This study investigates how viewers and creators on Japanese YouTube channels progress towards conspiracy theories. By categorizing channels based on ideological or financial motives and analyzing engagement metrics such as views, likes, and comments, we find that channels driven by monetization, particularly Monetized Conspiracists, promote conspiracy theories more vigorously. This indicates that financial incentives are a crucial factor in the proliferation of such content. Channels that package conspiracy theories in formats like entertainment or spirituality serve as gateways, facilitating viewers' progression towards more extreme conspiracy-laden content. Understanding these pathways is vital for crafting strategies to counteract the spread of conspiracy theories on social media.

This study investigates conspiracy theories surrounding artificial earthquakes in Japan using social media data. We analyzed the volume of posts and conducted network and time series analyses, identifying distinct periods of increased activity corresponding to earthquake events. The results indicate that the clusters formed by conspiracy theorists and debunkers are clearly separated, and there are fluctuations in the number and structure of users participating in discussions based on major earthquake incidents. While significant earthquakes triggered increased activity in both clusters, conspiracy theorists remained relatively active in propagating their theories even during quiet periods without earthquakes. Furthermore, we observed that the movement between the conspiracy theorist and debunker clusters was minimal, with differences in the types of information sources referenced by each cluster. Future analysis may elucidate the role of YouTube as an information source and highlight differences in language usage between the two clusters.

Social media has enabled users to easily share opinions and engage with like-minded individuals, but it has also increased exposure to biased information through filter bubbles and polarization. We propose Polarization Prism, a novel reflection tool that visualizes polarization on social media and reveals diverse perspectives often overlooked under polarization. The tool extracts differences in perspectives between polarized groups using natural language processing techniques, capturing multifaceted perspectives without relying on predefined axes like pro/con. Polarization Prism is designed to be immediately generated from social media posts about ongoing events, enabling news media outlets to raise awareness about opinion biases and present diverse perspectives to their readers. An online user study with 847 participants demonstrated that the tool effectively increases awareness of perspective diversity, particularly among users with higher media literacy. Our findings highlight the potential of Polarization Prism to mitigate polarization and underscore the importance of user media literacy in maximizing its effectiveness.

This study explores the effectiveness of network-based community detection methods in detecting misinformation and conspiracy theory communities on YouTube, specifically focusing on earthquake-related conspiracy theories. Given the role of YouTube as a platform for misinformation dissemination, especially during crises like earthquakes, it is essential to identify the communities formed by conspiracy theory propagators. This study compares the performance of network-based methods that leverage viewer behavior data with Natural Language Processing (NLP)-based clustering techniques that analyze video titles. By using the "conspiracy theory score" as an evaluation metric, the study quantitatively evaluates the differences in effectiveness between these methods. The results demonstrate that network-based approaches, particularly Node2Vec, are superior in identifying conspiracy theory communities and high-influence channels, achieving higher Recall and F1 score than NLP-based methods. These findings suggest that network-based methods are highly effective for detecting conspiracy theory communities, even when textual data is limited, offering valuable insights into how conspiracy theories spread within and across communities on YouTube.

Numerous forms of consumer-generated media (CGM), such as social networking services (SNS), are widely used. Their success relies on users' voluntary participation, often driven by psychological rewards like recognition and connection from reactions by other users. Furthermore, a few CGM platforms offer monetary rewards to users, serving as incentives for sharing items such as articles, images, and videos. However, users have varying preferences for monetary and psychological rewards, and the impact of monetary rewards on user behaviors and the quality of the content they post remains unclear. Hence, we propose a model that integrates some monetary reward schemes into the SNS-norms game, which is an abstraction of CGM. Subsequently, we investigate the effect of each monetary reward scheme on individual agents (users), particularly in terms of their proactivity in posting items and their quality, depending on agents' positions in a CGM network. Our experimental results suggest that these factors distinctly affect the number of postings and their quality. We believe that our findings will help CGM platformers in designing better monetary reward schemes.

For efficient policy-making, a thorough recognition of controversial topics is crucial because the cost of unmitigated controversies would be extremely high for society. However, identifying controversial topics is costly. In this paper, we proposed a framework to search for controversial topics comprehensively. We then conducted a retrospective analysis of the controversial topics of COVID-19 with data obtained via Twitter in Japan as a case study of the framework. The results show that the proposed framework can effectively detect controversial topics that reflect current reality. Controversial topics tend to be about the government, medical matters, economy, and education; moreover, the controversy score had a low correlation with the traditional indicators-scale and sentiment of the topics-which suggests that the controversy score is a potentially important indicator to be obtained. We also discussed the difference between highly controversial topics and less controversial ones despite their large scale and sentiment.

