Morphing Identity: Exploring Self-Other Identity Continuum through Interpersonal Facial Morphing Experience

Title of the Paper

Morphing Identity: Exploring Self-Other Identity Continuum through Interpersonal Facial Morphing Experience

Paper Information

• Subject Area: Human-Computer Interaction and Social Facial Perception
• Keywords: Face, Morphing, Self-Other Identity, Illusion, Interaction Design, Real-Time Systems, Telecommunication, GAN, User Experience, Social Cognition

Research Background and Problem

• Identified Issues or Challenges:
  • The face is a crucial component of self-identity and social interaction, but existing technologies or research are limited to static images or require controlled laboratory environments.
  • The application of cross-user facial morphing experiences in dynamic, real-time human-computer interaction remains underexplored.
  • The exploration of the shifting boundaries between self and other facial identities in a social cognition context is still an under-researched area.
• Significance:
  • Facial identity plays a central role in self-perception, body ownership, and social relationships.
  • Investigating the self-other boundary can provide insights for cross-disciplinary applications such as telecommunication, virtual reality, psychological experiments, and social application design.
• Research Motivation and Related Work:
  • To develop a system capable of real-time facial feature morphing, supporting interaction while retaining rich social context.
  • Drawing on psychological research on "illusory ownership" and GAN generation techniques to explore social cognition and identity perception in real-time scenarios.

Solution

• Proposed Solution:
  • Develop a real-time interactive system called “Morphing Identity” that can merge the facial identities of two users while preserving their facial dynamic signals.
  • The system uses GAN (Generative Adversarial Network) to generate a series of interpolated facial images, combined with facial reenactment technology, allowing users to control the dynamically changing facial representation during face-to-face interaction.
• Innovative Aspects of the Solution:
  • Combining GAN-generated static facial images with real-time dynamic facial signals to create a highly realistic and interactive experience.
  • Conducting live experiments across social relationships, such as family, friends, and strangers, to explore identity perception contrasts in different contexts.
  • Analyzing the self-other facial boundary from both quantitative and qualitative perspectives, introducing the concept of the "Self-Other Continuum."
• Implementation Steps and Techniques:
  • Design a technical pipeline, including StyleGAN2 for facial image mapping, interpolation, and generation, as well as facial dynamic signal reenactment technology (First Order Motion Model).
  • Collect user feedback in exhibition scenarios: set up a symmetrical semi-open exhibition space, providing interactive devices (e.g., mirrored displays and foot pedals) to enhance the experience.
  • Data analysis includes quantitative research on self-identity boundaries and qualitative studies of user experiences through post-interaction interviews.

Research Outcomes

• Specific Outcomes:
  • Users identified a "window" for facial identity boundaries, where at approximately 20% morphing rate, users still perceived the displayed facial image as their own.
  • Multiple possible user experiences were observed, including extended identity, shared identity, and even intertwined identity. These experiences demonstrated how the system could stimulate social cognition, such as enhancing intimacy, fostering familial connections, and creating shared narratives.
• Advantages Over Existing Solutions:
  • Real-time tracking and adjustment of facial dynamics provide a more natural interactive experience.
  • Offers quantitative analysis methods (e.g., using GAN latent vector distances), enhancing comparability with psychological experiments.
  • The system is applicable to diverse social scenarios, including video communication, virtual reality, and interactive museum exhibitions.
• Experimental or Evaluation Results:
  • Analysis of 9,269 data sets revealed quantitative patterns of self-facial boundaries and observed the effects of social relationships, age, and gender on these boundaries.
  • Interview feedback indicated that the experience could enhance intimacy and other positive social emotions, particularly during interactions with family members or partners.
• Limitations and Future Directions:
  • Limitations: Technical constraints make it challenging to fully generate and synchronize users' complete head dynamics; the polarization of experiences may require deeper exploration of individual differences.
  • Future Directions:
    1. Expand the research scope to include recognition windows for others' identities.
    2. Investigate additional factors influencing identity and interaction (e.g., video quality, network latency).
    3. Improve the precision of facial generation technology to support richer head dynamics.
    4. Apply this technology to more social scenarios, including education, healthcare, and corporate remote collaboration.

Conclusion

This study explores the boundaries of self and other identities in dynamic changes by designing a real-time interactive facial morphing system, proposing the theoretical framework of the "Self-Other Continuum." The findings contribute not only to understanding social cognition and identity perception but also provide significant insights and application scenarios for designing more interactive and emotionally engaging human-computer systems in the future.