Move over HAL, ANN says who's hot or not
By Michal Tellos
News Editor
August 26, 2009
 While hotornot.com may have been a lot of fun in Grade 12 computer science class, it may now be obsolete.
UWindsor student and researcher Josh Chauvin, a double-major in psychology and philosophy, has developed a computer program that judges the attractiveness of a human being.
The program utilizes an artificial neural network (ANN), which is pattern recognition software somewhat similar to the architecture of the brain.
Chauvin sought to provide evidence for the idea that assessments of attractiveness are biologically inherent, rather than objective.
ANN was trained by gathering 50 images of men and 50 images of women, and then having a subsequent 100 people rate the images on a scale of one to 10. Chauvin then provided ANN with 66 of the images along with how people rated them.
Chauvin then provided ANN with the remaining 33 images, without the population ratings, and ANN was able to construct a pattern, based on the 66 human responses, that could judge facial attractiveness.
Chauvin was inspired for the project by a philosophy about mind design where he first learned of ANNs.
While developing the project, Chauvin worked under the guidance of Marcello Guarini, a philosophy professor who helped with the computer aspects of the project, and Chris Abeare, a psychology professor who helped with the statistical portions.
While the facial attractiveness aspect of the project has generated interest and press attention, Chauvin states that this was never his long-term interest.
He notes that the project could have clinical applications.
“Child psychology definitely gave me some ideas about how the system could be applied to help diagnose certain congenital illnesses,” he said.
Specifically, Chauvin would like to target facial differences common to people with congenital diseases such as Down syndrome.
While Down syndrome is easy to identify, other illnesses, such as fetal alcohol syndrome, do not currently have genetic testing available.
Though Chauvin states that the project is still in its infancy, he notes that it can be even more interesting to students in the future.
“As the project develops however, and if the ANN can generalize attractiveness to an even larger sample of the population, I am sure it will be useful to the student population,” said Chauvin.
“If not useful, then at least a fun little gadget to rate people's attractiveness.”
|
