Generative AI in the Wild: Prospects, Challenges, and Strategies

Title of the Paper

Generative AI in the Wild: Prospects, Challenges, and Strategies

Paper Information

• Subject Area: Research on the application and impact of generative artificial intelligence in the creative industries and human-computer interaction
• Keywords: Generative AI, human-AI collaboration, transparency, user agency, creative industries, learning-using-assessing framework, user feedback

Research Background and Problems

• Identified Problems or Challenges:
  1. Generative Artificial Intelligence (GenAI) demonstrates immense potential but still faces technical limitations that affect its effectiveness in real-world applications.
  2. The role of GenAI in creative workflows extends beyond tool usage to a dynamic, iterative "human-AI co-creation" process, yet current research has not fully explored its complexity.
  3. Users face challenges in learning technical details, adapting to tool functionalities, and addressing non-functional issues.
• Why It Matters:
  1. As a disruptive technology, GenAI is driving the transformation of creative industries from traditional workflows to automation and collaboration.
  2. The creative industries demand high originality, making it crucial to study how GenAI supports creative generation.
  3. Users' real experiences and feedback reveal pain points in technology design, ethics, and functional limitations, providing direction for future development.
• Research Motivation and Related Work: Through background research, the authors found:
  • The HCI field has extensively studied user perceptions and human-computer interaction with conventional AI, but the unique characteristics of GenAI lack comprehensive exploration from a human-centered perspective.
  • While studies on the practical applications of GenAI, particularly in tasks like creative writing and music composition, exist, the dynamic learning and practice processes of users in real-world scenarios remain underexplored.
  • The authors aim to fill this research gap by focusing on GenAI usage behaviors in the creative industries.

Solution

• Proposed Solution or Method: The authors propose an analytical approach based on the LUA framework (Learning, Using, Assessing) to study in detail how humans learn, use, and evaluate generative AI.
• Innovative Points:
  1. Adopting a dynamic "human-AI co-creation" perspective rather than static tool usage examples.
  2. Refining the analysis through the iterative processes of learning, using, and assessing, uncovering how users address technical and non-functional challenges through experimentation and strategy.
  3. Exploring the multidimensionality of user perceptions, such as differing opinions on GenAI creativity and its profound impact on workflows.
• Implementation Steps and Key Techniques:
  1. Conducting semi-structured interviews with 18 participants to collect real-world experiences of using GenAI in the creative industries.
  2. Using grounded theory to code and identify themes from the interview content.
  3. Decomposing the human-computer interaction process into three parts—learning, using, and assessing—via the LUA framework and constructing a comprehensive cross-stage analysis.
  4. Extracting user agency behaviors, such as tool selection and the development of personalized "prompt strategies," to address controllability issues in practical use.

Research Findings

• Specific Findings:
  1. Learning Phase:
     • Abundant learning resources: Most users acquire technical knowledge from training courses (e.g., Deeplearning.AI) and online tutorials (e.g., Twitter, YouTube).
     • Challenges include inconsistent resource quality, rapid technological iteration, and a lack of non-English resources.
     • Users strategically obtain accurate information through official documentation and social media influencers.
  2. Using Phase:
     • GenAI enhances content creation efficiency, driving a shift in work modes from team collaboration to individual independent creation.
     • Usage challenges include uncertainty, inefficient feedback mechanisms, engineering-oriented technical design, and a lack of localized customization.
     • User agency is reflected in selecting appropriate tools, adopting multi-tool integration, and developing personalized prompt strategies.
  3. Assessing Phase:
     • GenAI-generated content excels in language fluency and diversity, enhancing non-native users' expressive abilities.
     • Users hold differing views on "new creativity," with some attributing creativity more to human leadership.
     • Challenges include a lack of clear copyright disclosure norms, content authenticity (hallucination issues), and regulatory compliance.
     • Users adopt strategies such as labeling content as "AI-generated," manually fact-checking, and avoiding high-risk scenarios.
• Advantages:
  • Provides a better explanation and construction of the complexity of "human-AI collaboration" in the GenAI era, advancing HCI research from static tasks to dynamic creative processes.
  • Reveals technical limitations and user strategies based on real-world cases, offering practical evidence for improving technology design and responsible AI.
• Experimental and Evaluation Results:
  • Users' proactive learning and iterative development of prompt techniques significantly enhance GenAI adaptability, combining traditional and emerging tools to create new workflow models.
  • Content sharing on social platforms has a broad impact on knowledge acquisition and technology dissemination.
• Limitations and Future Directions:
  1. Participants were primarily from the creative industries; future research should expand to other fields (e.g., academia, healthcare).
  2. Conduct cross-cultural comparative studies in non-single-country contexts.
  3. Continuously evaluate functional improvements and emerging challenges (e.g., content authenticity) as GenAI rapidly evolves.

This study lays the cognitive and design foundation for broader real-world deployment and the design of next-generation generative AI tools.