Presentation on theme: "Tim Holmes & Johannes Zanker. Establishing the drivers of aesthetic preference is hard! Aesthetic phenomena are among the most complex phenomena within."— Presentation transcript:
Tim Holmes & Johannes Zanker
Establishing the drivers of aesthetic preference is hard! Aesthetic phenomena are among the most complex phenomena within the domain of psychology, and this is often used as an excuse by psychologists for being able to say so little about them.” “Aesthetic phenomena are among the most complex phenomena within the domain of psychology, and this is often used as an excuse by psychologists for being able to say so little about them.” Gustav Fechner (1865)
Reductionism destroys the “gestalt”...
... and frequently ignores context!
Historical reliance on self-report measures Not always reliable Subject to bias (Nisbett & Wilson, 1977; Wilson & Schooler, 1991) Objective measures needed MRI/EEG/Eye-Tracking (Vessel, Starr, & Rubin, 2009; de Tommaso, et al., 2008; Santella & DeCarlo, 2004)
GDEA developed during PhD Evolutionary Algorithm (Darwinian) Competition for limited resource Genetic inheritance Survival of the fittest (Holmes & Zanker 2008, Proc. GECCO) Fitness determined for the complete phenotype based on aesthetic preferential looking measure (Holmes & Zanker 2012 i-Perception) Tested on GR/Colour Shape/Grouping/Op- Art/Commercial Packaging
Never tested with “organic” phenotypes Special Effect (game) software accessibility GDEA as a creative tool? Taking it to the streets... Live Science BA Grant (yay!)
GAZE DATA FROM EYE- TRACKER PROVIDES FITNESS THIS IS USED TO IDENITFY THE FITTEST PARENTS (BEST DINOSAURS) GAZE DATA FROM EYE- TRACKER PROVIDES FITNESS THIS IS USED TO IDENITFY THE FITTEST PARENTS (BEST DINOSAURS) USE DNA FROM FITTEST PARENTS TO CREATE NEW OFFSPRING MUTATIONS INTRODUCE NOVELTY USE DNA FROM FITTEST PARENTS TO CREATE NEW OFFSPRING MUTATIONS INTRODUCE NOVELTY Evolving Population Fitter Population Initial Population
1000ms ms 250ms 4 samples per generation
Colour (R,G,B) Body (Type, Height,Width, Orientation) Number of spots Spines (Number, Height) Back Legs (Type, Height, Width) Front Legs (Type, Height, Width) Tail (Length, Thickness, Curvature) Neck (Length, Thickness, Curvature) Head (Type, Height, Width, Orientation) 1,073,234,313,200,000,000,000,000,000,000,000 dinosaurs!
We expected clustering in colours from subject matter......wide distribution suggests preferred colour was dominating rather than best colour for a dinosaur If colour preference driving preference, we might have expected lots of blue dinosaurs The most consistently preferred colour in self- report R,G,B levels in self report closer to evolved for females
Lost a lot of data: Set-up Environment Still lots more analysis to be done... Aesthetics: Age effects on colour & head/eye/body ratios Effects of presentation duration on performance Cultural differences - 43 different nationalities Eye-movements
Methodology facilitates research into complex interactions Participants liked it! 78% didn’t get tired/bored Clear potential for development as a creative tool
GDEA being applied to Mondrians (RHUL), Symmetry (Liverpool/RHUL) and commercially (Acuity) NEW Matlab/Tobii Toolbox supports full gaze contingency meaning... real time feedback on gaze interactive stimuli (animation, sound) RHUL Baby Lab testing the methodology in pre- verbal infants Future Testing with Special Effect
My team of explorers: Johannes ZankerSzonya Durant Alice Lowenhoff Hayley Thair Elina NikolaidouJade Jackson Felix ZankerGary Dalton Lee Webster My support:
Questions? Live Science Website still up... New Matlab Tobii Toolbox... Launches with SDK 3.2 in 2012 Workshop this afternoon for details