Donate Menu

Japan’s Shift to the Right: Computational Propaganda, Abe Shinzō’s LDP, and Internet Right-Wingers (Netto Uyo)

January 15, 2022
Volume 20 | Issue 2 | Number 4
Article ID 5677


Abstract: In recent years, academic research and investigative reports have brought to light several cases of computational propaganda (i.e. orchestrated attempts to manipulate public opinion or the outcome of elections via social media), as well as proof that filter algorithms amplify right-wing conservative content on Japanese social media. Piecing together the scattered pieces of a puzzle, this article summarizes recent findings from various disciplines and provides an overview of the mechanisms and potential scope of computational propaganda in politics. Although there is no conclusive evidence of who is behind this anonymous activity, an analysis of these findings demonstrates increasing circumstantial evidence that certain factions of the LDP are actively using computational propaganda, either by commissioning external contractors or through its online support group, J-NSC, thereby condoning and appealing to anonymous Internet right-wingers (netto uyo). The concluding discussion assesses the effects and consequences of computational propaganda on the political sphere in contemporary Japan.


Keywords: Computational propaganda, social media, Liberal Democratic Party (LDP), Populism, Dappi


At the time I started conducting research on social media, in 2013, my team and I were rather naïve in approaching the subject of the digital transformation of the political public sphere. In analyzing Twitter data collected before and during the Japanese 2014 general election, we were particularly puzzled by the fact that 80% of the tweets in our election-related dataset (collected via Twitter’s Application Program Interface (API)) were duplicates and near-duplicates of very few original tweets (Schäfer, Evert, and Heinrich 2017). The amount of copied content was enormous, especially compared to the German-language data, which we collected simultaneously. That we were extracting much more data from the Japanese Twittersphere than from its German counterpart was not surprising to us as such, since Japanese was (and still is) the second-most-used language on Twitter at the time we collected our data (16%, compared to English at 34%; Statista 2013). The preference of Twitter over other social media platforms in Japan can be explained by sociocultural norms, with the possibility that the use of pseudonyms allows for greater anonymity on Twitter (Wang 2016). Nevertheless, the number of identical and especially near-identical tweets in our Japanese dataset appeared highly suspicious to us.

In order to pursue our original research plan, to study the scope and character of political deliberation on Twitter, we initially decided to treat these duplicates and near-duplicates as noise. We filtered out duplicates and implemented a simple but efficient algorithm to identify the near-duplicates in our dataset as well. Lacking the appropriate terminology and also the right frame when we conducted our research, we realized only in hindsight that we were in fact looking at a largely unknown phenomenon in 2014: political computational propaganda on Twitter. With this new understanding, we adapted our research question and methodology, eventually concluding that the abundance of near-duplicated election-related tweets was the product of targeted campaigns by networks of bots to inflate the frequency and importance of certain election-related topics either by massive retweeting or by posting numerous slightly altered versions of the same tweet (with small modifications such as varying links to blogs or videos with political content). The aim of this computational propaganda strategy is to manipulate public opinion by creating trending topics through pushing certain hashtags or highjacking existing ones. Rather than engaging in political discussion with other users, computational propaganda campaigns affect the work of journalists, politicians, decision-makers and citizens through artificially amplified trending hashtags or topics (Schäfer, Evert, and Heinrich 2017).

Although the nature of social media platforms and the anonymity of accounts on Twitter in particular make it extremely difficult to pinpoint who is behind this activity, we found linguistic as well as contextual evidence that two factions of netto uyo (Internet right-wingers, a Japanese version of the alt-right) were the most likely candidates. In view of fresh evidence that has surfaced recently, this article revisits our previous results and conclusions. I will summarize recent findings to provide a more detailed picture of the effects and consequences of the digital transformation of the political sphere in contemporary Japan.

Unsurprisingly, computational propaganda is not a uniquely Japanese phenomenon. Orchestrated bot activity or the strategic dissemination of fake news and misinformation pose a threat to the integrity and stability of democratic institutions in all corners of the world (cf. Woolley and Howard 2019 for international case studies). Recent studies have shown that in China domestic public opinion manipulation is extensive on Weibo, whereas Twitter is used for positive propaganda aimed at foreign audiences; on Russian Twitter, some 45 percent of Twitter activity is managed by highly automated accounts, according to recent studies. Moreover, Russian-directed campaigns have targeted political actors in various countries in the past, including Poland, Ukraine or Germany. Ukraine, sparked by the ongoing Ukrainian-Russian conflict, is at the frontline of Russian computational propaganda, with Russian botnets engaging with Ukrainian nationalist botnets on various social media, including Facebook and Twitter. In the US, computational propaganda flourished during the 2016 US Presidential Election, when misinformation was strategically used to mislead voters, eventually contributing to Trump’s victory. In Germany, despite there being no evidence that social bots have played an influential role in elections, substantial misinformation has circulated during pivotal political moments, with various social media accounts of Russia Today being one of the most influential sources.

Research on computational propaganda on Japanese social media platforms, however, is still scant. Despite my summary of recent findings will not offer bulletproof evidence that the Liberal Democratic Party (LDP) is employing extensive computational propaganda, it highlights the extent of computational propaganda in Japan and the circumstantial evidence of LDP manipulation of Twitter uncovered so far. The methods of computational social science alone cannot provide final evidence, nor can investigative journalism deliver conclusive information to reveal the whole picture. The problem is that the use of computational propaganda doesn’t offer up the obvious clues required for conviction in, say, a murder case: an identifiable victim, a confessed perpetrator, and a smoking gun covered with fingerprints. In legal terms, engaging in computational propaganda is structurally rather similar to activities such as money laundering—criminal offenses committed through networks of intermediaries and subcontractors to cover up illicit acts. Accordingly, the perpetrators in these cases often can be convicted entirely on circumstantial evidence. Similarly, as we will see, the most recent cases of computational propaganda that have brought to light the indirect and direct involvement of LDP members were revealed as ancillary results of separate court trials. Adding to the murkiness of the situation is the decisive role of Japan’s Internet right-wingers, an anonymous force in the latent sphere of social media platforms.

