Ramping up Control: How Judicial Harassment and Online Smear Campaigns Work in Tandem to Limit Dissent in Thailand - Appendix
Description
This appendix contains additional information on the techniques and findings adopted in research concerning digital repression techniques observed in Thailand from between 1 Jan 2021 - 31 Dec 2021, a time marked by pro-reform protests which were met with harsh counter-reactions by state and institutional actors. This document covers information and techniques adopted in conducting social media account selection, data collection, and big data analysis comprising social network analysis, topic modeling, and directed keywords embeddings and semantic map generation for the labeling of key narratives of interest.
Files
Steps to reproduce
This research employed an actor-based approach to social media data collection, rather than a keyword-based approach. Actor-based approaches offer the advantage of enabling; 1) comprehensive latent analysis of the content produced by the accounts themselves, and 2) a more complete overview of the network context in which these accounts operate. Accounts were identified using digital ethnographic and observational techniques over an observation period from July to December 2020, coinciding with the onset and peak of anti-establishment protests in this year. During this time our team recorded observations in an excel sheet three times per week with each session lasting approximately one hour until the final characteristics of interest and inclusion criteria were solidified. Certain accounts were found to frequently engage in smear campaigns targeting protestors and possessed more than 5,000 followers, indicating that their content was likely consumed and disseminated by potentially large audiences. Iteratively following these accounts during the observation period yielded a list of 23 X (formerly Twitter) and 19 Facebook accounts consistently engaging in such behavior. All tweets and Facebook posts from January 1, 2020, to February 1, 2023 from these accounts were collected using the Twitter API (academic tier, v. 2.0) and Facebook’s data collection tool, CrowdTangle. To build out additional layers of interaction networks, four subsequent data collection rounds were conducted to include: 1) individuals who interacted with content from the originally identified accounts (1 layer of abstraction), and 2) individuals who interacted with content from the accounts found in the first layer of abstraction, and so forth. This process produced a total of 555,683 unique posts across both platforms, along with accompanying metadata of the accounts and posts in question. Initial time series analysis revealed a disproportionately large number of posts between January 1, 2021, and December 31, 2021, among the accounts of interest. Notably, this time frame corresponds with peak moments in the offline anti-establishment protests in Thailand. In light of this observation, and to examine the interplay between these offline and online phenomena, the dataset for this analysis was filtered to this timeframe, resulting in a dataset comprising 44,008 posts disseminated by the accounts of interest.