Data Abstraction Elephants: The Initial Diversity of Data Representations and Mental Models

Title of the Paper

Data Abstraction Elephants: The Initial Diversity of Data Representations and Mental Models

Paper Information

• Subject Area: Data Abstraction and Visualization Design
• Keywords: Mental Models, Data Abstraction, Data Visualization, User Studies, Data Design, Visual Encoding, Data Communication

Research Background and Problem

• Problem or Challenge: Different individuals generate varying mental models for the same dataset, focus on different attributes, and select different visualization methods. This phenomenon has profound implications for the process of data visualization design but has not been thoroughly studied.
• Significance: Data abstraction and mental models play a critical role in the early stages of the visualization design process, providing guidance on how to understand and design visualizations that meet user needs.
• Research Motivation: To explore how individuals construct mental models in the absence of external data structure influences and to identify best practices for facilitating the expression and communication of user mental models, aiming to support more effective data abstraction and visual design decisions.

Solution

• Research Methods:
  • Created three datasets with diverse data abstraction potential (file system, miscellaneous drawer, power station) presented in paragraph form to avoid bias towards any specific data abstraction format.
  • Recruited 28 participants and asked them to express their understanding of the datasets through drawings, followed by semi-structured interviews.
• Innovations:
  • Used paragraph format to avoid abstraction bias.
  • Explored the richness of data abstraction and the processes of its formation and communication.
  • Provided an open database, including participant-generated sketches and interview transcripts.
• Implementation Steps:
  • Dataset design and pilot study.
  • Data collection via video conferencing, including drawing tasks and semi-structured interviews.
  • Inductive thematic analysis to code and analyze the collected data.

Research Findings

• Specific Outcomes:
  1. Developed a set of themes and codes describing mental models, data abstraction, their manifestations, and influencing factors.
  2. Illustrated the diversity of user data abstraction and its implications for visual design.
  3. Established an open database for reference in future research.
• Advantages:
  • Demonstrated the diversity and flexibility of user mental models, emphasizing their importance in data abstraction and visualization design.
  • Provided practical recommendations for the early stages of design.
  • Compared differences in data abstraction between users from computational and non-computational fields.
  • Built upon and supported findings from existing research on mental models.
• Experiments and Evaluation:
  • Conducted detailed experiments and coding analysis with 28 participants, revealing patterns in how users form mental models when processing datasets.
  • Found that participants proposed numerous original ideas regarding the tasks, content, and potential applications of the datasets.
• Limitations and Future Directions:
  • Limitations:
    1. Participant group was concentrated on individuals with high data literacy, excluding a broader range of user backgrounds.
    2. Dataset design did not encompass larger-scale, more complex, or dynamically changing data.
    3. Capturing mental models relied on participants' expressive abilities and the skills of the interviewers.
  • Future Directions:
    • Explore datasets with higher abstraction levels and more complex relationships to study their impact on mental model formation.
    • Expand the participant group to include users with lower data literacy.
    • Investigate how users construct mental models and choose appropriate data abstractions in the context of large-scale and dynamically changing datasets.

Conclusion

The study highlights the diversity and influencing factors of mental models and data abstraction when users process "small paragraph-style datasets." The findings provide important insights for data visualization design methods, such as how to more effectively capture user mental models and design abstraction frameworks that support data exploration. This further underscores the critical value of mental model research in the field of information visualization, paving the way for future in-depth studies on the interaction between mental models and user needs.