Rating platforms provide users with useful information on products or other users. However, fake ratings are sometimes generated by fraudulent users. In this paper, we tackle the task of fraudulent user detection on rating platforms. We propose an end-to-end framework based on Graph Convolutional Networks (GCNs) and expanded balance theory, which properly incorporates both the signs and directions of edges. Experimental results on four real-world datasets show that the proposed framework performs better, or even best, in most settings. In particular, this framework shows remarkable stability in inductive settings, which is associated with the detection of new fraudulent users on rating platforms. Furthermore, using expanded balance theory, we provide new insight into the behavior of users in rating networks, that fraudulent users form a faction to deal with the negative ratings from other users. The owner of a rating platform can detect fraudulent users earlier and constantly provide users with more credible information by using the proposed framework.

Information polarization has the potential of developing into various problems such as opposition conflict. In this study, we analyzed the information polarization was caused in Japan by using Twitter data during the 48th Lower House Election in 2017. We succeeded to separate the opposite opinion such as political criticism (liberal) and political supporting (conservative) by using tweet networks based on users' retweets. Furthermore, in the result of analysis focusing users who retweet the tweets, it suggests that the liberal users did not retweet conservative tweets although the conservative users retweeted a few liberal tweets. Also, users cannot even access the tweets which have a different opinion. We indicate that the relation of follower-followee causes selective contact since users' opinions are related to the followees' them by the correlation coefficient of 0.903.

Personalization, including both self-selected and pre-selected, is inevitable when tremendous amounts of media content are available. Since personalization, which is believed to encourage people to consume fewer diverse contents, can lead to fragmentation and polarization in society, exposing people to more diverse contents is important. In this paper, we investigate using web TV data what behavioral and environmental features influence users to consume more diverse contents. We tackle this problem by building a classification model that predicts whether users will consume contents that are comprised of more diverse sources and features. We measure the change of the consumed content diversity of users with the method in our previous work and design six features based on our basic research and previous studies. The result of our experiment shows that our model predicts with high accuracy the change of the consumed content diversity of users, which means that the increase of the diversity of the consumed content of users is explained well with our six features. Our analysis of the model shows that consuming contents across multiple genres and inadvertent exposure to various contents positively affect users and encourage them broaden their horizons. The former has especially strong influence. In addition, we demonstrate that active-seeking behaviors of users and the amount of content in their most interested areas negatively affect the increase of their consumed content diversity. Our findings in this study are expected to be applied to web media to moderate the effect of selective exposure and reduce fragmentation and polarization in society.

Information is transmitted through websites, and immediate reactions to various kinds of information are required. Hence, efforts by users to select information themselves have increased, which is fueling further improvements in recommendation services that can reduce such burdens. On the other hand, filter bubbles that only provide biased information to users are generated due to redundant recommendations. In this research, we analyzed behavioral changes prior to recommendation by clustering, and we found that user attributes and cluster contents are different among users with different behavioral changes. The proportion of users under forty and women was relatively large in the diversity-increasing group. We also proposed an article selection model in order to clarify the influence of the recommendation system on the behavioral changes. We compared our proposed model with the target data, and verified the model. Then, we evaluated the effect of the recommendation system on user behavior. Simulation results showed that diversity decreases in any case, but collaborative filtering can suppress diversity decreasing rather than non recommendation. In addition, it was found that the maximum category is easily strengthened, and that is considered to be one of the factors causing diversity decreasing, and a recommendation method that can suppress strengthening of the maximum category is effective to develop a recommendation system that can suppress the diversity decreasing. We also proposed an article selection model to clarify the influence of recommendation systems on behavioral changes. We compared our proposed model with the target data, verified it, and evaluated the effect of recommendation systems on user behavior. Our simulation results showed that diversity usually decreases, but collaborative filtering can suppress the diversity decrease more effectively than non-recommendations. We also found that the category that users are interested in the most is easily strengthened and is one factor that leads to less diversity, and a recommendation method that can suppress the strengthening of the category that users are interested in the most will be effective for developing a recommendation system that can suppress diversity decreasing.

In modern election campaigns, political parties utilize social media to advertise their policies and candidates and to communicate to the electorate. In Japan's latest general election in 2017, the 48th general election for the Lower House, social media, especially Twitter, was actively used. In this paper, we analyze the users who retweeted tweets of political parties on Twitter during the election. Our aim is to clarify what kinds of users are diffusing (retweeting) tweets of political parties. The results indicate that the characteristics of retweeters of the largest ruling party (Liberal Democratic Party of Japan) and the largest opposition party (The Constitutional Democratic Party of Japan) were similar, even though the retweeters did not overlap each other. We also found that a particular opposition party (Japanese Communist Party) had quite different characteristics from other political parties.

People frequently change their music listening behaviors to fit their mood and particular situation. Such changes can be interpreted at various scales, e.g., the genres, artists, or specific tracks. Users of music streaming services expect to discover new music. However, discovering new music is not always a pleasant experience. Herein, we show that small changes in listening trends are good for users of such services. In contrast, large changes are bad. We modeled user listening trends using a hidden Markov model, which was applied hierarchically to analyze the user trends at multiple scales. We evaluated the changes in user listening trends to find good changes. Additionally, we analyzed the relationships between user listening trends and user lifestyle.

