New research examines the connection between faces and political ideology, specifically between conservatives and liberals, finding: “Most clearly we see that both male and female right-wing composites appeared happier than their left-wing counterparts. The male composites were not obviously differentiated based on masculinity. The right-wing faces, particularly for females, might be perceived as more attractive.” Moreover, according to the study’s authors, “we did observe that among women contempt was non-trivially correlated with a predicted probability of left-wing ideology.”
WATCH: Hunter Biden Pleads GUILTY, Plus James O’Keefe EXPOSES BlackRock in New Undercover Video
The study using a large sample of 3233 Danish local political candidates. Various methods were applied, including convolutional neural networks, heat maps, facial expression analysis, and assessment of physical attributes like masculinity and beauty.
The study goes beyond previous research by identifying specific facial characteristics linked to political ideology and by examining limitations in existing methodologies which result in overestimations of the association between faces and psychological traits.
Historically, research has indicated that voters consider their preferred candidates more attractive, creating an electoral advantage. Prior literature generally considered facial impressions subjective, but recent studies showed considerable accuracy in predicting ideology using algorithmic facial classifications.
Heat-mapping was utilized to determine which facial segments inform the ideology classifications provided by deep learning algorithms. Besides highlighting potential areas for methodological refinement, heat-mapping also indicated which facial features were most informative in predicting political leaning. The study posed the question of why faces can inform about ideology, and what neural networks are detecting when correctly identifying ideology-linked faces.
Various explanations were explored, including facial morphology, with right-leaning politicians found to be more attractive, and masculine features being associated with conservative economic preferences. Facial expressions were also considered relevant, with left-leaning individuals showing more surprise and less disgust, according to previous research.
Attractiveness scores in females positively correlated with a model-predicted conservative ideology. Furthermore, happy facial expressions were associated with conservatism for both genders, with males’ right-wing predicted probability correlating 0.17 with happiness and -0.17 with neutrality, while females’ were 0.15 and -0.15 respectively.
Other facial expressions were infrequent in the sample. Among women, contempt was moderately correlated with a predicted probability of left-wing ideology. Other facial expression-ideology correlations were not statistically significant for either gender.
This study confirmed the privacy threat posed by deep learning approaches, with an around 60% successful prediction rate of political ideology. It also revealed that current methods that include non-facial information in the small boxes around a face could inflate prediction accuracies.
The findings also showed that model-predicted ideology correlates with independently classifiable facial features. Among females, high attractiveness scores were noted in those predicted to be conservative, aligning with previous research. Expressing happiness, which has been linked to extraversion in past studies, was also associated with conservatism in both genders. Self-presentation may also play a role, as left and right-leaning politicians might have different motivations for smiling.
