Trusted, Automoderated users
17,538
edits
Physiognomy: Difference between revisions
Jump to navigation
Jump to search
→Empirical study
| Line 12: | Line 12: | ||
* Wang & Kosinski (2017) used a deep neural network that, analyzing 35,326 'selfie' images, correctly determined homosexuality in 81% of cases for men, and in 74% of cases for women.<ref>https://osf.io/zn79k/</ref> This was compared to human judges, who could distinguish a man's homosexuality in 61% of cases and women's in 54% of cases (slightly above chance). This study has been heavily criticized, however, for being confounded by differences in facial expression, grooming, clothing, camera angle and other contextual factors unrelated to facial structure. | * Wang & Kosinski (2017) used a deep neural network that, analyzing 35,326 'selfie' images, correctly determined homosexuality in 81% of cases for men, and in 74% of cases for women.<ref>https://osf.io/zn79k/</ref> This was compared to human judges, who could distinguish a man's homosexuality in 61% of cases and women's in 54% of cases (slightly above chance). This study has been heavily criticized, however, for being confounded by differences in facial expression, grooming, clothing, camera angle and other contextual factors unrelated to facial structure. | ||
* Holtzman (2011) created a series of prototypical faces corresponding to each of the traits of the [[dark triad]], using the photos of 81 study participants, who completed self-report inventories designed to measure the levels of the dark triad traits. The participants were also evaluated in regards to their level of dark triad traits by their peers. It was found that observers could (above chance) correctly distinguish between high and low morphs of the various "dark traits", thus lending some evidence to the idea that these traits are correlated with a certain facial structure. This correlation was explained by several hypothesis, the facial traits and the dark triad being co-evolved, the facial traits influencing people's self perception and thus behavior, or that individuals are possibly conditioned to behave in a way 'congruent' with their facial structure by peers, through constant social reinforcement.<ref>https://www.researchgate.net/publication/232381817_Facing_a_psychopath_Detecting_the_Dark_Triad_from_emotionally-neutral_faces_using_prototypes_from_the_Personality_Faceaurus#pf6</ref> | * Holtzman (2011) created a series of prototypical faces corresponding to each of the traits of the [[dark triad]], using the photos of 81 study participants, who completed self-report inventories designed to measure the levels of the dark triad traits. The participants were also evaluated in regards to their level of dark triad traits by their peers. It was found that observers could (above chance) correctly distinguish between high and low morphs of the various "dark traits", thus lending some evidence to the idea that these traits are correlated with a certain facial structure. This correlation was explained by several hypothesis, the facial traits and the dark triad being co-evolved, the facial traits influencing people's self perception and thus behavior, or that individuals are possibly conditioned to behave in a way 'congruent' with their facial structure by peers, through constant social reinforcement.<ref>https://www.researchgate.net/publication/232381817_Facing_a_psychopath_Detecting_the_Dark_Triad_from_emotionally-neutral_faces_using_prototypes_from_the_Personality_Faceaurus#pf6</ref> | ||
* A Chinese study claimed to be able to tell whether someone is a criminal based on machine learning, but the technique turned out to detect smiling instead.<ref>https://twitter.com/davidjayharris/status/1103636069180993537</ref> | |||
==See Also== | ==See Also== | ||