A study of UX Practitioners Roles in Designing Real-World, Enterprise ML Systems

Document Title

A Study of UX Practitioners' Roles in Designing Real-World, Enterprise ML Systems

Document Information

• Subject Area: Human-Computer Interaction (HCI), User Experience Design (UX), Design and Development of Machine Learning and Artificial Intelligence Systems
• Keywords: Artificial Intelligence (AI), Machine Learning (ML), User-Centered Design (UCD), Design Thinking, Interdisciplinary Collaboration, Design Methodology, Ethical Design, Complexity Management

Research Background and Issues

• What problems or challenges did the authors identify?

  • There is a lack of sufficient research on the design methodologies behind AI/ML system functionalities, particularly in real-world applications outside of large tech companies. Existing design methodologies such as User-Centered Design (UCD) have limitations when addressing the complexity of enterprise ML systems.
  • Collaboration between design and technical teams, as well as how design is integrated into the internal mechanisms of system functionalities, remains unclear.
  • Although attention to ethical issues (e.g., fairness, transparency, explainability) has increased, how these issues are addressed in actual design processes is still ambiguous.

• Why is this issue important?

  • With the widespread application of AI/ML systems across industries, their design must simultaneously address complex technical challenges and user interaction needs.
  • Enterprises face high-risk application scenarios (e.g., medical diagnosis and fraud detection), where the consequences could be costly or even life-threatening.

• Research Motivation and Related Work

  • The authors aim to fill theoretical gaps in the academic field by studying the behavior of UX practitioners in interdisciplinary teams.
  • Existing literature often focuses on technical aspects or is limited to interaction design for AI/ML, neglecting broader system design and team collaboration.
  • Understanding AI/ML design cases in small and medium-sized enterprises can complement the limitations of case studies focused on "large tech companies" and provide more generalizable insights.

Solutions

• What methods or solutions did the authors propose?

  • Data collection through three research methods: online surveys (targeting product managers responsible for AI/ML projects), semi-structured group interviews with interdisciplinary teams, and in-depth interviews with individual UX designers.
  • Qualitative and quantitative analysis of survey results to extract applicable design methodologies and collaboration models.
  • Recommendations for improving HCI design methodologies based on practical needs, including the integration of high-fidelity prototyping and iterative processes.

• What is innovative about this solution?

  • The authors focus on small and medium-sized enterprises rather than resource-rich large tech companies, emphasizing the real challenges faced by smaller teams.
  • By combining practical UX design processes with ethical considerations, they propose extending participatory design to the context of complex AI/ML systems.
  • They investigate collaboration models within teams, particularly how designers' skills influence technical teams' functionality and algorithmic choices.

• What are the implementation steps and key technologies used?

  • A survey involving 27 product managers to explore design methodologies and ethical considerations.
  • Group interviews and the use of UML (Unified Modeling Language) tools to map AI/ML design workflows.
  • Qualitative coding methods to organize interview data into themes, identifying major challenges and successful cases.

Research Findings

• What specific findings were achieved?

  1. Adaptation of Existing Design Methods:
     • UX designers can adapt to the complexity of AI/ML, extending HCI techniques to areas such as data variable selection, sample data modeling, and post-deployment maintenance.
     • A parallel workflow separating UX prototyping from AI/ML prototype development can prevent redundant work.
  2. Importance of Team Collaboration:
     • The success of overall design relies on extensive interdisciplinary collaboration. Technical teams increasingly value design skills, and designers are involved in data understanding and selection processes.
  3. Deficiencies in Existing Design Approaches:
     • Traditional HCI methods are overly focused on UI and UX, neglecting the contributions of design to system functionalities and algorithms.
     • Prototype testing methods (e.g., Wizard of Oz) have limited utility in enterprise environments, particularly regarding the practicality of manually labeled data.

• What advantages does it have compared to existing solutions?

  • It encourages designers to directly participate in decision-making at the system functionality level, rather than being confined to user interface design.
  • It highlights the gap between theoretical design methodologies and everyday enterprise practices, while emphasizing the importance of post-deployment maintenance (e.g., preventing algorithmic drift).

• What are the experimental or evaluation results?

  • High-fidelity prototypes (using real data and polished designs) demonstrated superior performance in team communication and gaining user trust compared to low-fidelity prototypes.
  • Through participatory design demonstrations, the authors illustrated how to effectively address complex issues such as transparency and fairness.

• Limitations and Future Directions

  • Limitations:
    • The sample is concentrated in specific enterprise environments, which may not fully generalize to broader application scenarios.
    • The data relies on subjective feedback from interviews, which needs to be cross-validated with actual user practices.
  • Future Directions:
    • Deepen research on collaboration models between UX practitioners and developers to build improved HCI methodologies.
    • Develop specialized design tools and frameworks for designing complex, high-risk AI/ML systems.
    • Standardize the integration of ethical design considerations and provide toolkit templates for sustainable AI design across enterprises of different scales.