[Feature: Social Media and Society] Yuya Shibuya: Elections and Social Media

Social media (hereafter SNS) is no longer just a means of contacting friends and acquaintances; it has become a major distribution channel for news, entertainment, and even political information. Its social role in the information flow that supports our lives is now extremely significant. Its influence is also substantial during elections, where the situation in which voter behavior and public opinion formation are closely linked to discourse on SNS has entered a qualitatively different stage compared to the era when newspapers and television were the primary media.

On the other hand, SNS also serves as a breeding ground where misinformation (information that is incorrect as a result) and disinformation (information intended to deceive) spread in a short period of time. While algorithms that optimize displayed content based on posting and viewing history enable presentations tailored to user interests, they also carry the side effect of amplifying information of uncertain authenticity.

In Europe and the United States, there have been successive cases where such structures are said to have had a concrete impact on the electoral process. In the 2016 US presidential election, false information claiming that the Pope had endorsed a specific candidate spread, leading the Vatican to later issue a denial. It has also been pointed out that disinformation influenced decision-making in the UK's Brexit referendum. In the 2017 French presidential election, disinformation regarding candidate Macron's use of tax havens spread, and in Germany, false associations capitalizing on migrant-related incidents were disseminated. These cases demonstrate that the structural characteristics of SNS facilitate such spread, and the possibility of information manipulation, including involvement by foreign governments, is a reality. Furthermore, the existence of behavioral targeting by platforms cannot be ignored as a foundation that increases the spreading power of disinformation. The Cambridge Analytica case in the 2016 US presidential election is symbolic of this. It is said that political advertisements were delivered to exploit psychological vulnerabilities based on leaked user data. This event garnered global attention and led to voluntary measures being taken by platform operators.

On the other hand, there is a general perception in Japan that situations as serious as those in Europe and the US have not been confirmed. However, there have been cases where disinformation became an issue during elections; in the 2018 Okinawa gubernatorial election, information of unknown authenticity criticizing a candidate spread from specific sites. Even if large-scale intervention by foreign governments is not prominent, the risks of domestic division and manipulation of candidate evaluations cannot be overlooked. Moreover, as the number of voters using SNS as their primary information source has increased in recent years, there is a high possibility that situations similar to those in Europe and the US could occur.

In 2024, candidates and political parties that expanded their support based on SNS were prominent, further clarifying the character of SNS as the main battlefield for information distribution. Among these, the Hyogo gubernatorial election was suggestive in that unverified claims regarding candidates spread on YouTube and other SNS, and "clipped videos" accelerated the speed of dissemination. During election periods, when news organizations tend to suppress the volume of reporting out of consideration for fairness, a so-called "news vacuum" is likely to occur, making it easier for content with insufficient sources or verification to gain visibility. It has been pointed out that this vacuum is likely a factor behind the increased presence of information on SNS.

The author's research team collected and analyzed YouTube videos containing election-related keywords via API, covering the period from the announcement date to the ballot counting day of the 2024 Hyogo gubernatorial election and the House of Representatives election. As a result, it may be surprising that long-form videos, rather than shorts (60 seconds or less), occupied the top rankings in terms of view counts. Particularly in the Hyogo gubernatorial election, official channels of specific candidates and popular political channels showed strong influence, and views of the long-form videos they released grew, while official channels of traditional media struggled to gain views. This suggests that rather than reacting reflexively to short, provocative fragments, viewers were taking the time to consume information that matched their own interests.

There is also a flip side to this active choice by users. Users tend to stay within channels and similar content videos that fit their interests, and as a result, opportunities to encounter differing opinions are narrowed. The formation of filter bubbles and echo chambers leads to the solidification of conflict and the amplification of division. Additionally, the increase in long-form video viewing indicates the possibility that it alternatively filled the reporting vacuum, but if this is linked to the spread of content with a high verification burden, it will create a different kind of vulnerability. Furthermore, even on the same YouTube platform, while traditional media videos gained a certain amount of viewership in the House of Representatives election, this trend was weaker in the Hyogo gubernatorial election. Differences in the nature of the election, regional context, participants, and topic setting may have changed the aspects of distribution through algorithms and viewing behavior.

