Focussing on the 2019-2020 anti-extradition law protests in Hong Kong, this is the first media investigation in Malaysia to map how foreign online disinformation campaigns made its way to Malaysian shores and spread within local online communities.
In this eight-month-long investigation, we were able to identify disinformation surrounding the protests that have reached Malaysian audiences. We traced the origins of false information on the protests, followed their distribution paths and uncovered the main players that ‘imported’ them into Malaysian cyberspace. China’s state media, local Chinese-language media, local community-based Facebook pages and the Chinese embassy in Malaysia were among the distributors.
As Southeast Asia becomes a hotbed of strategic rivalry between China and the United States, pro- and anti-Beijing propaganda targeting Chinese diaspora in Malaysia has intensified over the last few years causing worsening polarization among Malaysians especially Chinese-speaking communities. It is a well known phenomena in Malaysia but nobody has attempted to quantify and analyze it. This investigation is a local showcase of how investigation of mis- and disinformation can be done.
In this country where Beijing exerts enormous influence both financially and politically, most politicians, media outlets, think tanks and academics avoid making public censure or criticism against China or Chinese Communist Party (CCP). However, the investigation, published in both English and Chinese, received 3,128 interactions on Facebook including 941 shares according to data from Crowdtangle, a rare achievement for a long-form and heavy story critical of CCP and its affiliates.
After it was published, the project lead (Kuek Ser Kuang Keng) was invited to present the investigation findings at various journalism conferences including the Trusted Media Summit 2021 (recording https://vimeo.com/641757048) by Google News Initiatives (sharing the same panel with ProPublica, AFP, Digital Forensic Research Lab, EU Disinfo Lab and Australian Strategic Policy Institute), APAC Information Operation Roundtable by DoubleThink Lab, and an upcoming forum on disinformation by Asia Fact-checkers’ Network.
To understand how such disinformation infiltrated the Malaysian online public sphere, who were the main creators and ‘importers’ of such information, and to gauge its reach, we collected 11,482 posts published on 384 public Malaysian Facebook pages in three major languages spoken in Malaysia (English, Bahasa Malaysia and Chinese) between June 2019 (when the protests began to peak) and March 2020 that contained keywords related to the protests using social media listening tool Zanroo. The dataset includes details of each post including the number of engagements, date and time of posting, links and media shared, as well as the number of followers of each page.
Our team then manually inspected posts in the dataset that had received a minimum of 20 engagements. We scrutinised each post for misinformation or other relevant information. For posts that contain false and misleading information, we traced their origins and distribution paths through different fact-checking techniques and social media tracking analytics tools like Crowdtangle, Google search and Baidu search. We also used spreadsheets to identify Facebook pages that functioned as a network by sharing the same posts to amplify disinformation. We used Perma.cc (https://perma.cc/), an online archival tool developed by the Harvard Law School Library, to archive the Facebook posts containing misinformation.
As this is the first such investigation in Malaysia, we also made our dataset public on Github (https://github.com/kininewslab/Hk_misinfo) to allow other journalists and researchers to use it.
What was the hardest part of this project?
As this is a first-of-its-kind investigation in both Malaysia and our newsroom Malaysiakini, the team started the project by researching various methodologies used by other journalists in online information investigation. To design our methodology, the 5-person team compiled and analyzed major stories on the topic, spoke to journalists and experts in this field, went through dozens of tutorials of online investigation tools and techniques, and tested several of those tools. We invited the Crowdtangle team to give us training on how to use it for misinformation investigation. It is not just a journalistic investigation but a capacity building project.
The most laborious process in the investigation is to identify false and misleading information in the posts that we have collected. Misinformation investigations usually start with specific pieces of misinformation that are found from previous fact-checking efforts or existing fact-checking databases. However, there were no local fact-checking efforts on foreign misinformation when we started the project. Hence we decided to manually read all the posts 11,482 that we’ve collected to pick out the ones with misinformation. We soon realized that it was humanly impossible. We then limited our scope to posts with a minimum of 20 engagements. This reduced the number of posts to 6,248. Most of the posts were in Chinese and only 2 team members read Chinese. Together with a part-time research assistant, we spent hundreds of hours to go through each of the 6,248 posts and even fact-check them if there’s information that has not been verified.
This investigation began in December 2019. The Covid-19 pandemic exploded in Asia the next month. Our progress was severely disrupted and the completion of the investigation was delayed by 12 months. Its publication is a miracle created by the team’s dedication and perseverance towards quality journalism.
What can others learn from this project?
1. The methodology of this investigation, which was stated at the bottom of the story, is highly replicable. In fact, we developed our methodology by replicating and adjusting the methodologies of previous investigations done by other journalists.
2. There are issues that can’t be solved by technology. We attempted to use AI and machine learning to identify misinformation in Facebook posts but failed to find any solution. Eventually hundreds of man hours were spent to solve the problem but the outcomes were rewarding.
3. Launching an investigation that requires new skills and tools might seem daunting but it is one of the best ways to upgrade the skill set and toolbox of a newsroom. It is an investment that is worth taking.
4. If you hit a wall during your investigation, seek help from other journalists or experts. Many are willing to help. There are plenty of online self-learning resources for journalists too. Journalism conferences like GIJN are a good place to identify specialists or experienced journalists in the topics that you are investigating.
5. Having a clear plan and knowing your milestones is key to completing a long-term investigation even in the event of a major disruptor like Covid-19.