So far, no lawsuit has revealed similar activities among the other major political parties. This might have various explanations. On the one hand, as the ruling party, the LDP is currently under higher scrutiny from investigative journalism, judicial prosecution, and academic research than the opposition parties. Naturally, this means that the statistical probability is significantly higher that questionable activities such as the ones described above are coming to light. Accordingly, if other parties were employing or tolerating computational propaganda as well, it would only be a matter of time before similar cases were disclosed. On the other hand, the current opposition parties were much slower than the LDP in adapting to the revised electoral law of 2013 and adopting efficient online tactics in their campaign strategies (Williams and Miller 2016), making them less likely to be engaged in computational political propaganda to the same extent as the LDP.


State of Research: Computational Propaganda and Amplification Bias

Recent research and investigative findings by journalists have not only revealed cases of computational propaganda but also provided profound evidence of a significant overrepresentation of conservative and far-right content on Japanese social media, especially Twitter. As early as 2017, an investigative report by IT Media Business, an online media portal specializing in IT-related content, showed that CrowdWorks, one of Japan’s biggest crowdsourcing services, had suspended a recruitment post seeking writers for a political blog called Japanese Politics (Nihon no seiji). The CrowdWorks post stated that the blog, despite its apparently nonpartisan and nonbiased title, was “limited to those with conservative views” and that “articles supporting the Democratic Progressive Party or the Communist Party will not be accepted.” Among the blog entries recommended in the recruitment post as reference samples for potential applicants were articles bearing titles such a “Of course we should revise Article 9 of the constitution and maintain an army,” or “We should not bother with South Korea,” and “People who vote for the Communist Party are anti-Japanese (han’nichi)”.

The media report provided reasonable suspicion that somebody was trying to muster a troll army to interfere in the outcome of the 2017 general election, to take place a month after CrowdWorks deleted the recruitment post. Sociologist Kimura Tadamasa (2018b, 2018a) went on to offer substantial empirical proof that conservative and right-wing opinions were in fact overrepresented online. Drawing on large-scale questionnaires and content analysis of the YahooNews comment sections in July and August 2016, Kimura shows that anti-Korean and anti-Chinese sentiments, a feeling of “unease” (iwakan) towards minorities, and “distrust towards the mass media” all were significantly overrepresented in Japan’s most popular online news portal at that time. In his discussion of the results, Kimura claims that this backlash of nationalist online discourse needs to be seen as the result of a suppression of conservative views in postwar Japan, a period dominated by liberal values, endorsed by “intellectuals as well as the mass media ”(Kimura 2018b, 141; cf. also 2018a, 289–293). Basing his interpretation on moral foundations theory (MFT), Kimura argues that as the foundations of the conservative pattern are suppressed, right-wingers turn to the Internet to give free rein to racist feelings of being deprived of their fair share by the minority groups and the elites supporting these minorities.1

Recent studies from the field of computational linguistics have proven that conservative content has a wider reach on social media than content from other political camps for the case of Japan. According to a study by Yoshida Mitsuo et al. (2021), partisan posts originating from conservative clusters on Twitter have a higher probability of reaching moderate users than those from liberal groups. Yoshida et al. (2021) cite various reasons for this phenomenon. Network analysis of 129,639,061 tweets collected from February 10, 2019 to October 7, 2020 that included the word “Abe” revealed that the conservative cluster enjoys higher reciprocity of followership from moderate accounts than the liberal cluster. This indicates that the cluster of conservative accounts has a significantly larger influence on the moderate cluster. Furthermore, on the level of tweet content, sentiment analysis revealed a more frequent use of emotion words in conservative tweets, especially those expressing “dislike,” “joy,” and “anger” (cf. diagram below).


Source: Yoshida et al. 2021.


Because filter algorithms on social media usually prioritize emotional expressions, the more fervent tweets from conservative clusters suggest another explanation for why they are more successful in using social media for political mobilization. Moreover, it appears that a higher linguistic similarity between conservative and moderate tweets, especially in the use of certain adjectives, may also contribute to the greater reach of tweets from the conservative camp.

Confronted with public and scholarly debates on how conservative and far-right political groups might benefit more from algorithmic amplification on social media, Twitter commissioned a group of researchers to analyze the functioning of its own filter algorithm (Huszár et al. 2021). In 2016 Twitter had introduced the optional “home timeline” in addition to the previous nonfiltered chronological feed of tweets.2 The analysis of Huszár et al. (2021) revealed that content from the LDP has double the reach of the Constitutional Democratic Party (CDP) on Twitter in the algorithmically prioritized home timeline (cf. the left diagram below).


Source: Huszár et al. 2021.


These results clearly demonstrate that the LDP is at an advantage over the opposition parties on social media, an effect that is usually known as algorithm bias. In most cases this bias occurs when self-learning algorithms are trained on data in which the bias is already inherent, with the result that the algorithm adopts this bias, reproducing and thus further amplifying it. This, however, does not answer the question why the training data, most likely Twitter’s own data stream, is biased in the first place. In the case of Japan, the LDP’s quick and efficient adaptation to the new law allowing for online campaigning even during election periods, far outpacing the other parties, might be taken as one explanation for the existence of this bias.


Hate Speech and Computational Propaganda

Besides the reliance on bots to artificially amplify the reach of certain tweets or hashtags and thereby manipulate public opinion online, another technique of computational political propaganda is the frequent use of professional or semiprofessional trolls3 to spread derogatory humor or defamatory verbal attacks against politicians, journalists, and academics. Legally categorized as hate speech, racist and sexist, personal verbal attacks have become prevalent alongside more moderate campaigns based on politicians’ ideological beliefs and job performance. Originally in the toolkit of methods wielded by secret services against ideological opponents, particularly during the Cold War, disinformation and defamation campaigns have become a core element of what is also known as “negative” propaganda in the age of the digital transformation of the political sphere.

