New Decision Protects Political Speech Critical of Outgoing Japanese Prime Minister Fumio Kishida
September 10, 2024
In the case of a user’s reply to a Threads post about the Japanese Prime Minister and a tax fraud scandal, it was neither necessary nor consistent with Meta’s human rights responsibilities for the content to be removed. This case grapples with the issue of how Meta should distinguish between figurative and actual threats of violence. The Board has repeatedly highlighted over-enforcement against figurative threats. It is concerning that Meta’s Violence and Incitement policy still does not clearly distinguish literal from figurative threats. In this case, the threat against a political leader was intended as non-literal political criticism calling attention to alleged corruption, using strong language, which is not unusual on Japanese social media. It was unlikely to cause harm. Even though the two moderators involved spoke Japanese and understood the local sociopolitical context, they still removed the content in error. Therefore, Meta should provide additional guidance to its reviewers on how to evaluate language and local context, and ensure its internal guidelines are consistent with the policy rationale.
About the Case
In January 2024, a Threads post was shared that shows a news article about the Japanese Prime Minister Fumio Kishida and his response to fundraising irregularities involving his party. The post’s caption criticizes the Prime Minister for tax evasion. A user replied publicly to that post, calling for an explanation to be given to Japan’s legislative body followed by the word “hah,” and referring to the Prime Minister as a tax evader by using the phrase “死ね,” which translates as “drop dead/die” in English. The phrase is included in several hashtags and the user’s reply also includes derogatory language about a person who wears glasses.
The user’s reply to the Threads post did not receive any likes and was reported once under Meta’s Bullying and Harassment rules. Three weeks later, a human reviewer determined the content broke the Violence and Incitement rules instead. When the user appealed, another human reviewer decided once more that the content was violating. The user then appealed to the Board. After the Board selected the case, Meta decided its original decision was wrong and restored the user’s reply to Threads.
Around the time of the original Threads post and the user’s reply, Japanese politicians from the Liberal Democratic Party had been charged with underreporting fundraising incomes, although this did not include Prime Minister Kishida. Since 2022, when former Prime Minister Shinzo Abe was assassinated, there has been some concern about political violence in Japan.
Fumio Kishida recently announced he will not seek re-election as leader of Japan’s Liberal Democratic Party on September 27, 2024 and is to step down as Prime Minister.
Key Findings
The Board finds that the phrase “drop dead/die” (translated from the original “死ね”) was not a credible threat and did not break the Violence and Incitement rule that prohibits “threats of violence that could lead to death.” Experts confirmed the phrase is broadly used in a figurative sense as a statement of dislike and disapproval. The content also points to this figurative use, with inclusion of the word “hah” expressing amusement or irony.
However, Meta’s Violence and Incitement rule that prohibits calls for death using the phrase “death to” against high-risk persons is not clear enough. Meta’s policy rationale suggests that context matters when evaluating threats but, as has been noted by the Board in a previous case, Meta’s at-scale human reviewers are not empowered to assess the intent or credibility of a threat, so if a post includes threatening statements like “death to” and a target (i.e., “a call for violence against a target”), it is removed. Repeating a 2022 recommendation, the Board calls on Meta to include an explanation in the policy’s public language that rhetorical threats using the phrase “death to” are generally allowed, except when directed at high-risk individuals, and to provide criteria on when threatening statements directed at heads of state are permitted to protect rhetorical political speech.
It is also confusing how this policy differs in its treatment of “public figures” and “high-risk persons.” Currently, medium severity violence threats against public figures are only removed when “credible,” compared with content removal “regardless of credibility” for other individuals. More confusingly still, there is another line in this policy that gives “additional protections” to high-risk persons. Internal guidance on this to reviewers, which is not available publicly, specifically indicates that “death to” content against such high-risk people should be removed. When asked by the Board, Meta said its policy offers greater protection for users’ speech involving medium severity threats at public figures because people often use hyperbolic language to express their disdain, without intending any violence. However, threats of high-severity violence, including death calls against high-risk persons, carry a greater risk of potential offline harm. In this case, Meta identified the Japanese Prime Minister as falling into both categories. The Board has real concerns about the policy’s definitions of “public figures” and “high-risk persons” not being clear enough to users, especially when the two categories interact.
In response to the Board’s previous recommendations, Meta has completed some policy work to strike a better balance between violent speech and political expression, but it has not yet publicly clarified who “high-risk persons” are. The Board believes providing a general definition with illustrative examples in the Community Standards would allow users to understand that this protection is based on the person’s occupation, political activity or public service. The Board offered such a list in the 2022 Iran Protest Slogan case.
The Oversight Board’s Decision
The Board overturns Meta’s original decision to take down the content.
The Board recommends that Meta:
- Update the Violence and Incitement policy to provide a general definition for “high-risk persons” clarifying that they encompass people such as political leaders, who may be at higher risk of assassination or other violence – and provide illustrative examples.
- Update internal guidelines for at-scale reviewers on calls for death using the phrase “death to” directed against “high-risk persons,” specifically to allow posts expressing disdain or disagreement through non-serious and casual ways of threatening violence. Local context and language should be taken into consideration.
- Hyperlink to its Bullying and Harassment definition of public figures in the Violence and Incitement policy, and other relevant Community Standards, where such figures are referenced.
For Further Information
To read public comments for this case, click here.
