Source-linked AI summary
Sharing Roughness with Hand-Outline Visualization to Reduce Sensory Asymmetry in VR Collaboration
Minju Baeck, Yoonseok Shin, Hyunjin Lee, Boram Yoon, Sang Ho Yoon, Woontack Woo
TL;DR
Asymmetric haptic access can keep tactile evidence private to one collaborator, leaving the other to infer roughness visually. The paper externalizes coarse roughness through hand-anchored cues based on line shape and motion, validated their visual–haptic correspondence, and evaluated cue visibility in collaboration. The visualization reduced completion time when shown on both hands, while non-haptic-user visibility was associated with higher confidence and perceived contribution without compromising social presence.
Problem
Asymmetric access to haptic hardware leaves tactile information available only to some collaborators, while cue design for crossmodal asymmetric collaboration remains unclear.
Method
The paper uses hand-anchored visual proxies based on contour shape and motion to externalize three coarse roughness levels, validated through two preliminary studies and a collaborative sorting study.
Results
Showing the roughness visualization on both users’ hands reduced completion time relative to no visualization, while non-haptic-user cue visibility was associated with greater confidence and perceived contribution.
Takeaways & Limitations
Abstract hand-anchored roughness cues can restore shared access to object-related tactile information and support more balanced participation in asymmetric VR collaboration.
Takeaways & Limitations
The study used dyads with fixed asymmetric roles and a controlled laboratory setting, leaving scalability, ecological validity, and generalization to other tactile qualities open.
Abstract
from arXiv · showhide
In collaborative VR, asymmetric access to haptic hardware creates a critical information gap: tactile evidence remains private to the haptic user, hindering the shared understanding needed for joint decision-making. While prior work has explored crossmodal sensory cues in virtual environments, it remains unclear how such cues should be designed for asymmetric collaboration, where collaborators receive information through different modalities. In our setting, the haptic user feels roughness through fingertip vibration, whereas the non-haptic user relies on vision alone. To reduce this asymmetry, we propose externalizing an object's tactile state through a glanceable hand-outline visual proxy. Specifically, we examine whether abstract visual roughness cues based on line shape and motion can encode three discrete roughness levels for both haptic and non-haptic users. Two preliminary studies establish a shared visual semantics by identifying visually distinguishable cues for non-haptic users and validating their visuo-haptic correspondence for haptic users. In a main study of a collaborative sorting task, showing this visualization on both users' hands significantly reduced completion time relative to a no-visualization baseline. Moreover, NU-side cue visibility was associated with higher confidence and perceived contribution for the non-haptic user. These findings show that hand-anchored abstract visual cues provide a lightweight means of externalizing object tactile state, reducing information asymmetry without compromising social presence.
1 INTRODUCTION
Collaborative VR can leave tactile information private to haptic users, creating asymmetric access during joint assessment. This work externalizes coarse roughness through dynamic hand-anchored visual cues and studies their shared visual–haptic meaning and placement.
- Asymmetric collaboration occurs when users have unequal access to devices, roles, or task-relevant information.
- Haptic users can directly access roughness, while non-haptic users must infer it through observation or conversation, creating a potential information gap.
- The approach uses glanceable hand-anchored proxies whose shape and motion encode three coarse roughness categories: smooth, medium, and rough.
- Two preliminary studies establish a shared visual–haptic roughness mapping by testing non-haptic visual discriminability and haptic correspondence.
- The contributions include a lightweight visualization method that externalizes object roughness without additional hardware.
- Cue placement is treated as a collaboration design variable because visibility may shape efficiency, participation, and workload equity.
2 RELATED WORK
Prior collaborative VR research has shared visual, spatial, and sometimes haptic information, but asymmetric access to tactile material states remains underexplored. This work connects crossmodal roughness perception with the question of how tactile-state cues should be distributed across collaborators.
- Collaborative VR supports design review and spatial communication, but sharing roughness under uneven haptic access remains less explored.
- Multi-user haptic systems can share roughness or weight, yet typically assume that multiple users have direct haptic access.
- Roughness can be shaped by visual motion and contour shape in addition to vibrotactile stimulation.
- Categorical tactile perception and visuo-haptic integration motivate using a small number of perceptually distinguishable roughness levels rather than high-fidelity reproduction.
- Shared visualizations of hidden partner states, such as hand force, can improve collaborator awareness and social presence.
- Two open questions concern whether tactile-state cues are interpretable to non-haptic users and how cue access should be distributed.
3 TRANSLATING HAPTIC CUES INTO VISUAL CUES
The paper develops hand-anchored visual proxies that translate vibrotactile roughness into three coarse, shared visual levels. Two preliminary studies identify distinguishable shape–motion cues for non-haptic users and validate their correspondence with haptic roughness levels.
