Given at Women Data Scientists DC November 16th 2016. This is a preliminary study into the images surfaced in Google Search related to the US presidential candidates.
2. Images &
Elections
• Competence inferred
from facial appearance
(Todorov et al., 2005)
• Children (5-13 yrs) use
same facial attributes to
predict election outcomes
(Antonakis, Delgas, 2009)
• Beautiful candidates win
votes (Berggren et al, 2010)
3. Gender Bias
• Stereotype exaggeration and
systematic underrepresentation
of women in search results (Kay et
al., 2015)
• Sentiment of women in news:
Happy, Calm, Submissive;
• Sentiment of Men in news:
Sad, Excited, Dominant (Referenced in
Kwak, An, 2016)
• Women are younger and smile
more (Kwak, An, 2016)
4. Where is Google directing our attention?
How did Google Image Box images represent each
candidate?
Which news sources delivering these messages did Google
promote?
5. Plan Scrape data: Request timestamp
Search query (‘donald
trump’, ‘hillary clinton’)
Image box Rank
News link
Image path
Download image to image path
Associate each image with
each dataframe row
Determine image frequency
Determine News Source
frequency
Identify unique images
Calculate sentiment for each
unique image
6. Google Scraper from here, and customised
Collection once per day, 8 weeks.
Data stored in MySQL database on AWS.
25. Donald Trump Hillary
Clinton
Average Age 66.4
(Currently 70)
52.5
(currently 69)
Average Smile
(Intensity?)
0.93 0.96
Gender prototypical: Women are younger and smile more