Bot Posts
"Biden has an awful climate plan takes money from fossil fuels supports fracking maybe will do some stuff for year 2050](https://www.newsweek.com/joe-biden-second-worst-climate-change-greenpeace-1439346)"
"[Speaking of corruption, check out this article on Harris lol](https://www.humorousmathematics.com/post/kamala-harris-corrupt-former-prosecutor-selected-as-joe-biden-s-2020-running-mate)" appeared 34 times in one day.
Limitations
Due to time limitations and computer resources, we needed to use a pre-trained natural language processing model.
Bias
- Reddit bots may have skewed our data. When reviewing neutral-rated comments for Trump, we found that bots would post the same comment multiple times a day
- The demographics of reddit users do not represent the demographics of the U.S. poplation. Reddit users are: 50% Americans, 64% 18-24 years-old, 70% non-Hispanic white people
- Sentiment analysis is a machine learning algorithm, and it doesn't do well with some types of language, such as sarcasm
Conclusions
Our Reddit comments did not show any strong favor for either candidate, which ran counter to our predictions. Reddit comments were on average middling, with most recent data giving trump and biden 0.08 and 0.06 sentiment scores respectively. Separated out by positive and negative sentiment, both candidate's positive sentiment hovered around 0.15 and their negative sentiments hovered around -0.2.
In polling, the current president hovered around 50% disapproval and 30% approval. This high disapproval rating is reflected in our wordcloud to some extent with words associated with negative behavior, such as "condone", "hostilities", and "reddiquette".
Ultimately, this data may be a great tool for campaigning and adjusting campaign advertising tactics to suit the targeted populations.