- 3.2 Preliminary Study 1: Visual Discriminability for Non-haptic Users: The Bradley–Terry ordering was sharp_fast > sharp_slow > sharp_static > curve_fast > curve_static > curve_slow > none, supported by roughness ratings.The stimulus effect was significant, χ2(6) = 81.08, p < .001, with Kendall’s W = .845.
- 3.3 Preliminary Study 2: Visuo-Haptic Correspondence for Haptic Users: Preliminary Study 2 validated a three-level visuo–haptic mapping by pairing visual candidates with fixed Low, Middle, and High vibrotactile levels.The selected mappings were sharp_fast for High, sharp_slow for Middle, and curve_fast for Low.
- 3.3 Preliminary Study 2: Visuo-Haptic Correspondence for Haptic Users: Condition significantly affected MatchScore and Perceived Roughness but not Confidence across the nine visual–haptic pairings.The reported effects were MatchScore χ2(8) = 35.20, p < .001, W = .28; Perceived Roughness χ2(8) = 120.00, p < .001, W = .94; and Confidence χ2(8) = 12.70, p = .122, W = .10.
- 3.3 Preliminary Study 2: Visuo-Haptic Correspondence for Haptic Users: Participants judged visual–haptic matching primarily through edge sharpness, motion speed, and temporal congruence between vibration and visual motion.Sharper and faster-moving cues were associated with higher roughness, while static variants appeared less aligned with rhythmic vibration.
- 3 TRANSLATING HAPTIC CUES INTO VISUAL CUES: Hand-anchored visual proxies externalize coarse roughness through line shape and motion rather than realistic tactile reproduction.The design targets rapid comparison and shared understanding across collaborators with unequal haptic access.
- 3.2 Preliminary Study 1: Visual Discriminability for Non-haptic Users: Preliminary Study 1 found that sharp contours were judged rougher than curved contours, while motion modulated perceived roughness in a shape-dependent manner.Across 21 stimulus pairs, 19 differed from chance after Holm correction; the two non-discriminable pairs were both within the Curve family.
4 EVALUATING CUE VISIBILITY IN ASYMMETRIC COLLABORA-
The main study examined how placing a hand-anchored cue on the haptic user’s and/or non-haptic user’s hand affects asymmetric collaboration. It used four visibility modes in a collaborative roughness-sorting task with asymmetric HU–NU dyads.
- Study Design: The 2×2 placement design produced four modes: Off, H, N, and Both.Off was the no-cue baseline; H showed the cue on the haptic user’s hand, N on the non-haptic user’s hand, and Both on both hands.
- Participants and Task: The study involved 24 HU–NU dyads, with one participant using a vibrotactile glove and the other interacting without haptic hardware.The main task required dyads to sort nine visually identical cubes into Low, Mid, and High roughness zones.
- Procedure: Mode blocks were counterbalanced using a Latin-square schedule to reduce practice and fatigue effects.Each mode was administered as a block with two trials, and cube roughness assignments varied across trials.
- Implementation: The system synchronized shared object states, interaction events, and outline-cue visibility across networked VR clients.The setup used two Meta Quest 3 headsets connected to separate PCs via Meta Quest Air Link.
- Measures: The experiment measured completion time, success rate, cube grasp time, workload, social presence, collaboration ratings, and mode preference.Collaboration ratings covered Confidence, Teamwork Quality, Contribution, and Partner Expression Trust.
5 RESULTS
The visualization improved sorting efficiency without reducing accuracy, while cue placement changed workload balance and which partner handled tactile-state evidence. Non-haptic users reported higher confidence and contribution when the cue was visible on their hands, while social presence remained stable.
- Objective Measures: Both reduced completion time to 38.19 s versus 49.83 s for Off, while success rate remained high at 91.0–95.1% across modes.The completion-time difference was significant; success rate did not differ across modes.
- Objective Measures: Cue placement shifted grasp activity between partners: HU grasp time was longer in H, whereas NU grasp time was longer in N.HU/NU grasp times were 13.27(6.96)/10.01(3.58) s in H and 9.38(3.50)/10.63(4.04) s in N.
- Workload: Both lowered workload relative to H and reduced the HU–NU workload gap compared with Off.Workload was lower in Both than H (p = .045), and the gap in Off was significantly greater than in Both, H, and N (all p ≤ .020).
- Workload: Both produced higher perceived performance than Off, while H produced higher temporal demand than Both and N.Performance means for Both were 78.64/80.64 versus 71.77/62.55 in Off; temporal-demand means for H were 52.14/46.23 versus 38.45/34.23 in Both and 41.36/34.55 in N.
