PresentationPDF Available

Neuroaesthetic of websites: evaluating the impact of expertise and exposure to website on aesthetic judgment

Authors:
Neuroaesthetic of websites:
evaluating the impact of expertise and exposure to website
on aesthetic judgment
International Association of Empirical Aesthetics congress
31 Aug. 2018
Giulio Gabrieli, Gianluca Esposito
BACKGROUND
Aesthetic experience: psychological state, attentive focus.
“Halo Effect”: are good looking websites more reliable?
Relationship between visual properties and aesthetic judgment.
Users’ expertise or exposure to web pages?
Implicit ratings?
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
RESEARCH QUESTION
AIM
Aim of this study is to investigate the role of expertise and
exposure on websites’ aesthetic judgment, both with implicit and
explicit ratings.
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
HYPOTHESIS
1. Visual elements of websites (symmetry, visual complexity,
brightness, number of images and colorfulness) are related to
aesthetic appreciation
2. Expertise and exposure play a role in aesthetic judgment of
web pages
PHYSIOLOGY & AESTHETIC
Emotion and physiology
Researchers demonstrated the relationship between individuals’
physiological activity and emotions.
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Physiological measurements:
ECG
EMG
EDA
Pupillometry
Emotions
Number of Images Visual Complexity
PHYSIOLOGY & AESTHETIC
Visual properties of web pages
From previous studies, we know that there are several effective
features that can be automatically estimated, including:
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS Colorfulness Symmetry Brightness
VISUAL COMPLEXITY
Quadratic Tree Decomposition
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Recursive X-Y partitioning
The more the number of leaves, the more complex an image is.
VISUAL COMPLEXITY
Quadratic Tree Decomposition
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Recursive X-Y partitioning
The more the number of leaves, the more complex an image is.
IMAGES AND TEXT
Space based Decomposition + Optical character recognition
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Automatic recognition of different elements matched with text
recognition
IMAGES AND TEXT
Space based Decomposition + Optical character recognition
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Automatic recognition of different elements matched with text
recognition
METHODS
Participants (N = 59) have been asked to rate 100 images of
websites and 50 emotional pictures (IAPS Dataset), while their
physiological activity was recorded (ECG, EMG, EDA,
Pupillometry).
Expertise and exposure levels have been assessed using a
questionnaire.
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
AVI14 DATASET
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
EXPLICIT RATINGS
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
IMPLICIT RATINGS
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
TRAINING
GENERATION
NEURAL NETWORK
PIP INSTALL PYSIOLOGY
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
https://github.com/Gabrock94/Pysiology
https://pysiology.readthedocs.io/
Electromyographic signals
Electrocardiographic signals
Electrodermal activity signals
Gabrieli G., Azhari A., Esposito G. (in press), PySiology: a Python Package for Physiological Feature
Extraction, Special issue of Smart Innovation, Systems and Technologies.
Gabrieli G., Azhari A., Esposito G. (2018), PySiology: a Python Package for Physiological Feature
Extraction, Italian Workshop on Neural Networks (WIRN), 13-15 June 2018, Vietri sul Mare, Italy.
IMPLICIT RATINGS
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
TRAINING
GENERATION
NEURAL NETWORK
Website properties paired with ratings
MACHINE LEARNING
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
6 GLM 6 CTree
3 implicit 3 explicit 3 implicit 3 explicit
Expertise and Exposure
None
Expertise
Exposure
Testing
5-points Likert Sc
Grouped (1-3, 4-5)
PRELIMINARY ANALYSIS
After preliminary analysis, 5 websites have been removed because
of statistically significant explicit ratings between males and
females (gender differences).
Two websites presented significantly different ratings by high
exposed and low exposed users, while three websites received
significantly different rating by experts and non experts.
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
MODELS’ ACCURACY
A bit of background here
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Symmetry : Colorfulness : Visual Complexity: brightness: n.° Images
*Avg. accuracy on 1000 tests
MODELS’ ACCURACY
A bit of background here
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Symmetry : Colorfulness : Visual Complexity: brightness: n.° Images
*Avg. accuracy on 1000 tests
MODELS’ ACCURACY
A bit of background here
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Symmetry : Colorfulness : Visual Complexity: brightness: n.° Images
*Avg. accuracy on 1000 tests
MODELS’ ACCURACY
A bit of background here
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Symmetry : Colorfulness : Visual Complexity: brightness: n.° Images
*Avg. accuracy on 1000 tests
CONCLUSION
We have been able to predict average web pages’ aesthetic
from websites' visual elements
We have found some differences due to expertise and
exposure on single pages’ ratings, but the effect is lost in
predictive models
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS FUTURE STUDIES
More differentiate users (age, expertise, exposure)
Different websites, with different visual properties
TAKE HOME MESSAGE
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
Using an implementation of this model before real user testing we
can optimize the design evaluation process, reducing both time
and costs.
Thanks
Questions?
DISCUSSION - GENDER DIFFERENCES
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
DISCUSSION - EXPOSURE DIFFERENCES
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
DISCUSSION - EXPERTISE DIFFERENCES
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
METHODS
Participants (N = 59) have been asked to rate 100 images of
websites and 50 emotional pictures, while their physiological
activity was recorded (ECG, EMG, EDA, Pupillometry).
Expertise and exposure levels have been assessed using a
questionnaire.
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
ESTIMATION OF IMPLICIT RATINGS
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
PIP INSTALL PRETTYWEBSITE
https://github.com/Gabrock94/PrettyWebsite
https://prettywebsite.readthedocs.io
BACKGROUND
AIM & HYPOTHESIS
PHYSIOLOGY &
AESTHETIC
METHODS
RESULTS
CONCLUSIONS
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.