Fractal Analysis of Interface Structure

Description

Measures the fractal dimension and self-similarity of layouts and information hierarchies, characterising complexity consistently across scales.

Fractal geometry describes structures that show similar patterns at different scales, quantified by a fractal dimension. Box-counting estimates this by covering an image or structure with grids of decreasing size and measuring how the number of occupied boxes grows. In UX the method has been applied in two ways. Visually, the fractal dimension of screen layouts and images correlates with perceived complexity and aesthetic preference, with mid-range values often preferred. Structurally, information hierarchies can be examined for self-similarity: whether each level follows the same organising pattern, which supports learnability. The approach is rigorous when measurement is standardised (consistent resolution, thresholding and scale range), when dimension estimates are reported with fit quality, and when results are related to user judgements or performance rather than treated as aesthetic verdicts.

Attributes

Loss type
Latent
Stage
Evaluation
Effort
Weeks

Steps

  1. Select artefacts to analyse: screen captures at fixed resolution or hierarchy trees of the IA.
  2. Pre-process consistently, for example converting screens to edge maps with a fixed threshold.
  3. Apply box-counting across at least 5 scales and fit log(N) against log(1/size) to estimate fractal dimension.
  4. For hierarchies, compare branching patterns across levels to assess self-similarity.
  5. Collect complexity or preference ratings from 20-50 users for a subset to calibrate interpretation.
  6. Use calibrated dimension values to compare designs and flag screens outside the preferred range.

Sources

  • Unsupervised Learning in Time and Space
  • The Fractal Geometry of Nature (Benoît Mandelbrot)

Also known as

  • Fractal Analysis of UX
  • Fractal dimension analysis
  • Multi-scale complexity analysis

Categories

  • Information Theory, Physics & Mathematics