- Collaboration Ratings: NU confidence was lowest in Off (3.23) and highest in Both (5.96) and N (5.90), while NU contribution was lowest in Off (3.51) and highest in Both (6.13).These patterns indicate that NU-side cue visibility narrowed confidence differences and shifted perceived contribution toward NUs.
- Social Presence and Preference: Social presence showed no significant effects of Mode, Role, or their interaction, and Both was most preferred by both roles.Preference rankings followed Both, N, H, and Off for both HUs and NUs.
6 DISCUSSION
The discussion frames hand-anchored roughness cues as shared, actionable tactile evidence rather than haptic replication. Cue placement influenced efficiency, participation, workload, and preference, with bilateral placement offering complementary benefits but no universal advantage.
- Shared tactile access: Tactile-state visualization restored shared access to object-related tactile information rather than reproducing haptic sensation.The authors position the approach as a coarse shared reference for collaborative material judgment.
- Efficiency and social presence: Showing task-relevant cues on both hands reduced completion time relative to Off while social presence remained stable across conditions.The contrast distinguishes actionable tactile evidence from avatar representations that primarily support interpersonal awareness.
- NU participation: NU-side placement reduced the NU’s information disadvantage and supported higher confidence, perceived contribution, and more active interpretation of object-related tactile state.The workload imbalance between HU and NU was also reduced when tactile-state information became visually accessible.
- Placement trade-offs: Both was most preferred by both roles, reduced completion time relative to Off, and produced lower workload than H.However, the results found no clear performance advantage of Both over N.
- Design implication: Cue placement should be treated as a collaboration policy because it changed how tactile information was distributed and used.Off restricted tactile information to the HU, N partially restored NU access, and Both made the cue a shared reference.
- Abstract tactile cues: Abstract metaphorical cues can support rapid collaborative judgments of coarse tactile categories without physically faithful texture replication.The hand-anchored proxy is suited to tasks requiring quick categorical judgments rather than precise material verification.
7 LIMITATIONS AND FUTURE WORK
The study’s evidence is bounded by its dyadic, fixed-role laboratory setting, its focus on roughness proxies, and its limited assessment of communication and shared percepts. Future work should test broader users, tactile qualities, contexts, placements, communication patterns, and perceptual alignment.
- Study scope: The experiment used dyads with fixed asymmetric roles, leaving larger groups, role switching, and heterogeneous devices untested.This limits direct extension to multi-user or dynamically reassigned collaborations.
- Tactile qualities: The visualization focused on line shape and motion speed as roughness proxies, so generalization to stickiness, softness, or compliance remains unclear.The limitation concerns both the selected tactile property and the specific visual encoding.
- Ecological validity: The controlled laboratory setup used specific hardware, requiring ecological-validity testing in more realistic collaborative scenarios.The authors identify realism of setting and equipment as open questions.
- Placement context: Cue-placement benefits may change with object size, visual clutter, or task complexity, so context-dependent policies require further testing.The authors do not establish that Both, H, or N is universally preferable.
- Communication: Natural voice communication was not systematically recorded or coded, leaving its interaction with cue placement and social presence unresolved.Future studies should measure verbal exchanges alongside visualization conditions.
- Shared perception: The study tested task-level shared use of the cue, not whether HUs and NUs formed identical subjective roughness percepts.The authors limit their claim to collaborative sorting performance and call for direct perceptual measurement.
8 CONCLUSION
The paper presents a dynamic hand-outline visualization for roughness in asymmetric VR collaboration. Preliminary studies supported three visual–haptic mappings, and the main study found improved efficiency, workload distribution, and NU confidence and contribution while maintaining accuracy.
- Visual–haptic mapping: Contour sharpness was the dominant visual cue for roughness, while motion modulated perceived roughness in a shape-dependent manner.Together, these findings supported three discrete visual–haptic mappings.
- Main findings: Hand-outline visualization improved collaboration efficiency, redistributed workload, and enhanced non-haptic users’ confidence and contribution while maintaining overall accuracy.The conclusion frames these outcomes as evidence that abstract roughness visualization can reduce sensory asymmetry.
- Implications: The work offers design implications for multimodal feedback in collaborative VR by supporting more balanced participation through abstract tactile visualization.The stated future directions include finer tactile qualities, multi-user contexts, and applications such as education.
DISCLOSURE ON GEN AI USAGE
The authors disclose using ChatGPT for icon assets, reflected user images in the teaser figure, and English editing or proofreading. They state that the scientific work and conclusions were developed and verified by the authors.
- Generative AI use: ChatGPT assisted with icon assets, reflected user images in the teaser figure, and English editing or proofreading.The disclosure attributes scientific content, experimental design, implementation, data analysis, and conclusions to the authors.