Analyzing hate speech on Twitter, we were able to show its actual magnitude, especially against female politicians from the opposition parties in Japan (Fuchs and Schäfer 2021). To gain insight into everyday verbal abuse on social media, we intentionally gathered data (a dataset of 9,449,645 words) in a period during which there was no major election looming or ongoing, January to mid-April 2018. In the first stage of our study, we conducted a manual sentiment analysis of a randomized sample of 50 tweets for each of the four quantitatively most salient female politicians in our dataset. Next, we analyzed in greater detail those tweets that included negative sentiments. In all four cases, one-third to almost one-half of the negative tweets contained abusive language. Because all female politicians in these cases belonged to oppositional parties such as the CDP, one can argue that it is significantly more likely for politicians of the opposition than female politicians of the LDP to receive tweets conveying verbal abuse and hate speech. To put it another way, only 4.5% of the 200 tweets we analyzed had positive connotations, with the ultraconservative former LDP politician Koike Yuriko almost the sole recipient of these positive comments.

Verbal abuse and hate speech against female politicians is mostly based on gender and sexuality, that is, the outward appearance or alleged misbehavior that does not match the conservative image of a woman. In some instances, misogynist or sexist verbal abuse and hate speech includes nativist comments or racist slurs. The female politician Renhō (CDP), holding a dual Japanese and Taiwanese citizenship, for example, was often called “anti-Japanese” (han’nichi) or “traitor” (baikoku) and thus verbally excluded from the national community of “the Japanese.” With antifeminist sexism and nativism/racism at the center of online hate speech against female politicians, it is obvious that the harshest forms of such verbal abuse come from the pro-Abe faction of netto uyo, who support certain female LDP politicians only if they belong to the group of neoconservative hardliners around Abe.

Netto uyo differ from the old right-wing movements in Japan known for their black trucks, parading the streets of urban areas propagating their extremist ideology, and playing militaristic music. In contrast, netto uyo are commonly defined as “individual ‘heavy’ net users who are sympathetic to right-wing/nationalist views” though very “few of them” are “real-life activists or members of right-wing/nationalist organisations” (Sakamoto 2011). Sociologist Tsuji Daisuke (2008; cf. also Nagayoshi 2021 for a more recent study, arriving at similar results), who has interviewed these Internet right-wingers, attributes to them certain ideological attitudes: an anti-Korean and/or anti-Chinese racism/exclusivism, the belief that the prime minister should visit the controversial Yasukuni Shrine, approval of a revision of the constitution allowing for the maintenance of a proper military, support for the public singing of the national anthem and display of national flags in schools, and general distrust of the mainstream media. It is important to add that similar to the alt-right in the United States, many netto uyo in Japan express outspoken sexism and antifeminism (Yamaguchi 2019), often in the form of abusive verbal attacks against female politicians on social media (Fuchs and Schäfer 2021).

The actual number of netto uyo is hard to estimate, since most never make an offline appearance, as in demonstrations organized by groups like the ultranationalist Zaitokukai. According to several questionnaire-based studies, the core of Internet right-wingers consists of 50,000 to 100,000 internet users who actively post messages or comments, as well as a latent population of 1.2 million passive netto uyo (roughly 1–1.5% of Internet users) (Tsuda, Kayama, and Yasuda 2013; Nagayoshi 2021). However, Sakamoto Rumi (2011) emphasizes that it would be a mistake to treat the netto uyo as a homogeneous group. She describes netto uyo as “aggressive and shrill, but fragmentary.”

The enactment of the so-called Hate Speech Act in 2016, during Abe’s third term in office, has been decried as ineffective by oppositional politicians and legal scholars, since it set no penalties and was limited to threats of physical violence, not verbal discrimination. Even more striking is that the new law is based on a very narrow understanding of hate speech, addressing only xenophobic or racist attacks. The law does not cover misogynist hate speech, a problem as serious as racist remarks. Although Japan has been doing fairly well in regulating offline hate speech through administrative measures on the local level (cf. Löschke 2021), ithas failed to curb the problem online. The May 2020 suicide of Kimura Hana, a popular female wrestler and cast member of the Netflix reality show Terrace House, who had been persistently attacked in social media, demonstrated this failure and the urgent need for political and legal action. It was only after this incident that the government began drawing up legislation to counter cyberbullying. Kimura was the victim of intersectional discrimination, verbally abused not only for being a woman but also because of her Indonesian descent. 

As our research shows, female politicians are particularly vulnerable to online hate speech. And although these verbal attacks and smear campaigns are usually aimed at individuals, they are also directed against the political or journalistic institutions and bodies these individuals represent, contributing to the growing distrust of these institutions (Hayashi 2017, 2020). Nevertheless, the legal situation in Japan is not that different from other countries. In Germany, for instance, Renate Künast, a prominent politician of the Green Party who was targeted in an online hate speech campaign following her controversial statements regarding pedophilia by a fringe group within her own party, eventually filed lawsuits against Twitter and Facebook, with the outcome of the trial still pending. As mentioned in the introduction to this article, computational propaganda includes quantitative strategies such as the creation and massive dissemination of near-duplicated content. In her lawsuits against Facebook and Twitter, Künast argued that submitting separate requests to delete each individual variant of verbal abuse posed an insurmountable, “lifelong task” for the victim. Her lawsuit tried to force social media platforms to automatically delete all “semantically similar” posts that violated the law or the regulations of the platform. In Japan courts at least seem to have acknowledged that retweeting defamatory content from another account is equal to the original libel. In 2020 the Osaka high court ordered journalist Iwakami Yasumi to pay 330,000 yen in compensation to former Osaka governor Hashimoto Tōru for retweeting slanderous tweets.

The remainder of this article summarizes recent investigative journalistic reports and research results supporting the hypothesis that single representatives or perhaps even whole factions of the LDP employ or tolerate computational propaganda. The following examples provide overwhelming circumstantial evidence that computational propaganda has become both a very common practice for the LDP under Abe.


The LDP’s Online Campaign Strategy and the Role of Computational Propaganda

Computational propaganda on social media can be defined as the orchestrated attempt to manipulate public opinion or the outcome of elections via fake news and verbal attacks on prominent politicians, often with the help of networks of trolls and bots. Social bots and troll armies can effortlessly disseminate radical opinions and “alternative truths” or spread misogyny or racism online, contributing to the widening of what metapolitical strategists call the Overton window to the right, that is, the breaking down of discursive barriers and a broadening of the framework of ideas that are acceptable to voice in public. And even if it is unlikely that computational propaganda can provoke radical swings in public opinion, during “sensitive political moments when public opinion is polarized” (Howard and Kollanyi 2016)—such as the election campaign of Donald Trump or the UK's Brexit referendum—computational propaganda can tip the scales (Hegelich and Janetzko 2016; Howard and Kollanyi 2016; Kollanyi, Howard, and Woolley 2016). Moreover, computational propaganda can generate “censorship through noise” (Pomerantsev 2019). In this cesspool of misinformation, it has become more and more difficult to distinguish between fakes and facts, the legitimate expression of opinion and orchestrated verbal attacks, especially online.