"Community Notes," implemented on X, is a system where X users can collaborate to provide background information for posts that may be misleading. Contributors can leave comments on any post, and when a sufficient number of contributors from various perspectives rate the comment as "helpful," that comment is displayed on the post. It is expected to play a role in fact-checking uncertain information. Elon Musk of X describes it as "the most trusted source of information on the internet," but on the other hand, there is persistent criticism regarding the reduction of human content moderation and an over-reliance on algorithmic solutions. In fact, some research results suggest that in the US presidential election, it has not been able to counter false or misleading claims, and that X's Community Notes is not functioning sufficiently.

According to an analysis conducted by the author's research team targeting the 2024 House of Representatives election, the most viewed posts with notes were related to the method of filling out national reviews, immigration policy, and vaccine-related claims, suggesting that they fulfilled a certain function as corrections or supplements to obvious errors. At the same time, however, it was confirmed that many users who post a large number of Community Notes tend to add annotations biased toward specific political parties or claims. Furthermore, many of the top-viewed notes took the form of partisan debates by addressing at least two or more political parties simultaneously. As a result, Community Notes functioned more as a mirror visualizing partisan conflict than as the ideal image of neutral fact-checking.

The use of generative AI also drew attention in elections from 2024 onwards. Because generative AI realizes video editing and text generation quickly, at low cost, and easily, it significantly increases the volume and speed of information distribution. While the prevailing view at this stage is that an impact fundamental enough to change elections as a whole has not been confirmed, it harbors serious risks in the medium to long term. In particular, the spread of deepfake technology has the potential to make it easy to generate and disseminate false information that damages the credibility of candidates.

Even in SNS posts related to the 2024 House of Representatives election, the existence of generated videos making it appear as if prominent politicians support specific policies has been confirmed. The problem is not just the presence or absence of fake videos. It lies in the fact that the automation of editing and summarization accelerates the reproduction of messages, distributing them at a speed that outpaces verification. Verification from a medium- to long-term perspective, such as the impact on the cognition of recipients including young people and the cumulative effect of political distrust, is essential.

The challenges surrounding elections and SNS are multi-layered issues where elements such as the spread of disinformation, social division, behavioral targeting, and generative AI are intricately intertwined. While situations as serious as those in Europe and the US have not yet occurred in Japan, cases of information distribution on SNS during elections from 2024 onwards show the reality of future risks. While SNS carries the danger of threatening democracy, it can also serve as an opportunity for voters to encounter diverse information and encourage political participation. What is important is the construction of a framework for multi-faceted efforts based on this duality. It is necessary to advance considerations through dialogue among diverse stakeholders, such as the promotion of fact-checking, improving the transparency of advertisements and recommendations by platform operators, and examining the necessity of information literacy education and regulations for users. At the same time, each individual user must also examine the information they encounter and cultivate an attitude of having multi-faceted perspectives. With the spread of generative AI, the information environment will become even more complex, and the speed of information distribution will accelerate. That is precisely why a system for ensuring the health of democracy from a medium- to long-term perspective, without being misled by short-term phenomena, is required. Connecting the power of SNS to participation and deliberation rather than division is a challenge for the future.

The author's research team is currently analyzing the dynamics of video posting and viewing on YouTube targeting the 2025 House of Councillors election. In comparison with 2024, changes in posting strategies and content composition are being observed, suggesting that the style of information distribution is not identical for every election and that the distribution channels themselves transform according to technological progress and social conditions. Therefore, it is vital to avoid generalization based on a single case and to accumulate empirical knowledge that supports the health of democracy through continuous data collection and comparative analysis.
Annotation text text - Annotation This is a text part displayed in a lighter color compared to normal text. Add 20 / 1000 blocks

* In the analysis by the author's research team introduced in this article, Mr. Hibiki Sumioku and Ms. Tomoka Nakazato provided great assistance in tasks including data analysis. I would like to express my deep gratitude here.