A Gaze-Driven Evolutionary Algorithm to Study Aesthetic Evaluation of Visual Symmetry

Makin, Alexis DJ ORCID: 0000-0002-4490-7400, Bertamini, Marco ORCID: 0000-0001-8617-6864, Jones, Andrew, Holmes, Tim and Zanker, Johannes M
(2016) A Gaze-Driven Evolutionary Algorithm to Study Aesthetic Evaluation of Visual Symmetry. I-PERCEPTION, 7 (2). 2041669516637432-.

[img] Text
MakinBertaminiJonesHolmesZanker2016.pdf - Unspecified

Download (291kB)


Empirical work has shown that people like visual symmetry. We used a gaze-driven evolutionary algorithm technique to answer three questions about symmetry preference. First, do people automatically evaluate symmetry without explicit instruction? Second, is perfect symmetry the best stimulus, or do people prefer a degree of imperfection? Third, does initial preference for symmetry diminish after familiarity sets in? Stimuli were generated as phenotypes from an algorithmic genotype, with genes for symmetry (coded as deviation from a symmetrical template, deviation-symmetry, DS gene) and orientation (0° to 90°, orientation, ORI gene). An eye tracker identified phenotypes that were good at attracting and retaining the gaze of the observer. Resulting fitness scores determined the genotypes that passed to the next generation. We recorded changes to the distribution of DS and ORI genes over 20 generations. When participants looked for symmetry, there was an increase in high-symmetry genes. When participants looked for the patterns they preferred, there was a smaller increase in symmetry, indicating that people tolerated some imperfection. Conversely, there was no increase in symmetry during free viewing, and no effect of familiarity or orientation. This work demonstrates the viability of the evolutionary algorithm approach as a quantitative measure of aesthetic preference.

Item Type: Article
Uncontrolled Keywords: Symmetry, aesthetics, preference, eye tracking, evolutionary algorithm
Depositing User: Symplectic Admin
Date Deposited: 19 Apr 2016 08:40
Last Modified: 15 Dec 2022 23:47
DOI: 10.1177/2041669516637432
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3000642