The question remains whether we can trace these computational propaganda activities back to the LDP, the major beneficiary of this development among Japan’s established political parties. A closer look at the history and scope of the LDP’s online campaign strategy is revealing. The strategy took shape as early as 2009, at a time when the party was in the opposition. It was designed by Seko Hiroshige, formerly a section manager in the public relations department at Nippon Telegraph and Telephone (NTT) and now the minister of economy, trade, and industry (METI) in the LDP government. The LDP’s online campaign strategy gained traction with the 2013 revision of the Public Offices Election Law, which lifted the ban on online campaigning during the legally defined period of election campaigns. After this legal revision—a change sponsored by the LDP only after the party had suffered its biggest defeat in the 2009 general election—the party sought the assistance of major advertising agencies such as Dentsū and IT companies to intensify the training of its candidates in how to effectively use social media and to create a powerful IT infrastructure for online campaigning.

The LDP also established the so-called Truth Team (T2), a “cross between a public relations department and cyber-security task force” to monitor illicit online activities––such as the use of fake candidate profiles and accounts” (Nishida 2015, 216).4 Misleading posts were to be clarified by publishing accurate information on a designated website, and LDP’s Truth Team would analyze posts aimed against the party’s candidates on social media as well as on blogs and bulletin boards. In cases of defamation, the team would request that the platform delete posts. The team’s director, Hirai Takuya, was also the official chair of the Special Mission Committee on IT Strategy and in 2018 was appointed minister for digital transformation in charge of information technology policy.


Aso Tarō at the founding meeting of J-NSC at LDP’s headquarter on June 9, 2010.
Right before Aso spoke, Abe Shinzō delivered a five-minute address (
Video on YouTube).


Another unique element of the LDP’s online campaign strategy was the 2012 endorsement of the Jimintō Net Supporters Club (J-NSC), making it the party’s official online support group. In the same year, Hirai, the head of the Truth Team, was appointed the official representative of J-NSC. The organization grew out of a volunteer group of about 1,800 people who distributed pamphlets backing the LDP in the 2009 general election, in which the LDP lost its majority vote. The Mainichi shinbun reported that the number of J-NSC members had grown to 19,000 by 2017. To become a regular member, one merely needs to be at least 18 years old and hold a Japanese passport; party membership in the LDP is not necessary, according to the group’s website. The stated aim of J-NSC is to “amplify the strength of the LDP in order to reach a Japan full of dreams, hope and pride.”

Investigative journalist Kajita Yōsuke describes the J-NSC as an organization that not only “posts messages beneficial to the LDP on the Internet” but is also “at the center of negative campaigns against the opposition or critical forces.” Their strategy is “to post messages such as ‘Diet member XY of the Constitutional Democratic Party of Japan is saying this or that’ on the members-only message board,” thereby mobilizing “other members (…) to unfold a negative campaign and lavish the respective Diet member with anonymous criticism” (Shūkan Posuto 2019, 42). Kajita argues that with its integration into the campaign strategy of the LDP, J-NSC has turned into the “working brigade” of the party’s computational propaganda efforts, whereas the Truth Team could be considered its “control tower” (Shūkan Posuto 2019, 42). To say that J-NSC is just an online version of a kōen-kai (a local party support group), then, would be a dangerous underestimation of both the activity and the reach of the group in the online world. According to an investigation by Miyatake Mine on his personal website, there is a sizable overlap between membership of the J-NSC and the pro-Abe faction of netto uyo. His analysis of several Twitter accounts of self-declared J-NSC members has found that many display the rising-sun flag (kyokujitsuki) in their profile photos and express radical right-wing attitudes in their account descriptions. 

If these investigative reports prove to be accurate, the computational propaganda methods of this officially endorsed online support group of the LDP are not only reminiscent of troll activities, but one might even argue that J-NSC is to Abe’s LDP what the St. Petersburg–based Internet Research Agency (IRA) is to Putin. In public, the LDP understandably denounces this kind of illegal activity. Accordingly, an LDP staff member at a general assembly of the organization in autumn 2017 publicly declared that J-NSC members should avoid all activities punishable by law, including “the publication of false or distorted information about a candidate for the purpose of preventing their election to office” and “malevolent defamation and public insults” (Mainichi shinbun 2017). Given the frivolous atmosphere at the meeting, however, attending J-NSC members must have taken this statement as strategic advice rather than a prohibition. A report on the event in the English edition of the Mainichi shinbun describes a subsequent Q&A session with the LDP representative, during which “a supporter asked whether using puns to poke fun at other parties' names was acceptable”:


Former lower house legislator Masaaki Taira, who is the head of the party's online media division as well as J-NSC, responded, “Since they're parodies, I believe they are acceptable,” to which the participants laughed. Another male supporter asked if it was problematic that he posted a collage of an opposition party chief with a Chinese People's Liberation Army soldier, and another collage of a female opposition candidate with a so-called “comfort woman” online. In response, Taira said, “I think that should be left up to your personal judgment.” Again, there was a ripple of laughter from the participants. (Mainichi shinbun 2017)


Given what we know today of computational propaganda and the LDP’s involvement, the levity at this event is reprehensible. In the following section, I will discuss the two cases that have revealed the direct engagement of LDP members in orchestrated negative campaigns: the case of former LDP interior minister Kawai Katsuyuki and the case of the Twitter account Dappi.


