Abstract
Complex images represent a high-dimensional challenge when trying to understand which aspects of their appearance are important when making aesthetic decisions.
We developed a method to efficiently map the relative importance of many parameters simultaneously via an evolutionary selection of parameter values. Using this method we aimed to establish the features contributing to appealing (experiment 1) or aversive (experiment 2) aesthetic experience. The experiment used complex, computer-generated, flower-like stimuli defined by more than 20 attributes that included colour, symmetry, angularity and number of petals. Participants were shown a selection of six stimuli and were asked to identify both the most and the least appealing (exp 1), threatening (exp 2a) or disgusting (exp 2b) stimuli on each trial. The first stimulus set was generated by selecting, at random, a value for each of the available parameters for each stimulus. After each set of responses the presentation likelihood of the attributes of the most/least appealing stimuli were increased/decreased, respectively. The next set of six stimuli were generated via random value selection from parameters with the newly-adjusted display likelihoods. Participants completed 80 trials comprising two interleaved processes.
We established the stability of each participant’s selection through the correlation of the display probability between the two processes and assessed between-participant agreement via bootstrapping. Symmetry, self-similarity and luminance contrasts with low spatial frequency were identified as components of appealing stimuli. Blotchy and brown petals were selected as disgusting while complexity, sharpness and symmetry were selected as threatening. These results accord with previous research into the visual characteristics of aesthetic experience. We thus show that our method can quickly and accurately ascertain the important aspects of complex visual stimuli in a variety of perceptual decisions.