933.1027942.
- ↑ 6.0 6.1 6.2 Jeremy N. Bailenson, Andrew C. Beall (2006). "Transformed Social Interaction: Exploring the Digital Plasticity of Avatars". Avatars at Work and Play: Collaboration and Interaction in Shared Virtual Environments (R. Schroeder and A.-S. Axelsson, eds.), pp. 1-16. Springer (author copy, Stanford Virtual Human Interaction Lab). doi:10.1007/1-4020-3898-4_1. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/bailenson-digital-plasticity.pdf. Retrieved 2026-10-06.
- ↑ 7.0 7.1 Jeremy N. Bailenson, Nick Yee, Kayur Patel, Andrew C. Beall (2008). "Detecting digital chameleons". Computers in Human Behavior, vol. 24, no. 1, pp. 66-87. Elsevier (author copy, Stanford Virtual Human Interaction Lab). doi:10.1016/j.chb.2007.01.015. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/bailenson-digital-chameleons.pdf. Retrieved 2026-10-06.
- ↑ 8.0 8.1 Jeremy N. Bailenson, Nick Yee (2006-12). "A Longitudinal Study of Task Performance, Head Movements, Subjective Report, Simulator Sickness, and Transformed Social Interaction in Collaborative Virtual Environments". Presence: Teleoperators and Virtual Environments, vol. 15, no. 6, pp. 699-716. https://doi.org/10.1162/pres.15.6.699. Retrieved 2026-10-06.
- ↑ 9.0 9.1 Jeremy N. Bailenson, Nick Yee, Jim Blascovich, Andrew C. Beall, Nicole Lundblad, Michael Jin (2008). "The Use of Immersive Virtual Reality in the Learning Sciences: Digital Transformations of Teachers, Students, and Social Context". Journal of the Learning Sciences, vol. 17, no. 1, pp. 102-141. Taylor and Francis (copy hosted by the LIFE Center). doi:10.1080/10508400701793141. http://life-slc.org/docs/Bailenson_etal-immersiveVR.pdf. Retrieved 2026-10-06.
- ↑ 10.0 10.1 Soo Youn Oh, Jeremy Bailenson, Nicole Krämer, Benjamin Li (2016-09-07). "Let the Avatar Brighten Your Smile: Effects of Enhancing Facial Expressions in Virtual Environments". PLOS ONE, vol. 11, no. 9, e0161794. https://doi.org/10.1371/journal.pone.0161794. Retrieved 2026-10-06.
- ↑ 11.0 11.1 Daniel Roth, Constantin Kleinbeck, Tobias Feigl, Christopher Mutschler, Marc Erich Latoschik (2018-03). "Beyond Replication: Augmenting Social Behaviors in Multi-User Virtual Realities". 2018 IEEE Conference on Virtual Reality and 3D User Interfaces (VR). IEEE. pp. 215-222. doi:10.1109/VR.2018.8447550. https://doi.org/10.1109/VR.2018.8447550.
- ↑ 12.0 12.1 Eugy Han, Mark R. Miller, Cyan DeVeaux, Hanseul Jun, Kristine L. Nowak, Jeffrey T. Hancock, Nilam Ram, Jeremy N. Bailenson (2023). "People, places, and time: a large-scale, longitudinal study of transformed avatars and environmental context in group interaction in the metaverse". Journal of Computer-Mediated Communication, vol. 28, no. 2. https://doi.org/10.1093/jcmc/zmac031. Retrieved 2026-10-06.
- ↑ Daniel Roth, David Mal, Christian Felix Purps, Peter Kullmann, Marc Erich Latoschik (2018-10-13). "Injecting Nonverbal Mimicry with Hybrid Avatar-Agent Technologies: A Naïve Approach". Proceedings of the Symposium on Spatial User Interaction (SUI '18). ACM. pp. 69-73. doi:10.1145/3267782.3267791. https://doi.org/10.1145/3267782.3267791.
- ↑ Jeremy N. Bailenson, Philip Garland, Shanto Iyengar, Nick Yee (2006). "Transformed Facial Similarity as a Political Cue: A Preliminary Investigation". Political Psychology, vol. 27, no. 3, pp. 373-385. International Society of Political Psychology (author copy, Stanford Virtual Human Interaction Lab). doi:10.1111/j.1467-9221.2006.00505.x. https://vhil.stanford.edu/sites/g/files/sbiybj29011/files/media/file/bailenson-political-cue.pdf. Retrieved 2026-10-06.
- ↑ Jeremy N. Bailenson, Shanto Iyengar, Nick Yee, Nathan A. Collins (2008). "Facial Similarity between Voters and Candidates Causes Influence". Public Opinion Quarterly, vol. 72, no. 5, pp. 935-961. https://doi.org/10.1093/poq/nfn064. Retrieved 2026-10-06.
- ↑ Nick Yee, Jeremy Bailenson (2007). "The Proteus Effect: The Effect of Transformed Self-Representation on Behavior". Human Communication Research, vol. 33, no. 3, pp. 271-290 (author's accepted manuscript). doi:10.1111/j.1468-2958.2007.00299.x. https://nickyee.com/pubs/Yee%20&%20Bailenson%20-%20Proteus%20Effect%20(in%20press).pdf. Retrieved 2026-10-06.
- ↑ 17.0 17.1 Portia Wang, Mark R. Miller, Jeremy N. Bailenson (2023-03). "The Belated Guest: Exploring the Design Space for Transforming Asynchronous Social Interactions in Virtual Reality". 2023 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW). IEEE. pp. 617-618. doi:10.1109/VRW58643.2023.00151. https://doi.org/10.1109/VRW58643.2023.00151.
- ↑ Daniel Roth, Peter Kullmann, Gary Bente, Dominik Gall, Marc Erich Latoschik (2018-10). "Effects of Hybrid and Synthetic Social Gaze in Avatar-Mediated Interactions". 2018 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct). IEEE. pp. 103-108. doi:10.1109/ISMAR-Adjunct.2018.00044. https://doi.org/10.1109/ISMAR-Adjunct.2018.00044.
- ↑ Daniel Roth, Gary Bente, Peter Kullmann, David Mal, Chris Felix Purps, Kai Vogeley, Marc Erich Latoschik (2019-11-12). "Technologies for Social Augmentations in User-Embodied Virtual Reality". 25th ACM Symposium on Virtual Reality Software and Technology (VRST '19). ACM. pp. 1-12. doi:10.1145/3359996.3364269. https://doi.org/10.1145/3359996.3364269.