TY - GEN
T1 - Cross-Domain Classification of Facial Appearance of Leaders
AU - Yoon, Jeewoo
AU - Joo, Jungseock
AU - Park, Eunil
AU - Han, Jinyoung
N1 - Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020
Y1 - 2020
N2 - People often rely on visual appearance of leaders when evaluating their traits and qualifications. Prior research has demonstrated various effects of thin-slicing inference based on facial appearance in specific events such as elections. By using a machine learning approach, we examine whether the pattern of face-based leadership inference differs in different domains or some facial features are universally preferred across domains. To test the hypothesis, we choose four different domains (business, military, politics, and sports) and analyze facial images of 272 CEOs, 144 4-star generals of U.S. army, 276 U.S. politicians, and 81 head coaches of professional sports teams. By extracting and analyzing facial features, we reveal that facial appearances of leaders are statistically different across the different leadership domains. Based on the identified facial attribute features, we develop a model that can classify the leadership domain, which achieves a high accuracy. The method and model in this paper provide useful resources toward scalable and computational analyses for the studies in social perception.
AB - People often rely on visual appearance of leaders when evaluating their traits and qualifications. Prior research has demonstrated various effects of thin-slicing inference based on facial appearance in specific events such as elections. By using a machine learning approach, we examine whether the pattern of face-based leadership inference differs in different domains or some facial features are universally preferred across domains. To test the hypothesis, we choose four different domains (business, military, politics, and sports) and analyze facial images of 272 CEOs, 144 4-star generals of U.S. army, 276 U.S. politicians, and 81 head coaches of professional sports teams. By extracting and analyzing facial features, we reveal that facial appearances of leaders are statistically different across the different leadership domains. Based on the identified facial attribute features, we develop a model that can classify the leadership domain, which achieves a high accuracy. The method and model in this paper provide useful resources toward scalable and computational analyses for the studies in social perception.
KW - Face
KW - Facial display
KW - Leadership
KW - Machine learning
UR - https://www.scopus.com/pages/publications/85093094426
U2 - 10.1007/978-3-030-60975-7_32
DO - 10.1007/978-3-030-60975-7_32
M3 - Conference contribution
AN - SCOPUS:85093094426
SN - 9783030609740
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 440
EP - 446
BT - Social Informatics - 12th International Conference, SocInfo 2020, Proceedings
A2 - Aref, Samin
A2 - Bontcheva, Kalina
A2 - Braghieri, Marco
A2 - Dignum, Frank
A2 - Giannotti, Fosca
A2 - Grisolia, Francesco
A2 - Pedreschi, Dino
PB - Springer Science and Business Media Deutschland GmbH
T2 - 12th International Conference on Social Informatics, SocInfo 2020
Y2 - 6 October 2020 through 9 October 2020
ER -