This research aims to develop a method for maximizing the accuracy of next place prediction (NPP) in places that are unfamiliar to each mobile phone users. NPP is a problem of predicting the next place of the user given his/her current place and current time. In places that are unfamiliar to the person, it is difficult to predict the next place based on the person's historical location data because there are just a few or no data in such places for each user. Furthermore, it is also difficult to rely on the regularity of human mobility because tourists' mobility is easily affected by many external factors, such as weather. Our research aims to solve the difficulties in NPP in unfamiliar places by focusing on contextual factors such as weather, transportation means, place of residence, and time.

Burst phenomena, which frequently occur on social media, are caused by such social events as flaming on the internet, elections, and natural disasters. To understand people's thoughts and feelings, we must classify their opinions from burst phenomena. Therefore, classification methods that categorize tweets are critical. However, since most classification methods focus on text mining, they cannot group tweets by topics because each tweet has poor linguistic similarities. We used a non-text-based classification method proposed by Baba et al. that groups tweets by topics, even if they have poor linguistic similarities, and verified its validity by comparing it with a text-based classification method in two different evaluations: qualitative and quantitative. In the qualitative evaluation part, we did a questionnaire survey and validated the suitability of the topic clusters created using both the non-and text-based methods. Since evaluating the similarity of every pair of tweets in each topic is difficult, we evaluated the similarity between sampled pairs in the survey and acquired more appropriate topic clustering results using the non-text-based method than the text-based method. In the quantitative evaluation part, we focused on the robustness of each method against data reduction. Many approaches analyze social media data, especially because collecting data from social media is comparatively easy. However, since collecting the whole data of burst phenomena is very costly due to the vast amounts of available social media data, robustness against data reduction is an important index to evaluate classification methods. With the non-text-based method, over 55% of the pairs of tweets in the same cluster were also included in the same cluster even when the data were reduced to 10% in all three of our example cases. In this paper, as a source we focus on Twitter, one of the most popular microblogging services. Using clustering to conduct detailed case analyses, we scrutinized three burst cases that include natural disasters and flaming on the internet and found that a non-text-based method more effectively classified tweets in burst phenomena than a text-based method.

Although it is crucial to transmit important information to those who require it during disasters, neither of the following questions have been answered: who contributes to information diffusion? How do users construct helpful relationships in social media? Unfortunately, most previous research has focused on the scale of information diffusion, instead of the flow of information and the paths traveled by the transmitted information. In this study, we calculate the information flow probability between pairs of users and experimentally verified its validity. Our results showed a maximum precision of around 75% when we consider pairs as friends whose probability scores equal 1.0. Since our method is simple, easy to compute, and needs only time-series post-shared data in social media, we believe that it can be applied to various kinds of shared information data. By applying it to disaster data, we can identify the core information diffusers and contribute to disaster mitigation in the future.

We propose a new analytical method to classify web user behavior based on such latent states of users as intention, interest, or motivation. First, we put the clickstream data of many users into a Hidden Markov Model in which the number of hidden states is large enough to build a state transition network. Since the variable hidden states represent different latent states of users, the movement on the state transition network can represent user behavior. Second, we divide each piece of clickstream data into sessions, which we classify using network movement as feature values. These cluster labels represent the latent states of users during their stay in the Web service. In this paper, we applied our method to the data of a social network game named Girl Friend BETA, which is an online game that is mainly provided on social networking services. We observed the following hidden states that represent the variable latent states of users, such as enthusiasm for the main contents of the service, playing basic content, and daily routines that are well observed by visiting the service: e.g. receiving login bonuses. Also, we classified the sessions by the latent states of users, such as light user sessions, low motivation sessions, and sessions in which users seem addicted to the main contents.

In this paper, we analyze the network structure of two SNSs, academic community system (ACS) and Amippy. From the viewpoint of network topology, the major characteristics of these data sets can be summarized as follows: low average shortest-path length, high clustering coefficient, presence of a power law degree distribution and negative assortativity. Based on our analysis, we propose a growth model of SNS networks. We conducted numerical simulations to compare actual data sets with networks generated by the proposed model. Results of simulations indicated that the processes of the CNN model and fitness model are needed to reproduce the networks of SNSs.

Recently, social networking services (SNS) have become a social phenomenon on the Internet. There are many small SNSs, including campus SNS, company SNS, and regional SNSs. Such user-limited SNSs are not fully utilized yet. To solve this problem, we propose a SNS model to simulate effective SNS encouragement methods. Simulations based on our proposed model were performed to confirm the influences of the encouragement methods on small SNSs. The simulations show that encouraging existing users' log-ins effectively encourages small SNSs.