STUDY 2: IN-DEPTH INTERVIEW
6.1 Third-Party Fact-Checkers
6.1.2.2 Information Processing and Fact-Checking
國
立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
determinant of third party fact-check results. Professionals employed other sources too such as digital tools, academic journals, official statements and so on as explained in the aforementioned section.
6.1.2.2 Information Processing and Fact-Checking
After information gathering stage, how information was processed and procedures for fact-checking will be examined. Information processing and fact-fact-checking were simultaneously examined as third-party fact-checking organization focused on fact-checking, unlike news media, where fact-checking to clarify information was a part of journalism. Examined codes include correcting before publication, constructing, contacting fake news creator, tracing false claims, correcting after publication, showing results, and training.
Since fact-checking was the main objective of Mafindo, procedures for fact-checking became highly relevant. For the respondents, fact-checking went beyond gathering information and producing content, but ensuring that fact-checking practices were done consistently and results could be trusted.
Correcting: Before Publication
Correcting is defined as the process of revising information content by having content checked, rewritten, and editing writing errors (Himma-Kadakas, 2016). Compared to news media, two forms of correcting were found during analyses. The first type of correcting was corrections made before publication, and the second type occurred after publication (discussed at a later section). Given the main objective of fact-checking was to check and clarify information, it was possible for corrections to be made from the beginning. Analyses showed that correcting at this stage did not refer to directly determining an information was fake news, but which potential components of an information were the most likely to be false.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Through the accumulated experience, respondents were able to identify potentially problematic components that helped them in directing their fact-checking procedures. Using the case of post-Palu fake news disaster, F2 said:
For the aftershock (fake news) narrative, at first glance, we were able to somehow know that big chance, it was untrue… the narrative used was a typical format used in the aftermath of a disaster.
Respondent F2 did not insinuate that fact-checkers could directly determine if information was false, but the possibilities of which part that needed checking and correcting. Knowing which possible problematic element helped fact-checkers to be more directed, such as identifying which news articles or photos they need to search to clarify the factuality of information.
Constructing
Constructing is defined as ascribing newsworthiness to the content by emphasizing certain aspects of the information (Boukes & Vliegenthart, 2017). Compared to news professionals who had to construct their content in certain manners, third-party fact-checkers did not have certain writing angles due to the lack of agenda or affiliation to business or politics. . This organization has SOP when conducting their fact-checking, but no SOP was tailored to fit into the case of Palu disaster. “It’s the same SOP for all types of fake news,” said F2, and when further questioned, F2 claimed that the organization only worked on the facts that occurred. Unlike news media that might have different approach to address different forms of cases (e.g., politics, disaster, society, etc.) respondents claimed that they did not have different approaches. Different approaches in this context include avoiding certain topics, glossing over an issue, or framing an issue into the desired picture by the editorials/team.
The SOP for fact-checking are divided into several steps, first, after the selection of which fake news to be debunked, fact-checkers would start to gather their information. After identifying
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
which claims to fact-check, respondents would start their work. During the fact-checking process, respondents proceeded fact-checking activities based on Mafindo’s code of ethics, which are impartiality, neutrality, non-discriminatory, focusing on the facts and so forth. At times, when problems in fact-checking arise, discussions with the fact-checking team occurred. For example, when fact-checkers were unsure about how to continue their fact-checking, “Sometimes it was the lack of credible sources; sometimes it was hard to look for supporting evidences,” said F2, who then proceed to give an example. A case was when no mainstream news media reported about the issue and information could only be found from blog sites which led to a team discussion on how to proceed. At the end, the team opted to publish the result by indicating as much as sources they could find and supporting evidences to back it up (e.g., academic journals).
Upon completion, discussions about the fact-checked results were to be discussed with the fact-checking team, comprised of the director of fact-checking committee (F1) and other members of the fact-checking team, comprised of seven other fact-checkers. Discussions involved evaluation if certain content is ready to be published, F3 said, “We check if the evidences were there, the conclusion was correct or the sources sourced were right.” Once the content was ready, and final check was conducted to ensure that the facts were correct, consistent, and coherent, results would then be published using their predetermined format to promote coherence.
For all of their fact-checked content, a consistent format was utilized for coherence and avoiding misconceptions. F1 elaborated:
We always use the same format; we usually begin with the topic, the categorization…
meaning is the information right or wrong, and if it is a wrong, what type of false information is it… next, we have the conclusion first and so forth.
This writing structure of fact-check reports was that the conclusion was placed at the top and at the beginning of the content. F1 explained that this was done to cater to Indonesian’s reading
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
habits where people tended to read the topic and jumped to their conclusion. He continued that there had been cases where the public misunderstood the content’s topic, which was related to the fake news, as the fact-checked content, and learning from these experiences, the organization decided to change their format and kept the current format. In order to avoid recurring misunderstandings, the organization decided to change their style and put their conclusion from the very beginning. By doing so, even if Indonesians would only read the first part of their content, they would be exposed to the fact-checked conclusion, placed at the very top of the page.
