Breaking the Midas Spell:Understanding Progressive Novice-AI Collaboration in Spatial Design
Project Overview
The document explores the integration of generative AI in spatial design education, emphasizing the significance of iterative human-AI collaboration over traditional AI tools that often deliver complete outputs in a single step, which can hinder user engagement and learning. Through a Wizard-of-Oz study focused on novice users' interactions with AI in spatial design, the research identified common workflows and specific user needs. The findings underscore the necessity for designing AI-assisted tools that promote creativity and enhance the learning experience for novice designers, suggesting that a more engaging collaborative process can lead to improved educational outcomes. Overall, the document advocates for a shift towards AI applications in education that foster progressive interaction and support the development of design skills.
Key Applications
Progressive human-AI collaboration framework in Spatial Design
Context: Education for novice spatial design learners
Implementation: Conducted a Wizard-of-Oz study to observe interactions between novice users and AI during spatial design tasks.
Outcomes: Enhanced user understanding of spatial design, improved creativity, and engagement through iterative AI assistance.
Challenges: Novice users struggled with complex spatial structures and often had unrealistic expectations of AI capabilities.
Implementation Barriers
Technical Barrier
Current AI models struggle with detailed and precise design tasks, particularly in spatial design contexts.
Proposed Solutions: Further development of specialized AI models tailored to spatial design tasks is necessary.
User Experience Barrier
Users often found AI interactions frustrating due to miscommunication and a lack of context-awareness from the AI.
Proposed Solutions: Implementing a history feature for AI to track user interactions and providing better guidance on AI capabilities.
Project Team
Zijun Wan
Researcher
Jiawei Tang
Researcher
Linghang Cai
Researcher
Xin Tong
Researcher
Can Liu
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Zijun Wan, Jiawei Tang, Linghang Cai, Xin Tong, Can Liu
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai