Effects of Alternative Scatterplot Designs on Belief

Research Background and Problem

• What problems or challenges did the authors identify?
  People tend to underestimate the correlation in positively correlated data within scatterplots, a bias that is widespread in typical scatterplot designs. Although methods such as adjusting the size and transparency of points in scatterplots have been proposed to correct this bias, existing studies primarily focus on the perceptual level and have not explored whether these methods can influence higher-level cognition, such as belief change.

• Why is this problem important?
  Scatterplots are one of the most commonly used tools in data visualization, and accurate interpretation of data correlations is crucial for scientific communication, decision support, and even social policy-making. If scatterplot designs can be optimized to influence people's beliefs, it could significantly enhance the effectiveness and societal impact of visualizations. At the same time, the potential risks of misuse (e.g., manipulative graphics) also warrant in-depth investigation.

• Research Motivation and Related Work
  Existing research shows that the shape of point clouds in scatterplots affects the perception of correlation, but it remains unclear whether these perceptual improvements can extend to higher-level cognition. Researchers have also noted that data visualization has a dual-edged impact on audiences at both emotional and cognitive levels. New designs may help correct biases but could also mislead. Therefore, this study focuses on exploring whether optimized scatterplots can influence belief change at a cognitive level.

Solution

• What methods or solutions did the authors propose?
  The authors proposed a scatterplot design that adjusts point size and transparency to reduce the visual salience of points outside the regression line, thereby emphasizing the center of the point cloud and optimizing the representation of positive correlation. They also tested the specific effects of these designs on participants' belief changes compared to traditional scatterplots.

• What is innovative about this solution?
  This design approach extends previous research on correlation perception by applying perceptual optimization mechanisms to the study of belief change. This is the first time the profound impact of scatterplot design on cognition and attitudes has been incorporated into a research framework.

• What are the implementation steps and key techniques used?

  1. Pre-experiment: Screened emotionally neutral statements with weak correlations from a large pool of variable relationship statements based on belief and emotional neutrality ratings.
  2. Experimental Design:
     • Control group: Presented with traditional standard scatterplots.
     • Experimental group: Presented with optimized scatterplots with adjusted point size and transparency.
  3. Data Presentation: Presented data in the form of scatterplots attributed to the British NHS to enhance the realism and credibility of the results.
  4. Belief Assessment: Evaluated participants' belief changes regarding the strength of variable relationships before and after visualization. Collected additional data on participants' graph literacy, confidence in belief defense, and emotional bias ratings toward the statements.

Research Findings

• What specific findings were obtained?

  1. Compared to traditional designs, the optimized scatterplots significantly enhanced participants' belief changes regarding the correlation between variables.
  2. Participants with lower graph literacy were more influenced by the optimized scatterplots, while belief changes in those with higher literacy were diminished.
  3. Participants with higher confidence in belief defense were more susceptible to visual presentation, leading to belief changes.

• What advantages does it have over existing solutions?
  The optimized design not only corrected the underestimation of correlation at the perceptual level but also significantly intervened in participants' beliefs. This demonstrates that more pronounced adjustments to point clouds in visual design can directly influence higher-level cognition without altering other dimensions of the data.

• What were the experimental or evaluation results?

  • Results showed that the optimized design, building on perceptual improvements, further enhanced cognitive and belief-level impacts. The intensity of belief changes in the optimized design group was significantly higher than in the traditional group, with a medium effect size (Cohen’s d = 0.40).
  • Regarding the interaction between graph literacy and confidence in belief defense, the optimized design effectively reduced the interference of confidence or literacy on belief changes.

• Limitations and Future Directions

  1. Limitations:
     • The study tested only a limited number of statement samples and did not examine strong beliefs or statements with different emotional polarities.
     • The use of a simple 7-point Likert scale may not capture more nuanced belief change data.
     • The diversity of scatterplot adjustment parameters was not further explored.
  2. Future Directions:
     • Investigate the effects of statements with other beliefs and emotional polarities.
     • Study the potential impact of optimized designs on behavior change.
     • Validate the differentiated effects of various optimization strategies (e.g., adjusting only transparency or size) on beliefs.
     • Consider the influence of social and educational backgrounds on design cognition preferences.

This study is the first to validate that simple point adjustment strategies in scatterplots can influence higher-level belief cognition. It provides important theoretical support for designing interactive data graphics that balance scientific rigor and social acceptability, while also pointing to research directions for identifying and addressing potentially misleading data visualization designs.