The Entoptic Field Camera as Metaphor-Driven Research-through-Design with AI Technologies

Title of the Paper

The Entoptic Field Camera as Metaphor-Driven Research-through-Design with AI Technologies

Bibliographic Information

• Subject Area: Application of Artificial Intelligence (AI) technologies in design research; focus on the potential impact of technology on real-world experiences.
• Keywords: Research-through-Design (RtD), Artificial Intelligence (AI), Generative Adversarial Networks (GAN), image synthesis technology, technological mediation, metaphor-driven design, AI literacy.
• Conference: CHI Conference on Human Factors in Computing Systems (CHI ’23), April 23-28, 2023, Hamburg, Germany.
• Authors: Jesse Josua Benjamin, Heidi Biggs, Arne Berger, et al.

Research Background and Problem

• Problem or Challenge:

  • The widespread application of AI technologies in the field of image processing has significantly transformed the processes of image generation and modification, yet the profound impact of these changes on real-world experiences remains underexplored.
  • Beyond obvious user application scenarios, the "reality shaping" effects of AI technologies often remain hidden, despite their potential to provide critical insights into ethical, aesthetic, and epistemological dimensions.

• Significance:

  • Current discussions on AI technologies are predominantly focused on their explicit functionalities or technical issues, neglecting their potential mediation effects.
  • As AI technologies become increasingly integrated into public applications, understanding how they alter perceptions and experiences of reality is crucial for developing responsible AI design.

• Research Motivation and Related Work:

  • The authors were inspired by a theory of technological mediation, which posits that technology, akin to language, can render the world "readable" or comprehensible in specific ways.
  • This study aims to explore the profound impact of AI on real-world experiences by introducing a novel metaphor (entoptic phenomena) and opening new pathways for design research.

Solution

• Methods and Solutions:

  • The authors propose a metaphor-driven research method, utilizing "entoptic phenomena" as an analytical tool to emphasize how data and model interactions in AI technologies resemble the collaboration between the eye and brain in the visual system.
  • The metaphor is materialized through the design prototype "Entoptic Field Camera," a GAN-based image processing tool.

• Innovations:

  • By employing a metaphor-driven design process, AI technologies are treated as materials for design experimentation rather than mere tools or objectives. This approach transforms technical issues into a design space that designers can intuitively manipulate and reflect upon.
  • A non-functional metaphor (abstract dynamic interactions between technologies) is translated into a tool to explore the impact of AI technologies on knowledge generation and practice.

• Implementation Steps and Key Technologies:

  • Phase 1: Using GAN architectures and open-source platforms (e.g., RunwayML) with models such as BigBiGAN and HiFiLL to perform style generation and content filling on user-input images.
  • Phase 2: Developing a mobile-friendly web application that allows users to capture and experience image processing under technological intervention in real-world scenarios.
  • Phase 3: Testing and reflecting on the prototype through first-person design practices by the authors and field experiments conducted by research participants.

Research Outcomes

• Specific Outcomes:

  • Designed and deployed the "Entoptic Field Camera" as a tangible prototype.
  • Collected reflections and practical insights from users during field experiments, focusing on how AI technologies influence perceptions of reality.

• Advantages Compared to Existing Solutions:

  • Shifted from the traditional functional improvement of AI technologies to a deeper exploration of their structural impact on real-world experiences.
  • Enabled a transition in AI technology application from simplified interactions to diversified subjective reflections, inspiring deeper value creation in design.

• Experimental and Evaluation Results:

  • The experiments revealed AI-related practices, including:
    • How AI-generated images prompt users to reflect on their relationship with the self and the world.
    • The creative potential of GAN models in image processing, challenging traditional photography's demand for authenticity.
    • How users, through participation, develop specific literacies related to generated images (i.e., understanding and interpreting these images).

• Limitations and Future Directions:

  • Limitations:
    • The experimental participants were primarily academic teams closely associated with design research, limiting sample representativeness.
    • The broader applicability of metaphor-driven design in practical societal contexts remains unconfirmed.
  • Future Directions:
    • Incorporate broader participatory design practices to ensure research outcomes serve the general public and diverse groups.
    • Further explore the potential contributions of metaphor-driven design to AI technology regulation and ethical design.
    • Investigate how 21st-century design practices can transcend the limitations of current design materials and expand into more complex frameworks of technological mediation.

This study, through metaphor-driven design exploration, provides a new perspective on the profound impacts of AI technologies while offering methodological references for human-computer interaction and design research.