How the Role of Generative AI Shapes Perceptions of Value in Human-AI Collaborative Work

Research Background and Issues

• What problems or challenges did the authors identify?
  Traditional research often adopts a binary perspective of "whether to use AI" to study human-AI collaboration, overlooking the nuanced differences in the roles AI can play in collaboration. Moreover, past studies have primarily focused on traditional AI rather than Generative AI (GenAI), which not only identifies patterns but also generates new content. Its unique capabilities warrant deeper exploration. This study aims to address how GenAI influences creativity and the perception of value in work.

• Why is this issue important?
  With the widespread application of Generative AI across industries, the design of its roles in collaboration and its impact on human participants and external evaluations will profoundly shape future creative workflows and industry standards.

• Research Motivation and Related Work
  Current discussions on AI-human collaboration have not sufficiently examined the impact of AI roles on the perception of outcomes. Particularly with Generative AI, which can participate in tasks as a "driver" or an "advisor," the effects of these role differences on work quality, creativity, and value perception remain unclear. This study seeks to fill this theoretical and empirical gap.

Solutions

• What methods or solutions did the authors propose?
  The authors developed a structural model to theorize how different GenAI roles alter work quality and creativity, influence internal creative evaluations, and affect external evaluators' perceptions of value.

  • GenAI as a "driver" is responsible for generating initial content, which is subsequently modified by humans.
  • GenAI as an "advisor" provides feedback and optimization suggestions for human-generated content.

• What is innovative about this solution?
  The study introduces a dual-role framework for integrating Generative AI into workflows (driver and advisor) and empirically analyzes its specific impact on creativity and value perception in work outputs. This fine-grained role classification goes beyond the simplistic binary explorations of previous studies, proposing a more complex collaboration framework.

• What are the implementation steps and key technologies used?

  1. Experimental Design: Randomly assigning two AI roles (driver/advisor) and two task types (functional tasks/creative tasks) to examine AI's effects under different conditions.
  2. Survey and Measurement: After completing tasks, participants answered questionnaires about AI's impact on work quality improvement, effort, creativity evaluation, and external value perception.
  3. Data Analysis and Modeling: Structural Equation Modeling (SEM) was used to analyze experimental data and test hypotheses.

Research Findings

• What specific findings were obtained?

  1. When GenAI worked as an "advisor," participants perceived a significant improvement in work quality. When GenAI acted as a "driver," it was seen as accomplishing more tasks, diminishing the visibility of human contributions.
  2. While AI improved work quality and increased internal evaluations of creativity, it also raised concerns about the devaluation of AI-assisted work by external observers.
  3. External observers might assign higher "reverse enhancement" ratings to work heavily reliant on AI, contrary to initial expectations.

• What advantages does it have compared to existing solutions?

  1. Clearly distinguishes between multiple roles of Generative AI and their functions in different collaboration modes.
  2. Introduces heterogeneity effect analysis based on task types, providing a more nuanced research perspective.
  3. Uses empirical data to reveal the complex trade-offs between creativity and value perception.

• What are the experimental or evaluation results?

  • Compared to the driver role, the advisor role of GenAI significantly enhanced participants' perception of the quality of final outputs (model path coefficient b=0.260, p<0.01).
  • When GenAI acted as a driver, participants perceived a greater workload contribution from AI (path coefficient b=-0.821, p<0.001).
  • Regarding "creativity" evaluation, AI's quality improvements significantly increased creativity perception (b=0.303, p<0.01), but the reduction in creativity perception due to AI leading over human contributions was not significant (b=-0.131, non-significant).

• Limitations and Future Directions:

  1. Limitations:
     • The study sample primarily consisted of white college students, lacking diversity and limiting the generalizability of the results.
     • The experimental design was conducted in a laboratory setting, which may not fully reflect real-world work contexts.
     • Creative tasks may have been influenced by high cognitive load, affecting result differentiation.
  2. Future Directions:
     • Expand the sample to include participants from more diverse professions and industry backgrounds.
     • Consider longitudinal studies to observe changes in perceptions of GenAI over time.
     • Explore the role of transparent AI participation in mitigating societal biases against AI-generated work.

Through this research, the paper provides profound insights into designing "human-GenAI" collaboration models that genuinely enhance human capabilities.