When do Data Visualizations Persuade? The Impact of Prior Attitudes on Learning about Correlations from Scatterplot Visualizations

Title of the Paper

When do Data Visualizations Persuade? The impact of prior attitudes on learning about correlations from scatterplot visualizations

Paper Information

• Subject Area: Data visualization and science communication, particularly its role in attitude and belief change
• Keywords: Data visualization, belief updating, attitude change, uncertainty representation, science communication, political polarization, persuasiveness, correlation estimation

Research Background and Questions

• What problems or challenges did the authors identify?

  • Data visualization is a critical tool for science communication, but its actual persuasiveness, particularly when challenging existing attitudes and beliefs, remains uncertain.
  • In existing research, people's reception and interpretation of data are often influenced by their prior attitudes and values, a phenomenon especially evident in politically polarized topics (e.g., COVID-19 vaccines).
  • It remains unclear whether visualizations using statistical evidence can effectively change audience attitudes or behaviors.

• Why is this issue important?

  • In current societal issues (e.g., public health and economic policies), guiding the public to correctly understand data is crucial. However, when data conflicts with the audience's prior attitudes, designing effective visualizations to promote belief updating or even attitude change is a critical challenge.

• Research motivation and related work:

  • Psychological theories distinguish between beliefs and attitudes, and prior studies have shown that attitudes influence how people process data. However, systematic research on this attitude influence in the field of data visualization is limited.
  • Previous studies have focused on how data visualizations promote learning and belief updating, but the mechanisms of attitude change and the impact of uncertainty representation require further exploration.

Proposed Solution

• What methods or solutions did the authors propose?

  • Conducting experiments with data visualizations to study how audience attitudes influence their interpretation of statistical relationships and belief updating.
  • Using model comparisons and experimental data analysis to explore the potential impact of uncertainty representation (e.g., confidence intervals) on belief updating and attitude change.

• What is innovative about this solution?

  • Building on existing research on belief updating, the study further measures how attitudes influence belief changes and whether attitudes themselves change through data visualization.
  • Special experimental conditions were designed to examine the effects of different types of uncertainty representation (dynamic and static confidence intervals).

• What are the implementation steps? What key techniques were used?

  • Experimental design:
    • Analyzing two topics (COVID-19 vaccination and labor union membership) and studying participants' attitude and belief changes.
    • Pre-testing participants' global attitudes, then presenting a series of visualizations and measuring their belief updates (correlation estimation) and attitude changes.
    • Using three visualization conditions: static linear regression plots (Line), hypothetical outcome plots (HOP), and static uncertainty representations (Ensemble).
  • Methods and design:
    • Representing uncertainty graphically and measuring participants' understanding of the data through visual scales for beliefs and attitudes.
    • Statistical models were used to evaluate the main drivers of belief changes, including participants' attitude strength and uncertainty representation conditions.

Research Findings

• What specific findings were achieved?

  • Belief Changes:
    • Strong prior attitudes (e.g., strong opposition to COVID-19 vaccines) were associated with smaller belief changes; participants with conflicting data were more likely to stick to their original views.
    • Uncertainty representation in data visualizations further reduced belief changes, particularly in the strong-attitude group.
  • Uncertainty Representation:
    • Uncertainty representations (e.g., hypothetical outcome plots and static displays) increased participants' subjective uncertainty about the data but did not significantly align participants' understanding of confidence intervals with the actual intervals.
  • Attitude Changes:
    • While data visualizations could change beliefs, they did not significantly impact global attitudes. Only in the labor union topic was there a slight possibility of attitude change observed.

• What advantages does it have compared to existing solutions?

  • The study explored the interaction between beliefs and attitudes and quantified the statistical relationship between attitude strength and belief updating.
  • It provided new insights into the mechanisms through which uncertainty visualizations influence beliefs and attitudes.

• What were the experimental or evaluation results?

  • Participants with strong attitudes were more resistant to belief updates when faced with conflicting data.
  • The negative impact of uncertainty representation was more pronounced in the strong-attitude group.
  • Participants' attitudes showed no systematic change, especially among those with initially strong attitudes.

• Limitations and future directions

  • Limitations:
    • Belief changes may be influenced by whether participants actively remembered the data or by the design of the experimental tasks (e.g., whether they encouraged incremental reasoning).
    • Measuring attitudes may not fully capture the complexity of human cognitive mechanisms.
  • Future directions:
    • Explore new experimental designs that combine active reasoning tasks with enhanced persuasive information.
    • Measure long-term belief changes and sustainable attitude shifts, and investigate how data visualization design interacts with moral and social cognition to achieve persuasive goals.

This study provides valuable insights into the effectiveness of data visualization in persuasive science communication and raises new challenges for mitigating cognitive resistance and enhancing attitude change.