Cases of Computational Propaganda: Kawai Katsuyuki and Dappi


The Kawai Case

On October 19, 2020, the Chūgoku shinbun reported details from the affidavit of an IT consultant to former interior minister Kawai Katsuyuki. The testimony was made public during the court proceedings against Kawai, accused of having purchased votes by paying 29 million yen to about 100 politicians, mayors, and supporters of his wife Kawai Anri, who was running for a seat in the Hiroshima district in the 2019 House of Councilors election against Mizote Kensei (an LDP lawmaker and a longtime political opponent and outspoken critic of Abe Shinzō). The LDP desperately wanted to keep both seats in the Hiroshima district, and further revelations brought to light that the LDP headquarters had transferred the extraordinarily large amount of 150 million yen to the local Hiroshima office prior to the election, thus having directly or indirectly financed the Kawais’ violation of the Public Offices Election Law. In 2021 the Tokyo district court  sentenced Kawai Anri to a year and four months in prison, suspended for five years, and her husband to three years in prison without probation and a fine of 1.3 million yen for violating electoral law (Niiya 2021). Kawai was a very close confidant of Abe until it became clear that he was behind the bribery. 


Shiomura Ayaka in a tweet from October 20, 2021, accusing Kawai of
having hired the IT consultant to run a negative campaign against her in the 2017 election.


In the course of the prosecution, the IT consultant was questioned when it was discovered that Kawai had ordered him to delete from his computers all data related to the briberies once the accusations became public. Besides admitting the deletion of the data, including a list of recipients of the bribe money, the unidentified consultant also confessed that Kawai had hired him to employ negative computational propaganda against his political opponent in the 2017 general election. According to the transcript of the affidavit published in the Chūgoku shinbun, the IT consultant not only took measures to improve Kawai’s reputation by manipulating results on search engines regarding critical issues related to him, but he had also “manufactured and published negative articles on a blog” in order “to worsen the image of the opposing candidate.” Kawai won by only 20,000 votes, exactly the kind of narrow margin in which computational negative propaganda could actually have made the difference. Kawai’s opponent was Shiomura Ayaka, formerly of Your Party (Minna no tō) and now of the CDP, who had already been under harsh attack on social media in the 2014 election (Schäfer, Evert, and Heinrich 2017). In the same year, Shiomura was verbally assaulted with blatantly sexist comments from two male LDP members during her speech on the city's childbearing policy at the Tokyo Metropolitan Assembly on June 18.

The affidavit revealed that the consultant’s activity was not confined to publishing negative blog entries. For these blog posts to have impact, the IT consultant had to circulate them widely via social media. As we know from related research, this usually requires the help of a network of trolls or bots extensively retweeting and liking posts related to the blog. Since this use of computational propaganda was not part of the legal allegations against Kawai in his bribery trial, and with the identity of the consultant undisclosed, the details of his activities remain obscure.


The Dappi Case

Much like Kawai’s case, the second instance of computational propaganda came to light unintentionally during an ongoing prosecution. CDP lawmakers Konishi Hiroyuki and Sugio Hideya went to court to enforce the disclosure of the identity of the Twitter account @dappi2019, the source of persistent attacks on them. With 170,000 followers as of November 2021, the account, which began posting in June 2019, has an enormous reach by Twitter standards. The ideological stance of the account’s profile clearly overlaps with the pro-Abe fraction of netto uyo: “Loving Japan. Hating the biased reporting of the mass media. Following the live broadcasts of the parliament” (see image below).


Image source: BuzzFeed News.


Besides the direct attacks on the two CDP lawmakers, the account is known for reposting videos from Toranomon News, a program broadcast on DHC TV, an online channel funded by the Japanese cosmetics company of the same name. In the past, DHC TV has disseminated conspiracy theories regarding the last US election, and Toranomon News is famous for hosting historical revisionists from the far right, including such controversial figures as Hyakuta Naoki and Fujii Genki. The account frequently posts highly edited and biased videos of Diet recordings as well. A video uploaded via the account in June 2021 is a good example of Dappi’s activities. A BuzzFeed News fact-check of the video, which denounces Edano Yukio’s (CDP) response to former prime minister Suga Yoshihide's Diet speech as “too pathetic” (aware sugiru Edano) and classified the content as false (Hatachi 2021b; see image below).


Image source: BuzzFeed News.


The Asahi shinbun conducted a detailed linguistic analysis of the account, looking at all 5,110 tweets and retweets sent from June 22, 2019, to October 1, 2021, the day the account stopped posting messages. According to the survey, the term “opposition” was the most frequently used noun (1,151 times), followed by "CDP" (1082 times), "DPJ" (Democratic Party of Japan) (710 times), "Communist Party" (391 times), and “mass media” (433 times). Of the total 5,110 tweets, 3,669 included videos or images, many of which were copied or cropped from Toranomon News (Asahi shinbun 2021a).

A brief analysis of Dappi’s activity via BotOmeter showed that the account was not a bot itself, nor did it have an unusually large number of bots or fake accounts among its followers. The account seemed to belong to a real person. However, in the course of the trial it became clear that the operator of the account was not an individual but an unnamed small-scale Internet company. According to its corporate registration, the company’s portfolio includes website design, consulting services, and online advertising (Asahi shinbun 2021b). As an investigative report published by the online media outlet Litera found, the company behind Dappi had received various direct contracts from the LDP for the maintenance of official LDP websites, document production, and other services, including work for LDP’s Tōkyō Branch Federation (Tōkyō-to Shibu Rengō-kai), the fund management group of former METI minister Obuchi Yukō (Litera 2021). Furthermore, it was revealed that the company had been awarded annual contracts worth 40 million yen from System Shūnō Center Company, Ltd. (Kabunushi Kaisha Shisutemu Shūnō Sentā). Both BuzzFeed News and Litera have reported that System Shūnō Center, which specializes in financial services, is in fact a dummy company of the LDP, with the LDP itself the major client and top-ranking LDP officials—including Japan’s current prime minister, Kishida Fumio—having served as members of the board of directors in 2001 and 2003–2005 (Hatachi 2021b; Litera 2021).

An analysis of the daily activity of the Dappi account via BotOmeter shows that almost all tweets were posted during regular office hours, roughly between 9 a.m. and 9 p.m., and rarely on weekends. This can be taken as robust evidence that @dappi2019 isn’t some lone-wolf netto uyo who posted more than 5,000 tweets over little more than two years, but rather that it was employees of the Internet company who were posting during their workday (Hatachi 2021b).

