Understanding Bias in Perceiving Dimensionality Reduction Projections

2025-07-28

AI Summary

This study identifies a pervasive “visual appeal bias” in dimensionality reduction (DR) projection selection: practitioners consistently prefer aesthetically pleasing, visually salient projections—even when they poorly preserve underlying data structure—over more structurally faithful alternatives. Method: Through controlled user experiments (varying color encoding, exposure duration, and other visual parameters) and rigorous behavioral data analysis, we formally define and empirically validate this cognitive bias for the first time, identifying color labeling and brief exposure as key amplifying factors. Contribution/Results: We propose actionable mitigation strategies—including fidelity-aware visual cues and adjusted interaction timing—to counteract the bias. Our findings provide both theoretical grounding and practical guidance for visualization design, DR tool evaluation, and human-in-the-loop analytical workflows, advancing evidence-based practices in visual analytics.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Dimensionality Reduction/Feature Selection
• Data Mining & Knowledge Management: Data Visualization & Summarization

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Web data visualization
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Selecting the dimensionality reduction technique that faithfully represents the structure is essential for reliable visual communication and analytics. In reality, however, practitioners favor projections for other attractions, such as aesthetics and visual saliency, over the projection's structural faithfulness, a bias we define as visual interestingness. In this research, we conduct a user study that (1) verifies the existence of such bias and (2) explains why the bias exists. Our study suggests that visual interestingness biases practitioners' preferences when selecting projections for analysis, and this bias intensifies with color-encoded labels and shorter exposure time. Based on our findings, we discuss strategies to mitigate bias in perceiving and interpreting DR projections.

Problem

• Identifies bias favoring aesthetics over structural faithfulness in dimensionality reduction
• Explains why visual interestingness biases projection selection for analysis
• Proposes strategies to mitigate bias in interpreting dimensionality reduction projections

Innovation

• User study verifies visual interestingness bias
• Bias intensifies with color labels
• Strategies to mitigate dimensionality reduction bias