Automated Goal Classification and Analysis of Facebook Political Advertisements

Natalie Appel
Natalie Appel

Natalie is a rising junior (’23) from Westchester, NY. She graduated from Scarsdale High School in 2019 and is now pursuing a degree in Economics with a minor in Data Analysis. She has worked with the Wesleyan Media Project’s Delta Lab since July, 2020 studying the influence of political advertisements on elections. Outside of work she loves playing on Wesleyan’s club volleyball team, working as an Orientation Leader, and watching nature documentaries. After graduating from Wesleyan she hopes to work as a data scientist.

Lexie Silverman
Lexie Silverman

Lexie is a rising junior (‘23) studying Economics and Computer Science. She is from Ventura, CA and graduated from The Thacher School in 2018. Before coming to Wes, she took a gap year during which she spent time studying abroad in Granada, Spain and interning in Silicon Valley. Her experience at Wesleyan, especially with the WMP Delta Lab this summer, has cemented her interest in applied cognitive science and behavioral studies, which she hopes to implement in her future academic and professional careers. On campus she works as a tutor for the SCIC, plays on the women’s club soccer team, and enjoys picnicking on College Row. 

Abstract: Over the last decade, the scale and influence of political digital advertising has increased dramatically. Campaigns are taking advantage of the ever-expanding audience that social media platforms have grown to offer and utilize online advertisements as an important component of their election strategy. There is an overwhelming amount of data available about these ads, especially from large platforms like Google and Facebook, but the sheer volume of ads makes it difficult to analyze. The Wesleyan Media Project has a team of hand coders who have reviewed 3,000 Facebook ads from the 2020 U.S. Election and classified them based on a number of variables, including the perceived goal of the ad. This information is extremely useful for our analysis of the sentiment and content of these ads, but hand coding is a time intensive process that cannot keep up with the scale of digital advertising. Our goal is to automate part of this process by training a machine learning classification model on the information we have from the hand labeled data set. In doing so, we can utilize computational analysis to gain insight into campaigns’ advertising strategies on social media platforms. 

FINAL-QAC-apprenticeship-poster

Live Poster Session: https://wesleyan.zoom.us/j/96016348679?pwd=dWVkbVRYbFpiNHYzREtpdjB1WjduZz09
Thursday, July 29th 1:45-2:45pm EDT