Temporal analysis of the account @dappi2019 (as of November 22, 2021),
conducted by the author via


Since the Dappi scandal surfaced right before the last general election, the case was hotly debated in the House of Councilors on October 13, 2021. Comparing the Dappi case to the Kawai case, CDP politician Mori Yūko in her statement summoned the prime minister to state that no members of the LDP would engage in computational propaganda and thereby unfairly distort the results of the upcoming election. Instead of confirming or outrightly rejecting the accusations, Kishida responded evasively with tu quoque generalizations, stating that “not only the members of my party, but all members of the Diet and every candidate running in the election should … engage in political activities only in accordance with the Public Offices Election Law” (Litera 2021).

In December 2021, new information regarding the operator of Dappi and its tight connection to the LDP surfaced. According to a recent investigative report on BuzzFeed News, Diet member Yamamoto Yūji (LDP), the former minister of agriculture, forestry, and fisheries, had business connections with the president of the web company as well (Hatachi 2021a). The first hearing in the lawsuit of the two DPJ lawmakers Konishi Hiroyuki and Sugio Hideya took place on December 10. It remains to be seen what new information the trial will unearth.



Various scholars and commentators have observed a rightward shift in Japanese politics in recent decades. One of the most outspoken critics of this tendency, political scientist Nakano Kōichi, argues that although this move was facilitated by the collapse of the political left, “the right itself drifted rightward,” too. Without a successful extreme right-wing party on the national level, the LDP, formerly “a ‘broad church’ that included some liberals as well as ‘One Nation’ conservatives,” turned into “a much more consistently rightwing party in socioeconomic and education policy, as well as in foreign and security policy” (Nakano 2015; 2016, 23; 2018). With a powerful pro-Abe troll army fighting the political opposition online and the algorithms of social media amplifying right-wing conservative content, the LDP’s successful implementation of computational campaign strategies since 2009 has decisively contributed to this rightward shift in Japanese politics.

As this article demonstrates, a close ideological proximity and symbiotic relationship existed between the LDP under Abe and a certain faction of netto uyo. On the one hand, since politics has expanded beyond the public sphere of official campaign platforms and the mass media and into the sphere of social media, the LDP has tightened its links with the pro-Abe activists in the netto uyo. The LDP has not only closed ranks with strong offline support organizations, such as Nippon Kaigi, but also managed to mobilize portions of the netto uyoku (or netto uyo) through its online support group, J-NSC, to push its nationalist agenda on social media. On the other hand, the beginning of the 2000s provided excellent “discursive opportunity structures for a certain faction of netto uyo” (Koopmans and Statham 1999), allowing it to link its nationalist online discourse to larger politics, the right wing of the LDP in particular (Higuchi 2018). Put differently, it was not only with the support of ultranationalist lobby organizations such as Nippon Kaigi but also with the help of netto uyo, manipulating a huge cyber army of bots on Twitter pushing a similar nationalist agenda, as well as the tactical use of social media and computational propaganda that the LDP was securing its dominant position in the past decade. 

Both the right wing of the LDP and the far right tend to use the empty signifier “anti-Japanese” (han'nichi), evidence of how the extremist discourse of Internet right-wingers has interlinked with Abe's nationalistic views. The strategic deployment of “anti-Japanese” as a discursive marker to denounce political opponents from the left in fact originated in the latter half of the 1990s. It was at this time that right-leaning magazines such as Seiron and Shokun! began to use the term “anti-Japanese”, attaching this label to proponents of what their authors regarded as the supposedly “masochistic” view of Japanese history. Not only did these publications blame critical historians for this view, they even declared NHK and left-leaning media outlets such as the Asahi shinbun enemies of the Japanese people. Jōmaru Yōichi (2011) has found a drastic escalation in the frequency of the term han'nichi in both magazines: it appeared in only two articles between 1985 and 1989, in six articles between 1990 and 1994, in 26 articles between 1996 and 1999, in 24 articles between 2000 and 2004, and in 52 articles between 2005 and 2009. With the advent of image bulletin boards such as 2channel and subsequently social media, the expression “anti-Japanese” quickly migrated into the online realm (cf. Itō 2019). Drawing on an analysis of the online communication of netto uyo on these platforms, Sugano Tamotsu argues that netto uyo are in fact longtime readers of these magazines who have a particular preference for articles by writers affiliated with Nippon Kaigi (Sugano 2015).

Moreover, Jōmaru and Sugano find a pro-Abe bias in these right-leaning magazines,  which tend to cover the LDP leader disproportionally and favorably. Abe, for his part, has been fighting his own battle against the mass media since around 2009; in particular, he blames the reporting of left-leaning mass media outlets as an important factor in his resignation in 2007. Even though Abe does not explicitly use the term “anti-Japanese” himself—instead describing representatives of opposing liberal or left-leaning organizations as “disgraceful adults” (hazukashii otona)—he has appealed “to the online nationalists and suggested that he agreed with their conservatism,” his rhetoric echoing “the online activists' complaints about the left-wing bias of traditional media outlets” (Murai and Suzuki 2014). In 2012 Abe informed the public that he considered online video broadcasts and posts on social media “the fairest and most interactive forums for lively political discussion” because it enabled him to “avoid the arbitrary editing by the mass media” (Murai and Suzuki 2014). According to sociologist Tsuda Daisuke, his critical stance towards the liberal media in particular has become an integral part of Abe's online identity since the beginning of the 2010s (Tsuda, Kayama, and Yasuda 2013).

Abe thus successfully implemented a political strategy reminiscent of the rhetoric of right-wing populists: a combination of anti-elitism and anti-pluralism. It is typical of this kind of populism that its proponents “claim that they, and they alone, represent the people” (Müller 2016). By referring to political opponents or journalists as enemies of the people, right-wing populist politicians rhetorically “legitimize” their right to exclusively represent “the people,” often by claiming that “whoever does not support populist parties might not be a proper part of the people” (Müller 2016). As Nagayoshi Kikuko (2021, 29) has convincingly demonstrated in her research, Abe’s right-wing populist rhetoric and outspoken nationalism has made him broadly electable. According to her findings, the majority of netto uyo vote “more often … for the LPD, the established conservative party, than for the newly emerging radical right-wing party, the PJK, or for Makoto Sakurai, who is a radical right-wing candidate.” Time will tell whether netto uyo will continue to support the LDP to the same extent under Abe’s successors.

