TL;DR
This paper introduces the "Social-Aware VR Conference Group Display Configuration" (SVGD) problem, addressing the challenge of overwhelming and irrelevant user displays in large-scale VR conferences. They propose a 2-approximation algorithm called SVC that balances personal preferences and social utilities to optimize user experience within limited display slots, outperforming baselines by 75% in solution quality.
What Problem Does This Paper Address?
Virtual Reality (VR) has become a popular platform for conferences and exhibitions, especially since the pandemic. However, current VR platforms often display all users uniformly, leading to a cluttered and overwhelming experience. This "obstructed view" can hinder social interactions, reduce user satisfaction, and even lead to "VR-induced Social Isolation." The core problem is how to intelligently configure what each user sees in their VR display, considering their personal preferences, social connections, and the limited number of display slots available, to maximize their overall satisfaction and facilitate meaningful interactions.
Methods
The paper introduces a new problem formulation: Social-Aware VR Conference Group Display Configuration (SVGD). This problem aims to find an optimal "SVD $k_u$-configuration" for each user, where $k_u$ is the maximum number of other users that can be displayed in their VR headset.
The proposed solution is a 2-approximation algorithm called Social Utility-Aware VR Conference Group Formation (SVC). SVC works by:
- Quantified Social Benefit Network (QSB Network): It constructs a network that embeds social utilities and personal preferences.
- Configuration Comparison: SVC iteratively selects between two strategies:
- P-Configuration: Prioritizes users based on individual personal preferences.
- S-Configuration: Prioritizes users based on social utilities (i.e., users who have high social utility with each other).
- Effective SVD Utility Ratio: To make decisions, SVC calculates an "effective SVD utility ratio" which normalizes the utility gained by adding a user to a display slot. This helps in making efficient choices within the limited $k_u$ slots.
- Iterative Selection: SVC greedily selects the configuration (P or S) that yields the greater effective ratio in each step, updating the network accordingly. This process continues until no more users can be added or all display slots are filled.
Datasets
The paper evaluates their approach on five real-world datasets:
- Timik
- Pokec
- Youtube
- SMMnet
These datasets represent social networks, and the authors assign personal preferences and social utilities based on existing work (Ko et al. 2020; Chen and Yang 2022).
Key Insights and Ideas
- The "VR-induced Social Isolation" is a real problem: Simply displaying everyone in VR doesn't foster connection; it can actively hinder it by creating overwhelming and irrelevant visual noise.
- Balancing is key: Neither purely preference-based nor purely social-based configurations are optimal. The sweet spot lies in a balanced approach that considers both what a user wants to see and who they should see for better interactions.
- Limited display slots are a crucial constraint: The $k_u$ parameter is not just a technical limitation but a core element that forces prioritization and intelligent selection.
- Co-display matters: The ability for two users to see each other (co-display) is essential for social utility to take effect, highlighting the importance of reciprocal visibility.
Key Results
- The proposed SVC algorithm outperforms 11 baseline approaches by an average of 75% in terms of solution quality (total SVD utility).
- SVC is also efficient, with a time complexity of $O(\max_{u}(k_u) \cdot |V| \cdot \log(|E|))$, making it suitable for large-scale VR events.
- Sensitivity tests show that SVC consistently achieves the best performance across different values of $k_u$ and $\lambda$ (the weighting factor for social utility).
Limitations and Future Work
- Obstructed View in Future Work: The authors acknowledge that their current work focuses on display configuration but plan to extend SVC for scenarios where the user's view might be physically obstructed, which is a different problem than display slot limitations.
- Generalization: The authors aim to extend SVC for more generalized scenarios, implying that the current formulation might be specific to conference settings.
- Parameter Tuning: While the paper mentions $\lambda$ can be set by users or learned, the optimal way to learn or set these parameters for diverse user groups is an area for further exploration.
Summary updated