Other respondents concurred that they employed such a consistent format for all types of fake news, as they minimized the potentials for people to misunderstand the content. Web observations showed similar results, showing that all fact-checked content related to post-Palu disaster fake news and other fake news exhibited the same format, such as providing conclusion at the beginning, attaching references and including screenshots. Compared to journalism’s inverted pyramid style in newswriting, fact-check professionals first elaborated what they meant when they labelled a content as “false”, further categorizing it into seven types of disinformation, namely (1) satire or parody, (2) misleading content, (3) imposter content, (4) fabricated content, (5) false connection, (6) false context, and (7) manipulated content. After the categorization, fact-checkers then included the original narration, picture, or video before pointing out which components of the disinformation that was wrong. At the end of the content, fact-checkers would include their references and links. Unlike newswriting, fact-checked content did not provide interview excerpts from sources nor description about the contextual background. On the other hand, content was very visual by including various screenshots to highlight the important aspects of the information, and shorter narrative texts. This was done to highlight the most important aspects and avoid lengthy content that might drive readers away.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Contacting False News Creator
Secondly, contacting false news creator is defined as questioning claims from the source suspected of providing the wrong information (Graves et al., 2016). All respondents strongly rejected the idea to contact false news creator for two reasons. First, it was unnecessary to look for the creator, as the focus of their job was to check the factuality of a claim. According to the respondents, it was irrelevant to know who first created the fake news and they could conduct fact-checking without knowing who exactly created it. As an example, during post-Palu disaster’s aftershock fake news, it was only important to fact-check the factuality of the event. Investigation or releasing information about the creator might not satiate public’s anxiety as the public prioritized information about the aftershock itself.
Secondly, tracking the information creator might lead to misjudging someone and provide the wrong information to the public. This was one of the main rules at Mafindo, all fact-checkers are not allowed to release information about the person who created the fake news, unless it was clear who gave the (false) statements. For example, the case of FPI in the aftermath of Palu disaster indicated clear sources (i.e., Ministry of Communication and Information, and FPI). During the interviews, respondents stated that they made no particular efforts in trying to identify the culprit(s). F2 clarified:
We are not allowed to contact fake news perpetrators. Aside from the job scope that it not being our responsibility, we want to avoid the risks of pointing fingers at the wrong person;
we avoid doxing and causing harms to others.
According to F2, they would not want people to look for the person named in their content and harassed for something they might not have done. As an example, the respondent gave the case of Boston bombing where doxing was conducted and the person named was harassed. At the end, it was found out that the person had nothing to do with the attack but the damage has been
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
done as the person conducted suicide. Learning from this case, members of Mafindo did not want this to occur and avoid the crowd to act as “judges”. Therefore, contacting false news creator was never conducted by respondents, regardless of the case.
Tracing False Claims
Tracing false claims refers to tracking down the false information to its original source (Graves et al., 2016). Similar to the second code, this was not conducted by the fact-check professionals All respondents stated that what mattered the most was the factual information.
Although tracing false claims might lead to more information, respondents stated that it was out of their range to trace false claims. First, respondents claimed that they did not have the necessary resources and it was the police who had access to utilize tools to trace claims.
Tracing false claims in the context of third-party might refer to look back into previous content related to the claim or issue. F1 stated that they might trace back a case to provide context and deeper understanding about the case. A reason for doing this was due to some fake news with narratives that derived from previous cases or might lack the contextual background. During the event of post-Palu disaster fake news, tracing false claim was conducted when the history of the issue was needed, especially when it was related to politics. However, in cases such as the aftershock, tracing false claims was not conducted as it was a singular and non-continuous event that derived from Palu disaster itself.
Correcting: After Publication
Unlike the first part of correcting that occurred before publication, correcting after publication took place when errors were found on the published content. This second type of correcting was more similar to correcting made by news professionals. When mistakes occur, fact-check professionals would take down the wrong content, and re-upload it again along with the
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
edited version. This was done in order to maintain transparency, and to build trust within the public that they were willing to admit their mistakes and make the right changes. Respondents admitted that mistakes could happen due to internal mistakes in fact-checking, but most of the times, errors were thought to originate from the sources they used. F3 further elaborated:
Since we rely a lot on news media, if news media’s sources said something wrong, automatically, we would be wrong. In this case, we will release a post to clarify the mistakes.