As we have seen, although there is no smoking-gun evidence that the LDP is systematically employing computational propaganda, the findings presented in this article suggest deepening circumstantial evidence that certain factions of the LDP are using computational propaganda by commissioning external contractors through its online support group, J-NSC or by condoning the anonymous activity of pro-Abe netto uyo. Further ethnographic, sociological, and computational research is necessary to ascertain whether the lawsuits involving Kawai and Dappi were singular events or just the tip of the iceberg. What is clear is that the LDP largely benefits from the current situation. 

Of course, computational propaganda is not just for winning an election. The stakes are high, and the potential consequences are not limited to politics but extend throughout the democratic system. With its far-reaching connective architecture and functionality, social media can facilitate the strategic mainstreaming of marginal and even extremist political positions (as we have seen with regard to the populist term “anti-Japanese”). And normalizing radical positions via social media is a central tool of what right-wing populists and others on the far right describe as “metapolitical” strategy: the strategic dissemination of extremist ideas and discourses in the “pre-political” sphere of culture in order to pave the way for their inclusion in the political public sphere. With the help of botnets and troll armies, it has become much easier than in pre-digital times to disseminate racist and misogynist views via hate speech and to spread “alternative truths.”

The possible outcomes of these developments reach far beyond party politics. They include a barbarization of language in general and political discourse in particular; the potential distortion of public opinion and the manipulation of elections; an increasing distrust of democratic institutions, including mass media, academia, and the political system; and the growing polarization of societies, which heightens their vulnerability to computational propagandistic interference not only within a country but potentially from the outside as well (Brown 2021).


Asahi shinbun. 2021a. "Dappi no tweet, dare ga tōkō?" [Who is posting Dappi’s tweets?]. 3 December 2021 (accessed 3 December 2021).

Asahi shinbun. 2021b. "'Dappi no tweet ha meiyo kison': Rikken gi’in ga webbu kankei kaisha teiso" ["Dappi’s tweets are libel”: CDP lawmakers file lawsuit against an Internet company]. 13 October 2021 (accessed 13 October 2021).

Asahi shinbun AJW. 2020a. "Abe Forced to Call Off Bid to Extend Retirement Age of Prosecutors." 18 May 2020 (accessed 20 May 2020).

Asahi shinbun AJW. 2020b. "LDP Brushes Off Millions of Protest Tweets over Prosecutor." 13 May 2020 (accessed 15 May 2020).

Brown, J.D.J. 2021. "Russian Strategic Communications toward Japan: A More Benign Model of Influence?" Asian Perspective 45(3): 559–586.

Fuchs, T., and F. Schäfer. 2021. "Normalizing Misogyny: Hate Speech and Verbal Abuse of Female Politicians on Japanese Twitter." Japan Forum: 33(4): 553-579.

Hatachi, K. 2021a. "Dappi hasshin-moto no shachō ha 'yūjin'” [The sender of Dappi’s tweets is a "Friend"], BuzzFeed News, 2 December 2021 (accessed 3 December 2021).

Hatachi, K. 2021b. "Yatō hihan o kurikaesu akaunto 'Dappi' no un’ei hōjin? Jimintō shibu ya kokkai gi’in ga torihiki, seiji shikin shūshi hōkoku-sho nado de akiraka ni" [Who is the cooperation behind the account “Dappi” that has been criticizing the opposition?]. BuzzFeed News 11 October 2021 (accessed 12 October 2021).

Hayashi, K. 2017. Media fushin: nani ga towarete iru no ka [Distrust towards the media: What is being questioned?]. Tōkyō: Iwanami Shoten.

Hayashi, K. 2020. "The Silent Public in a Liberal State." In The Crisis of Liberal Internationalism, edited by Y. Funabashi and G. J. Ikenberry, 325–358. Washington, DC: Brookings Institution Press .

Hegelich, S., and D. Janetzko. 2016. "Are Social Bots on Twitter Political Actors? Empirical Evidence from a Ukrainian Social Botnet." Proceedings of the Tenth International AAAI Conference on Web and Social Media. Cologne: ICWSM.

Higuchi, N. 2018. "The Radical Right in Japan. In The Oxford Handbook of the Radical Right. Oxford: Oxford University Press.

Howard, P. N., and B. Kollanyi. 2016. "Bots, #Strongerin, and #Brexit: Computational Propaganda during the UK-EU Referendum." Social Science Research Network (accessed 29 May 2017).

Huszár, F., et al. 2021. "Algorithmic Amplification of Politics on Twitter." arXiv (accessed 22 October 2021).

Itō, M. 2019. Netto uha no rekishi shakaigaku: andāguraundo Heisei-shi, 1990–2000-nendai [A historico-social history of the Internet far right: An underground history of the Heisei period, 1990s–2000s]. Tōkyō: Seikyūsha.

Jōmaru, Y. 2011. Seiron no kenkyū [A study of Shokun! and Seiron]. Tōkyō: Iwanami shoten.

Kimura, T. 2018a.Haiburiddo esunogurafi: Nettowaku komyunikeshon kenkyu no shitsuteki hoho to jissen [Hybrid ethnography: The quantitative method and praxis of network communication research]. Tōkyō: Shin’yōsha.

Kimura, T. 2018b. “'Netto yoron' de hoshū ni tatakareru riyū: Jisshō-teki chōsa dēta kara" [Reasons for being attacked by conservatives in the “internet public opinion”: Based on data from an empirical analysis]. Chūō kōron 132(1): 134–141.

Kollanyi, B., P. N. Howard, and S. C. Woolley. 2016. "Bots and Automation over Twitter during the First US Presidential Debate." COMPROP Data Memo 1: 1–4 (accessed 5 January 2022).

Koopmans, R., and P. Statham. 1999. "Political Claims Analysis: Integrating Protest Event and Political Discourse Approaches." Mobilization: An International Quarterly 4(2): 203–221.

Litera. 2021. "Netto uyo Dappi un’ei to no torihiki o hōjirareta Jimintō damī hōjin no jitsumei" [Real name of the LDP dummy company reportedly doing business with the online right-winger Dappi revealed!] Litera, 15 October 2021 (accessed 18 October 2021).

