Mixplorer: Scaffolding Design Space Exploration through Genetic Recombination of Multiple Peoples' Designs to Support Novices' Creativity

Title of the Paper

Mixplorer: Scaffolding Design Space Exploration through Genetic Recombination of Multiple Peoples’ Designs to Support Novices’ Creativity

Paper Information

• Domain: Human-Computer Interaction, Design Space Exploration, Creativity Support Tools
• Keywords: Design Exploration, Creativity Support Systems, Social Design Space Exploration, Genetic Recombination, Novice Designers

Research Background and Problem

• Problems and Challenges:

  • Novice designers face significant challenges in design space exploration due to their lack of experience and expertise, which limits their ability to generate high-quality and diverse solutions.
  • The design space for creative tasks is often ambiguous and ill-defined, making it difficult for traditional tools and algorithms to effectively support exploration activities.
  • Design fixation negatively impacts the diversity and creativity of design outcomes.

• Significance:

  • Encouraging novice designers to overcome design fixation and explore more innovative solutions is crucial for their long-term development and creative capabilities.
  • Technology-supported design exploration can help novices improve design quality and enhance collaboration and learning outcomes.

• Motivation and Related Work:

  • The research aims to develop tools that support novice designers in exploring the design space, overcoming design fixation, and stimulating creativity.
  • Related work has demonstrated the contributions of computer-supported design tools (e.g., parametric exploration, rule-based systems, and genetic algorithms) to design space exploration. However, traditional genetic algorithms face challenges in defining objective functions for creative design.

Solution

• Method/Solution:

  • Development of Mixplorer, a system based on genetic algorithms infused with the concept of social design exploration, enabling the "mixing" of designs from other designers to generate new designs.
  • Mixplorer features two core functional modules:
    1. Garden Design Interface: Allows users to create garden designs through drag-and-drop interactions.
    2. Garden Exploration Interface: Generates new designs by combining multiple designs through genetic crossover operations.

• Innovations:

  • Adopts a "social design space exploration" approach, allowing multiple users to upload designs in real-time as inspiration for the genetic algorithm, thus addressing the issue of insufficient initial design samples.
  • Replaces traditional fitness functions with human-in-the-loop selection, enabling users to choose designs that suit their preferences, solving the challenge of defining objective functions in creative design.

• Implementation Steps and Key Techniques:

  • Define the genetic representation of gardens (including object genes, structures, and the positions of individual elements) to support genetic operations.
  • Implement various genetic crossover operations (e.g., gene-level crossover, structure-level crossover) and mutation operations to generate new designs.
  • Use two heuristic methods to validate generated designs, ensuring design quality and diversity.

Research Outcomes

• Specific Outcomes:

  • Mixplorer enables novice designers to quickly generate diverse designs and significantly enhances creativity and exploratory capabilities.
  • Through the design mixing process, novices can better overcome design fixation and produce more novel designs.

• Advantages Compared to Existing Solutions:

  • Social design space exploration addresses the issue of insufficient initial design samples, and the genetic algorithm dynamically utilizes user-provided designs.
  • Human-referenced selection improves the adaptability of design tools to open-ended problems.

• Experimental or Evaluation Results:

  • Study 1 (Expert Interviews): Six landscape design instructors found Mixplorer easy to use and suitable for integration into classroom teaching, effectively helping novices consider more design possibilities.
  • Study 2 (Controlled Experiment): Among 64 participants, the creativity scores were highest when using the mixing functionality, resulting in more novel designs. Design exploration support further enhanced the diversity of design solutions.
  • Creativity Support Index (CSI) scores indicated that mixed designs better supported novice designers' creativity compared to no exploration or random exploration.

• Limitations and Future Directions:

  • Experiments were primarily conducted in simulated environments, requiring validation of the system's feasibility and effectiveness in real classroom settings.
  • The quality of generated designs has not yet been evaluated. Future research should involve expert assessments to further measure the practical value of design outcomes.
  • The applicability of the design mixing functionality to other design domains (e.g., interior design) remains to be tested.

Conclusion

Mixplorer demonstrates both theoretical and practical potential in supporting novice designers' exploration of design spaces through technology. The system employs an innovative social design space exploration approach and has been validated through extensive experiments to support creative practices and design outcomes. Future research directions include expanding application domains, evaluating design quality, and exploring the potential for multi-user collaborative design.