A key limitation of the study is that Facebook itself is practically a black box. Facebook’s algorithms personalize the delivery of content on people’s newsfeeds, making it difficult to see what people actually see.
Compared to Twitter, which has a transparent API for searching content using keywords, Facebook does not have a very robust search API.
For these reasons, this research project does not purport to represent all content published on Philippine content channels on Facebook.
Rather, it used data collected through two separate systems: (1) the Sharktank, a tool developed by Rappler to monitor public pages, groups, and posts on Facebook, and (2) CrowdTangle, a social media discovery tool owned by Facebook and which allows newsrooms to monitor pages and groups from Facebook, Twitter, and Instagram.
Initial data captured by the Sharktank came from Facebook groups connected to 26 fake accounts that were investigated as part of Rappler’s propaganda series. This investigation suggested that one account alone which was determined to be fake was connected (as of October 6, 2016) to about 2.9 million members of various overseas Filipino groups associated with Ferdinand “Bongbong” Marcos Jr, and other hobby groups.1
The Sharktank’s scope of monitoring grew as the tool discovered content from other pages and groups whose posts were shared within these original groups. Newly-discovered pages and groups were added when they were determined to be posting politically-relevant content and when they were found to be posting hoaxes. Some pages and groups were added because these were either submitted to Rappler’s fact check team or discovered via Facebook’s fact check tool.
Once a page was detected and determined to be relevant to the study, the Sharktank then used scrapers to capture all publicly available posts published within these groups and pages for as far back in time as possible. This became the basis of the graphs on this study which estimated volume of content produced by these groups and pages every day. This also makes the Sharktank a useful tool for viewing the historical background of a page, website, or group.
As of writing, the Sharktank now monitors content on over 16,000 public pages and around a thousand public groups. This number continues to grow every day. Within these pages and groups, it is able to track as many as 10,000 unique hyperlinks every day or around a million links every 100 days. These include links from websites operated by newsgroups, government, and the academe. But a significant number of these websites are managed anonymously and have no known authors. In some cases, bylines indicated are not traceable to real people.
Complementary data generated through the fact-checking efforts of Rappler and other Facebook third party fact-check partners in the Philippines indicate that at least 500 of these websites could be hoax sites. Domain registration lookups indicate that many of these sites were created after the Duterte administration took over.
Comparing domain registration data with the history of posts captured by the Sharktank that included content from the domain shows that many of these appear to be “disposable sites.” They were active only for a very short period of time.
As the country prepared for the 2019 elections, Rappler expanded the scope of monitoring to include pages in the name of candidates and meme pages which were observed to have, from time to time, shared political content.
To complement and countercheck data captured by the Sharktank, this study also made use of data captured via CrowdTangle, a dashboard that facilitates the monitoring of the interactions inside Facebook pages and groups.
By exploring CrowdTangle, Rappler discovered additional pages and groups created on Facebook which bear the name of the senatorial bets not directly connected with networks already monitored by the Sharktank. Afterwards, data on content and engagement from these pages was collected and grouped into four main categories: official Facebook pages, support pages, support groups, and hate/troll pages/groups.
A separate list of pages and groups of each candidate was created in the tool to determine the total number of accounts they have. However, closed groups and pages whose links have been removed are not included in the list.
The total number of posts within each page and group was captured through CrowdTangle’s Leaderboard section, which summarizes engagement and content statistics for a given list of groups and pages. Also captured through this process were accumulated engagements per candidate as well as total followers and members of these groups and pages.
Below are links the to the other parts of this series:
- Social media, disinformation and the 2019 Philippine elections
- Social media, disinformation and the 2019 Philippine elections: Overview of Summary and Findings
- Social media, disinformation and the 2019 Philippine elections: Background
- 2019 Election Findings: How Was Social Media Used by Senatorial Candidates?
- Social media, disinformation and the 2019 Philippine elections: Conclusions and Recommendations
1Chay F. Hofileña, “Fake accounts, manufactured reality on social media,” Rappler, October 9, 2016, https://www.rappler.com/newsbreak/investigative/148347-fake-accounts-manufactured-reality-social-media