Löschke, A. 2021. "Administrative Measures against Far-Right Protesters: An Example of Japan’s Social Control." Social Science Japan Journal 24(2): 289–309.

Mainichi shinbun. 2017. "19,000-strong LDP supporters’ group debates how to knock down election rivals." 18 October 2017 (accessed 22 October 2017).

Müller, J.-W. 2016. What Is Populism? Philadelphia: University of Pennsylvania Press.

Murai, S., and T. Suzuki. 2014. "How the Japanese Social Media Users Discussed the Senkaku Controversy." In The Dispute Over the Diaoyu/Senkaku Islands: How Media Narratives Shape Public Opinion and Challenge the Global Order, edited by T. A. Hollihan, 169–199. New York: Palgrave Macmillan.

Nagayoshi, K. 2021. "The Political Orientation of Japanese Online Right-Wingers." Pacific Affairs 94(1): 5–32.

Nakano, K. 2015. Ukeika suru Nihon seiji [Japanese politics tilting to the right]. Tōkyō: Iwanami shinsho.

Nakano, K. 2016. "New Right Transformation in Japan." In Disasters and Social Crisis in Contemporary Japan: Political, Religious, and Sociocultural Responses, edited by M. R. Mullins and K. Nakano, 23–41. London: Palgrave Macmillan.

Nakano, K. "History, Politics, and Identity in Japan." In Identity, Trust, and Reconciliation in East Asia: Dealing with Painful History to Create a Peaceful Present, edited by K. P. Clements, 201–222. London: Palgrave Macmillan.

Niiya, E. 2021. "Kawai Drops Appeal, Accepts Prison Sentence for Vote-Buying." Asahi shinbun AJW, 21 October 2021 (accessed 22 October 2021).

Nishida, R. 2015. Media to Jimintō [The media and the Liberal Democratic Party]. Tōkyō: Kadokawa shinshō.

Pomerantsev, P. 2019. This Is Not Propaganda: Adventures in the War against Reality. London: Faber & Faber.

Sakamoto, R. 2011. “'Koreans, Go Home!' Internet Nationalism in Contemporary Japan as a Digitally Mediated Subculture." Asia-Pacific Journal 10(2) (accessed 17 November 2017).

Sugano, T. 2015. Nippon kaigi no kenkyū [An inquiry into the Japan Council]. Tōkyō: Fusōsha.

Schäfer, F., S. Evert, and P. Heinrich. 2017. "Japan’s 2014 General Election: Political Bots, Right-Wing Internet Activism, and Prime Minister Shinzō Abe’s Hidden Nationalist Agenda." Big Data 5(4): 294–309.

Shūkan Posuto. 2019. “'Han-Abe' wo tataki makuru 'senden kōsaku butai' no sosei" [The birth of the propaganda division fighting the 'anti-Abe camp’]. 12 December 2019 (accessed 1 November 2021).

Statista. 2013. "Most-used languages on Twitter as of September 2013." (accessed 17 November 2017).

Sugiyama, S. 2020. "Former top Abe aide and wife arrested in vote-buying scandal." Japan Times, 18 June 2020 (accessed 17 November 2021).

Tsuda, D., R. Kayama, and K. Yasuda, eds. 2013. Abe seiken no netto senryaku [The Internet strategy of the Abe administration]. Tōkyō: Tsukuru shuppan.

Tsuji, D. 2008. Intānetto ni okeru ‘ukei-ka’ genshō ni kansuru jisshō kenkyū: Chōsa kekka gaiyō hōkokusho [An empirical study of right-wing phenomena on the Internet: A summary report] (accessed 17 November 2017).

Wang, S.S. 2016. “To Tweet or Not To Tweet: Factors Affecting the Intensity of Twitter Usage in Japan and the Online and Offline Sociocultural Norms.” International Journal of Communication, 10: 2637–2660.

Williams, J. A., and D. M. Miller. 2016. "Netizens Decide 2014? A Look at Party Campaigning Online." In Japan Decides 2014: The Japanese General Election, edited by R. J. Pekkanen, S. R. Reed, and E. Scheiner, 144–152. London: Palgrave Macmillan.

Woolley, S., and P.N. Howard. 2019. Computational Propaganda: Political Parties, Politicians, and Political Manipulation on Social Media. New York: Oxford University Press.

Yamaguchi, T. 2019. "Netto uyo to feminizumu" [Internet right-wingers and feminism].) In Netto uyo to ha nani ka [What are Internet right-wingers], edited by N. Higuchi et al., 164–187. Tōkyō: Seikyūsha.

Yoshida, M. et al. 2021. "Japanese Conservative Messages Propagate to Moderate Users Better Than Their Liberal Counterparts on Twitter." Scientific Reports 11(1): 19224






According to MFT, moral foundations for the conservative pattern include a strong “in-group” feeling (i.e., loyalty towards the peer group, pride, anger against deceivers) and a longing for “authority” (a strong desire for the upholding of tradition, obedience to authorities/legitimacy, and respect).


In 2016 Twitter introduced the so-called home timeline, which orders tweets in an individual’s feed via machine-learning algorithms based on a personalized relevance model. Until then, Twitter presented tweets only in reverse chronological order.


 I want to make clear that I do not consider “trolls” a label for real humans but rather an umbrella term for a contemporary Internet phenomenon.


LDP’s Truth Team derived its moniker from a campaign by the same name initiated by US president Barack Obama in 2012. The aim of Obama’s truth team was “to recruit two million supporters to help debunk attacks on the president's record and hit back at his Republican rivals” (CBS News). On the original website, the truth team was described as “a network of supporters of President Obama who are committed to responding to unfounded attacks and defending the President’s record. When you’re faced with someone who misrepresents the truth, you can find all the facts you need right here—along with ways to share the message with whoever needs to hear it.”

Fabian Schäfer

Fabian Schäfer received his PhD from Leipzig University in 2008. He was appointed full professor and chair of Japanese studies (modern and contemporary Japan) at the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) in 2013. His research interests include the digital transformation of the public sphere, computational social sciences, corpus-based discourse analysis, and media history and media theory.