A Measurement Model for Visual Complexity in HCI: Focusing on Visual Elements in Mobile GUI Design

Ruiqi Zhang, Oosung SonElectronics 2025 · DOI

TL;DR

This paper introduces a novel measurement model for visual complexity in mobile GUI design, addressing a gap in existing research that primarily focused on web interfaces. By quantifying the influence of individual visual elements like icons, text, and images, the model offers designers a more intuitive and data-driven approach to optimize user experience and accelerate design iterations.

What Problem Does This Paper Address?

In human-computer interaction (HCI), visual complexity significantly impacts user experience, influencing first impressions and cognitive load. While research on visual complexity has grown, most studies have concentrated on web interfaces. Mobile interfaces, with their limited screen real estate and touch-based interactions, present unique challenges. Existing models often lack granular insights into how specific visual elements contribute to overall complexity, hindering designers' ability to make targeted improvements. This paper aims to fill this gap by developing a quantitative model specifically for mobile GUI design, focusing on the independent contributions of various visual elements.

Methods

The core of this research is the development of a multiple linear regression (MLR) model to quantify visual complexity in mobile GUIs. The authors identified four key metrics as independent variables:

  • Number of icons (X1)
  • Number of characters (X2)
  • Number of images (X3)
  • RGB sum entropy (X4), which quantifies color diversity and distribution.

The dependent variable is visual complexity (γ). The model aims to determine the influence weights (regression coefficients) of these independent variables on visual complexity. To ensure the reliability of their data collection, the researchers employed a two-pronged approach:

  1. User subjective ratings: Participants rated interface screenshots on a 7-point Likert scale.
  2. Objective measurement: Compressed file size of the interface screenshots was used as a proxy for visual complexity.

A correlation analysis was performed to validate the relationship between these two methods, demonstrating that compressed file size is a reliable predictor of visual complexity. The MLR model was then developed using data from 200 mobile interface screenshots.

Datasets

The study utilized a dataset of 200 mobile phone interface screenshots. These screenshots were sourced from 100 top applications in both the English and Chinese markets to ensure a diverse and representative sample of commonly used interfaces. The selection aimed to cover a range of visual complexity levels (low, medium, high) and included interfaces with at least three visual design elements to avoid overly simplistic designs.

Key Insights and Ideas

  • Mobile-specific complexity matters: The unique constraints of mobile interfaces necessitate a distinct approach to measuring visual complexity compared to web interfaces.
  • Color and distribution are key: The RGB sum entropy, representing color diversity and distribution, emerged as the most significant predictor of visual complexity, highlighting the importance of these aspects in mobile GUI design.
  • Icons' impact is nuanced: While icons are crucial visual elements, their sheer number alone did not show a statistically significant impact on visual complexity in this study. This suggests that other factors, like icon layout or internal design, might play a more significant role, or that the number of icons in mature mobile interfaces is often within a range that doesn't drastically alter complexity.
  • Quantitative metrics empower designers: Providing designers with intuitive, quantitative metrics can significantly enhance design efficiency, reduce the need for extensive user testing, and ultimately improve the user experience.

Key Results

The developed MLR model achieved an R-squared value of 0.647, indicating a good fit for the data. The key findings regarding the influence of visual elements on visual complexity are:

  • Number of icons (X1): Showed a statistically insignificant effect on visual complexity (p > 0.05).
  • Number of characters (X2): Had a statistically significant effect (p < 0.05).
  • Number of images (X3): Had a statistically significant effect (p < 0.05).
  • RGB sum entropy (X4): Demonstrated a highly significant effect (p < 0.001), with a regression coefficient of 0.278, making it the strongest predictor.

The final regression formula derived is: γ = 0.014X1 + 0.003X2 + 0.071X3 + 0.278X4 – 0.090. This formula quantifies visual complexity based on the identified metrics.

Limitations and Future Work

The authors acknowledge several limitations and suggest avenues for future research:

  • Subjective user differences: The model does not fully capture the subjective differences in how individual users perceive complexity.
  • Sample universality: While efforts were made to create a diverse dataset, it may not cover every possible mobile interface scenario, limiting the universality of the findings.
  • Future research directions:
    • In-depth icon analysis: Further research could explore the distribution characteristics within icons and their arrangement on the interface, rather than just the count.
  • Additional metrics: Incorporating metrics related to interface interactions could provide a more comprehensive understanding of visual complexity.
  • Automated tools: Combining this model with AI for metric extraction could lead to the development of automated visual-complexity measurement tools.
  • More advanced models: Exploring more sophisticated models beyond linear regression could potentially capture more complex relationships.

Summary updated

Abstract

Visual complexity has significant research value across various fields, influencing both first impressions and user experiences. In the context of human–computer interactions, studies on visual complexity in user interface design have evolved progressively alongside technological advancements, and the quantification of visual complexity has been demonstrated to be significant. However, existing research has primarily focused on web interfaces. In contrast, mobile interfaces have less maneuverable space and rely on finger interactions, necessitating more targeted studies. Additionally, while model research has gradually emerged in recent years, there are still issues regarding the influence weight of each visual element in the mobile interface that need to be addressed. This study employed a multiple linear regression analysis to develop a measurement model based on the visual elements used in mobile GUI design, addressing the weighting of each metric from the designer’s perspective. Different from other studies, this study focused on the independent visual elements of the mobile phone interface. This measurement model offers designers intuitive and quantitative reference metrics, enhancing the efficiency of design iterations. Furthermore, it lays the groundwork for the development of visual-complexity measurement tools.
BibTeX
@inproceedings{zhang2025measurement,
  author = {Ruiqi Zhang and Oosung Son},
  title = {{A Measurement Model for Visual Complexity in HCI: Focusing on Visual Elements in Mobile GUI Design}},
  booktitle = {Electronics},
  year = {2025},
  doi = {10.3390/electronics14050942},
}
Figure