Intra, Extra, Read all about it! How Readers Interpret Visualizations with Intra- and Extratextual Information

Research Background and Problem

• Problem and Challenges: The authors found that readers' interpretation of data visualizations depends not only on the direct information within the visualization (intratextual information) but is also significantly influenced by external information (extratextual information) they possess. However, it remains unclear which types of intra- and extratextual information readers use and how they integrate these sources to understand visualizations.
• Significance: Better understanding how readers integrate these two types of information and formalize their meanings can explain why seemingly similar readers interpret the same visualization in vastly different ways. This understanding is crucial for designing visualization tools and methods that align more closely with users' cognitive processes.
• Research Motivation and Related Work: The authors compared the concepts of "intratextual/extratextual" from textual analysis and argued that this framework is applicable to the study of visualization interpretation. The study was inspired by prior literature on how individuals incorporate external information such as personal beliefs and background knowledge into chart interpretation. Its uniqueness lies in systematically analyzing the roles and integration of these two types of information in the specific interpretation process.

Solution

• Research Method: The authors designed an experiment based on semi-structured interviews, inviting six undergraduate students to observe four real-world data visualizations and conduct in-depth interviews about their interpretation processes.
  • Two analytical techniques were employed:
    1. Thematic Analysis: Used to categorize and summarize the intra- and extratextual information mentioned by participants.
    2. Diffractive Reading: A feminist analytical tool focusing on uncovering subtle differences and details in interview data to reveal how participants integrate information to form deeper understandings.
• Innovations:
  • Systematically identified the types of intra- and extratextual information readers use when interpreting visualizations.
  • For the first time, summarized three specific ways individuals integrate these two types of information:
    1. Guiding Focus: Readers use external information to determine which internal information is more important.
    2. Skipping Detail Assumptions: External information simplifies chart reading, sometimes leading to accurate interpretations and sometimes to misunderstandings.
    3. Deriving Higher-Level Meaning: Combining intra- and extratextual information to construct insights that go beyond the chart itself.

Research Findings

• Specific Findings:
  1. Type Classification: Participants utilized various intra- and extratextual information, including variable understanding, comparative value analysis, statistical principles, personal experiences, and personal identities.
  2. Integration Methods: Described three typical ways of integrating intra- and extratextual information to form interpretations.
• Comparison with Existing Solutions and Advantages:
  • Existing research primarily focuses on how domain knowledge influences chart interpretation. This study uniquely highlights the indispensable role of non-dominant epistemologies (e.g., life experiences, personal identities).
  • The study reveals that external information can both enhance understanding depth and introduce inaccurate assumptions, presenting the "double-edged sword" nature of information integration.
• Experimental Results:
  • Thematic Analysis found that participants tended to focus on high-level global features (e.g., variables, trends) rather than specific numerical details.
  • Diffractive Reading analysis revealed that participants often incorporated information not directly presented in the visualization through their personal experiences, identities, or prior knowledge.
• Limitations and Future Directions:
  • Limitations:
    1. Small sample size (6 undergraduate students) with relatively homogeneous backgrounds (STEM students).
    2. Only four data visualization samples were selected, not covering all visualization styles and topics.
    3. Findings rely on participants' explicitly expressed views, potentially overlooking unmentioned or difficult-to-articulate interpretation processes.
  • Future Research Directions:
    1. Expand the participant sample to include diverse identity backgrounds and cognitive styles.
    2. Explore how to design visualization tools and frameworks that actively support the interpretation of non-dominant information.
    3. Investigate whether "deep insights" necessarily depend on the integration of intra- and extratextual information.

Conclusion

This study systematically reveals the diversity of intra- and extratextual information readers use in visualization interpretation and their integration methods, providing robust theoretical support for understanding how individuals generate diverse interpretations of visualizations. Furthermore, the study proposes an important new direction: data visualization design should place greater emphasis on users' personal experiences and non-dominant cognitive styles to help them form accurate and profound insights more efficiently.