class: center, middle, inverse, title-slide # Understanding public opinion &<br>the dissemination of information<br>in new media environments ### Michael W. Kearney📊<br/>School of Journalism <br/>Informatics Institute <br/>University of Missouri ### <table style="border-style:none;padding-top:30px;" class=".table"> <tr> <th style="padding-right:75px!important"> <a href="https://twitter.com/kearneymw"> <i class="fab fa-twitter"></i> </a> </th> <th style="padding-left:75px!important"> <a href="https://github.com/mkearney"> <i class="fab fa-github"></i> </a> </th> </tr> <tr style="background-color:#fff"> <th style="padding-right:75px!important"> <a href="https://twitter.com/kearneymw"> <span class="citation">@kearneymw</span> </a> </th> <th style="padding-left:75px!important"> <a href="https://github.com/mkearney"> <span class="citation">@mkearney</span> </a> </th> </tr> </table> --- <!-- As new media continue to shape [and reshape] digital journalism, it is increasingly important to both identify and understand new patterns in the spread of information. With this in mind, I present my research examining user networks and the dissemination of political information on Twitter and outline the trajectory of my broader research project, which leverages computational/data-science methods to identify and evaluate automation and the spread of information in a fast-paced digital world. --> ## Roadmap + Research interests/background + Problems and opportunities in new media + Completed study: partisan networks on Twitter + In-progress study: detecting bots and fake news + The big picture --- class:inverse,middle,center # Research interests/background --- ## Political communication + Political use/effects of Facebook - Following political candidates [on Facebook] has no effect on **political participation** <cite>(Computers in Human Behavior, 2015)</cite> - **Interpersonal uses** of Facebook for politics include identity expression and responding to social expectations–not to win people over <cite>(Communication Research Reports, 2017)</cite> + Perceptions of candidates and media - Sexism continues to shape **candidate image** following 2016 election <cite>(Banwart & Kearney, 2018)</cite> - Understanding perceptions of **trust** in news sources <cite>(Trusting News, 2017)</cite> --- ## Quantitative research methods + **Using** - Chapters on research design <cite>(Handbook of Political Advertising, 2017)</cite> and statistical methods for analyzing categorical <cite>(Sage Encyclopedia, 2017)</cite> and longitudinal data - Analysis via multilevel <cite>(Hall, Kearney, & Xing, 2018; Kearney, R&R)</cite> and latent variable <cite>(Hall, Kearney, & Xing, 2018; Communication Research Reports, 2017)</cite> models + **Teaching** - Taught undergraduate and graduate classes on quantitative research methods - Served as quantitative research/statistical consultant <cite>(Center for Research Methods & Data Analysis, University of Kansas)</cite> --- class:inverse,middle,center # Problems and opportunities<br>in new media --- ## *New* problems + Concerns over media fragmentation and the rise of new media - Expansion of **media choice** <cite>(Holbert et al., 2000; Prior, 2007)</cite> and growing importance of **social validation** <cite>(Messing & Westwood, 2012)</cite> - Effects of **under-regulated user-level agenda setting** <cite>(Garrett, 2013)</cite> e.g., proliferation of misleading or fake news <cite>(Stanford History Education Group, 2016)</cite> --- ## *New* solutions? + Technological advancements introduce new possibilities - Answer new and novel questions by **capturing communication in real-time** <cite>(Handbook of Political Advertising, 2017)</cite> - Understand large and complex data sets via **data science**, e.g., mapping topic salience and political sentiment on Twitter during the 2016 election <cite>(Kearney, 2018)</cite> --- class:inverse,middle,center # Completed study:<br>partisan networks on Twitter --- ## Partisan networks + We know **a lot** about exposure to partisan news - e.g., theories of motivated reasoning and cognitive dissonance <cite>(Prior, 2013)</cite> + We know relatively **little** about partisan [user] networks - We lacked<sup>1</sup> necessary tools to examine networks and interactions on a large scale <cite>(Garrett, 2013)</cite> <br> <footnote><sup>1</sup> The reason I learned to code and created {rtweet}, an R package for collecting and analyzing Twitter data <cite>(CRAN, 2016)</cite></footnote> --- class: tight ## About {rtweet} <img src="img/rtweet-logo.png" width="120px" align="right" /> + On Comprehensive R Archive Network (CRAN) [](https://opensource.org/licenses/MIT)[](https://cran.r-project.org/package=rtweet) + Growing base of users [](http://depsy.org/package/r/rtweet) + Fairly stable [](https://travis-ci.org/mkearney/rtweet)[](https://codecov.io/gh/mkearney/rtweet?branch=master)[](https://www.tidyverse.org/lifecycle/#maturing) + Package website: [rtweet.info](http://rtweet.info) [](http://rtweet.info/) + Github repo: [mkearney/rtweet](https://github.com/mkearney/rtweet) [](https://github.com/mkearney/rtweet/)[](https://github.com/mkearney/rtweet/) --- ## Selective exposure on Twitter + Twitter's **asymmetrical connections** make it unique - Estimate political ideology via user networks <cite>(Barbera, 2015)</cite> + Study design: - Track the follow-decisions (networks) of a random sample of partisan and non-partisan Twitter users during the 2016 election --- ## Hypotheses 1. Twitter user **networks will cluster according to partisanship** - Partisan network homogeneity or 'network polarization' -- 2. Network polarization will **increase with proximity to election** - *Change* in partisan network homogeneity --- ## Method 1. Identified every follower from 12 source accounts representing 3 groups - Democrats: maddow, paulkrugman, salon, huffpo - Republicans: DRUDGE, PalinUSA, seanhannity, foxnews - Moderates: SI, AMC_TV, survivorcbs, americanidol -- 1. Randomly sampled 20,000 followers from each group -- 1. Filtered out inactive/bot-like users -- 1. Randomly sampled 1,000 users from each group --- <p style="text-align:center"> <img src="img/flowchart-labelled-1a.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-1b.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-2a.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-2b.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-3a.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-3b.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-4a.png" /> </p> --- <p style="text-align:center"> <img src="img/flowchart-labelled-4b.png" /> </p> --- ## H<sub>0</sub>: partisanship unrelated to networks <p style="text-align:center"> <img width="70%" src="img/partisancluster.sim.null.png" /> </p> --- ## H<sub>1</sub>: partisanship related to networks <p style="text-align:center"> <img width="70%" src="img/partisancluster.sim.model.png" /> </p> --- ## H1: partisanship related to networks <p style="text-align:center"> <img width="70%" src="img/partisancluster.results.png" /> </p> --- <p style="text-align:center"> <img width="90%" src="img/sourceaccounts.png" /> </p> --- <p style="text-align:center"> <img width="90%" src="img/sourceaccounts.verify.png" /> </p> --- ## H2: polarized networks -- + The number and rate of **in-group**<sup>1</sup> versus **out-group**<sup>2</sup> follow-decisions will be higher among partisan (Dem/GOP) users than non-partisan (moderate) users. - For moderates, in/out-group classification were based on empirical network leaning at time of initial data collection <footnote><sup>1</sup> e.g., Democrat user following a Democrat elite</footnote><br> <footnote><sup>2</sup> e.g., Republican user following a Democrat elite</footnote> --- <p style="text-align:center"> <img width="90%" src="img/np.unweighted.png" /> </p> --- <p style="text-align:center"> <img width="90%" src="img/np.weighted.png" /> </p> --- ## Discussion + Additional **evidence of partisan selective exposure** in follow-decisions of Twitter users + Proximity to election predicted higher in-group follow-decisions **only for partisan users** - Non-partisan users still preferred to follow in-group, but not at a significantly higher rate over time -- Biggest division in politics today? It's not Democrats and Republicans; it's partisans and non-partisans --- class:inverse,middle,center # In-progress study:<br>detecting bots and fake news --- ## Two related projects 1. **tweetbotornot**: Identifying, [machine] learning, and classifying "Twitter bots" - A powerful bot-detecting model trained on over 40k Twitter-listed bots 1. **textfeatures**: Exporting tools to streamline the *text-to-machine-learning-model* process - Easy to use R package for converting text into machine learning-friendly numeric features - Part of pipeline used to detect fake news stories --- ## tweetbotornot <img src="img/tweetbotornot.png" width="120px" align="right" /> Open-source project with novel and prolific method for generating labelled bots <p style="text-align:center"> <img width="45%" src="img/tweetbotornot-kearneymw.png" /> <img width="45%" src="img/tweetbotornot-netflix_bot.png" /> </p> --- ## textfeatures Converts textual data into 100+ numeric features per observation–great for building machine learning models <p style="text-align:center"> <img width="65%" src="img/textfeatures.png" /> </p> --- class:inverse,middle,center # The big picture --- ## Digital journalism + New potential for **spreading of false or misleading information** and **manipulating public opinion** - Especially if **trust in media** is purely a function of partisanship and online echo chambers + The defining question of our time is a communication one: - How do we ethically **communicate** in an increasingly-connected and stratified digital world? --- ## Trajectory + Working toward building a **media reliability system** - Use citation metrics, fact-checking sites, theory driven journalism dictionaries, and social media data to measure accuracy, political bias, etc. + Two goals: - Facilitate **source-checking** and **debunking of fake or misleadingnews** stories - Offer a **badge/linking interface** to a database with source-related information --- ## Collaborations + I also enjoy working on other projects where my main contribution is related to data/analysis - Two tests of social displacement through social media use <cite>(Hall, Kearney, & Xing, 2018)</cite> - A test of reinforcing spiral of influence between news use, political talk, and social media expression - Journalist interactions on Twitter - Tracking inconic photos on social media --- class:inverse,middle,center # That's it \o/ <br> # Questions?