For the respondents, there was a chain reaction when the original sources cited had the wrong information. It was not necessarily due to the incompetence of news media for making mistakes, but it could be due to the sources cited by news media themselves made errors that were unknown. Since fact-check professionals mostly utilized second-hand resources, like news media, there might be an issue with the extent news media might influence fact-checkers. When mistakes occurred due to mistakes from fact-checkers’ source (e.g., news media), this suggests that respondents might have little control over the “factuality” of their information. Although multiple resources were used to cross-check information along with other resources (e.g., digital tools, academic journals, etc.), there might still be possibilities for errors to occur not because fact-checkers were careless, but the lack of resources to conduct thorough first-hand investigations.
Despite that, respondents claimed that in the context of post-Palu disaster fake news, all respondents stated that there were no mistakes on their behalf and all clarifications related to the disaster’s fake news had been thoroughly fact-checked.
Showing Results
The next code, showing results is defined as the effort to communicate findings through information content produced by news professionals (Graves et al., 2016). Third-party fact-checkers gave lots of efforts and considerations to ensure that their results could be trusted. Trust
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
was an overarching theme for this code as respondents tried hard to build trust and to expose people to fact-checking. Unlike news media, Mafindo is relatively new and unknown to the wider public, prompting them to position themselves differently. The organization aims to be known as a neutral and objective fact-check organization yet by mining information from news media, this might be a challenge. Due to distrust in news media, sourcing information from news media might be detrimental to the end-result of third-party fact-checkers. To address this challenge when presenting their results, third-party fact-checkers would always include various sources and incorporate digital tools whenever possible. According to F1, having several sources might diminish the level of bias, yet this might not guarantee that the end-result would be bias-free.
However, with the current situation and resources the organization had, this might be the extent the organization could do. Furthermore, respondents were direct with their results; fact-checkers did not discriminate case(s) regardless of who was involved in the case. If result pointed that an information was false, it would be published as false followed by its explanation. As a result, respondents were at the receiving end of public criticism. F2 elaborated:
Sometimes when you are debunking fake news related to A, people might accuse us of supporting A and vice versa, so whatever we do, we would still be criticized. Despite that, we still tried to remain neutral and impartial in our fact-checking.
By remaining neutral and impartial regardless of the types of fake news or public’s opinions, all respondents acknowledged that they were trying to convey their effort to show that their results were indeed bias-free and objective. Even during post-Palu disaster fake news, they tried to convey these values and to treat all fake news equally regardless of who was involved.
Cases related to political-infused false information were examples; considering the involved parties included the government and a religious group, observation noted that fact-checked results were as consistent as other fact-checking content. Clarifications were made and backed up with
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
necessary evidences without any bias towards any parties.
Training
Training was only formally offered when this organization was in partnership with Google News Initiatives, or specifically after joining the alliance, CekFakta, in 2018 (see chapter 6.2). All respondents acknowledged that no formal training was offered before the alliance or when they first joined. “We are all self-learners,” said F1, and further continued to explain that when a new member joined the team, the person would be introduced on how to use the digital tools (e.g., Google reverse image) but it was not a formal program where trainers were present.
Fact-check professionals had to learn as they go along with their work, and a reason for the lack of formal training might be due to the lack of resources. F2 did acknowledge that their resources were not as large as news media, and even news media might have challenges in inviting international fact-checking trainers due to financial limitations. Nonetheless, it was difficult to evaluate the effectiveness of this method. Web observations showed that in the event of post-Palu disaster fake news, there were no problems with fact-checking results published by the organization, such as wrong conclusion or wrong resources used. Although respondents did not mention any problems faced that occurred due to the lack of training, this might be a subjective evaluation by them. For example, when errors appeared on their fact-checking results, it might not always be due to the blunders made by their resources, but internal human-errors as a result of minimum training.
6.1.2.3 Information Distribution
After information gathering and processing sessions are conducted, the researcher examined the stage of information distribution. Four subcodes related to information distribution were examined, namely social media, showing results, news websites, and traditional media. In
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
general, the main modes of communication for the respondents were social media, but both traditional media and news websites were still employed to distribute information.
Social Media
In terms of social media, as the main communication channel, respondents considered their Facebook forum as the most crucial. When debunking post-Palu disaster fake news, their social media platforms were the main channels. Currently, Mafindo had two main Facebook accounts, one as the forum and fan page. The forum enabled the public to post fact-checking content as long as they followed the predetermined format and regulations. On the other hand, the second platform was their fan page, where all fact-checked and important fake news clarifications were posted.
Importance in this context referred to widely viral fake news or fake news with great public importance (e.g. natural disaster, politics, etc.)
In addition, all respondents also acknowledged the relevance of using social media platforms, they acknowledged that fake news were mostly disseminated via social media and most Indonesians were often online on social media. Utilization of social media was to counter fake
In addition, all respondents also acknowledged the relevance of using social media platforms, they acknowledged that fake news were mostly disseminated via social media and most Indonesians were often online on social media. Utilization of social media was to counter fake