IntroductionThe overarching goal of this research is to build a series of models that predict how an online community shares, reacts to, and interprets the news, using a dataset harvested from the social networking site Reddit.  https://www.reddit.com I used Reddit because it has a publicly available API, and clearly structured data. On reddit, users form into communities called subreddits- these online communities were the focus of my study research.I divided this document into three sections based upon methods and technologies.Section 1 outlines how I  used web APIs to collect data and scripting languages to sterilize and model the data. Section 2 outlines my experiments with statistical natural language processing models (Statistical NLP) specifically sentiment analysis and latent dirchilet allocations.Section 3 outlines my neural language processing models (Neural NLP), including recurrent neural networks and deep learning.Throughout this body of work, I developed skills that prepared me to undertake broader questions, beginning with simple sentiment analysis (tying words to connotation scores) and progressing to using a mixture of deep and recurrent neural networks to tackle more open-ended questions.I conclude that Sentiment Analysis, Topic Modelling, and Word Embedding models do not have the representational capacity to model how people will interpret the news. On the other hand, recurrent neural networks and deep learning have representational capacity but require large datasets. A promising research direction for the future would be optimizing deep learning models for smaller data sets. My future interests in this body of work aim to inquire about the volume of data necessary to ensure the fidelity of the results. Harvesting the DataHarvesting data from Reddit using python is straightforward with their well-documented API, called PRAW:[https://praw.readthedocs.io/en/stable/getting\_started/quick\_start.html]. For documentation on how API calls work in general, you may consult the following: [https://tray.io/blog/how-do-apis-work].

My research looks at the struggle between greater simplicity and greater complexity online conversations. I lay out two different feedback loops that shape online conversations, the Simplicity loop and the Complexity Loop. By online conversations, I am referring to the process by which people come together and decide what means what. Are GMOs good or bad? Is the president a sinner or a rogue? Was the film genius or mad? The first feedback loop, which I will call the Simplicity Loop has three parts that generate each other. 1) A simple conversation becomes 2)popular and 3) generates a consensus which leads to greater simplification. The second loop, which I will call the Complexity Loop also consists of three parts. 1) An intricate and complex conversation 2) breaks into diverging positions 3) generating new ideas and positions. The Complexity Loop generates and recombines ideas, but does not make them popular. I then lay out how to study the two loops using Yule I measure. The measure is a statistical index of linguistic complexity and can be used to study the degree to which online conversations are becoming more simple or complex. I then illustrated the utility of the theory and used the Yule I measure to study conversations on Reddit, an online messaging board.