Partners Data integrations for better customer experience. Product Coaching Increase satisfaction and improve product adoption with complimentary training. Upcoming Events Join us for live webinars and other events, like Khoros Engage. On-Demand Watch webinars and events on your own time. Blog Insights, tips, news, and more from our team to yours.
Customer Stories Case studies with successful customers to see how they did it. Channel integrations Integrations to connect with your customers, wherever they are Tech integrations. Developer Information Technical overviews and links to developer documentation.
Leadership Meet the team that leads the team. Newsroom Press releases and other announcements. Contact Us Need anything? Social Responsibility Our commitment to do more and do better. Highly Influential. View 4 excerpts, references methods and background. Social Networks that matter: Twitter under the Microscope.
Influence and passivity in social media. View 2 excerpts, references background. What is Twitter, a social network or a news media? Why we twitter: understanding microblogging usage and communities. View 2 excerpts, references background and methods. Beyond Microblogging: Conversation and Collaboration via Twitter.
However, there are a few differences when looking at the details. For example, Tables 4 and 5 indicate that in the Facebook network, the retweet rate increased from 0.
Furthermore, the comment rate L decreased from 0. Thus, it means that the retweet and quote tweet enhanced the cooperation, i. Finally, our experiments indicate that the cooperation and comment rates fell in the QT reward game compared to those in the RT reward game in all networks including the Facebook and the Twitter networks.
In general, the retweet rate in the QT reward game increased to that in the RT reward game; the quote tweet rate kept high in all networks, which indicates a large number of quote tweeting. In this sense, it can be treated as both posting and comment activities to some degree. In the QT reward game, it cannot be said that information provision became less active, but rather society itself was more active. This study investigated the influence of retweets on users in social media networks.
We have proposed new models by extending the conventional reward game with the introduction of the retweet and quote tweet mechanisms. In these models, an article experiences two rounds of retweets. We found that the retweet mechanism causes users to read articles posted by others who are close but unknown on the network, thereby expanding the potential readership of the article posters. Thereafter, we investigated the cooperation article posting and comment rates of the agents, which would change with the existence of retweet mechanisms.
We found that retweets could motivate agents to post new articles, and quote tweets slightly suppressed the posting activities while improving the commenting activities. In the connecting nearest neighbor networks, the cooperation rate appeared to exhibit the most significant increase when u was near 0. In the future, we plan to study the proposed RT and QT reward games by varying the costs and rewards and by implementing meta-rewards and negative rewards in our model.
Moreover, we will conduct several experiments using other real-world networks and will apply the multiple world genetic algorithm [ 23 ] to analyze the diverse strategies for individual agents. Chugh, R. Article Google Scholar. Culnan, M. MIS Q Exec 9 4 , Conway, B. Zhao, D.
Yu, A. An investigation of online social networking impacts. Ellison, N. Burke, M. Toriumi, F. Hirahara, Y. In: International Conference on Social Informatics, pp.
Osaka, K. Axelrod, R. Political Sci. Yan, Y. E 67 5 , Leskovec, J. ACM Trans. Kupavskii, A. Peng, H. Macskassy, S. Anti-homophily wins the day! Tang, X. Soc Simul. Springer Miura, Y. Download references. A preliminary version of this journal paper appeared as an article of proceedings: Benito R. The current paper proposed additional new game models that reflect not only retweets but also quote tweets and conducted experimental analyses based on more extensive experiments.
You can also search for this author in PubMed Google Scholar. All authors YY, FT, and TS conceived the idea and participated in the discussion of designing model and planning experiments and thus almost equally contributed to the work. YY mainly designed and implemented the code for experiments. TS and YY composed the draft of the manuscript. All authors read and approved the final manuscript. Correspondence to Toshiharu Sugawara. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Reprints and Permissions. Understanding how retweets influence the behaviors of social networking service users via agent-based simulation. Comput Soc Netw 8, 18 Download citation. Received : 01 March Accepted : 29 August Published : 13 September The work of [ 24 ] and [ 35 ] explore polarization and consensus within the context of opinion dynamics models.
The work of [ 33 ] show how intervention to counter the spread of disinformation, might result in an undesirable effect which is further polarizing the social network. The work of [ 16 ] provides a stochastic network model of dynamic communicators. Users have an influence value, which determines their ability to get their message broadcast by other users. The model undergoes non-polarized and polarized phases which provide an identity per user affecting the way cross group messages can then be propagated retweet effect.
The authors show that introduction of polarization into the model, enables users of high influence to accumulate even higher relative influence and consequently getting more message broadcast.
The implications are that the dialogue then becomes dominated by a smaller group of accounts rather than allowing more equally weighted exchanges. The modeling approach allows users to accumulate social capital over time as type of currency which can mirror the effect of the memory of an account by other users over time.
This way influencers can find an opportunity to use such a period of polarized discourse to capitalize upon it providing a direct motive for their activity. The role and effect of influencers on topics ranging from opinion leaders in political discussions to forming customer attitudes has been investigated [ 3 , 15 , 25 , 41 ].
Users with higher number of followers are able to broadcast their message to a bigger audience, which in turn makes their message more prone to getting retweeted. The objective for classifying users and their tweets into two tiers is to investigate the relative capability of members of the two groups to broadcast their message and acquire attention, comparing to the same quantity before the hyper-polarizing event.
This seeks to answer this question as to whether, and how, a polarizing event causes disparity among users participating in the discussion of Brexit, within the Twitter network. Reinforcing the modeling approach of the results of [ 16 ] with a data-driven investigation, the outcome is that a polarizing event such as an election, results in an increase in relative influence of users of the top tier compared to users of bottom tier.
This hypothesis is tested by investigating quantities such as Gini coefficient, retweet ratio and favorite ratio of top and bottom tiers. Also, the work of [ 1 ] puts forward the evidence that exposure to opposing views can induce polarization and in this paper we show that it is the event which causes people to see that changes affecting their views are possible which induces the polarization.
It can then be speculated that influencers could do this on smaller scale disagreements, e. The next section discusses the data used for the investigation, the subsequent section presents the key metrics used upon the data for the analysis of the topic, then the results and the description of the findings are displayed, and the final section provides a discussion with a brief summary of the outcomes.
The data used in this study arise from Twitter where the subselection is based upon the discussion surrounding Brexit. The reference to the dataset [ 5 ] Harvard Dataverse provides a collection of tweet ids and user ids on the topic of Brexit which was collected between January and October The dataset contains more than 50 million tweet ids, which can be used to retrieve the original tweet objects through the Twitter API which is valuable for inferring specifics regarding the Brexit context.
This dataset also provides a list of users whose tweets are present in the dataset along with the number of tweets each user has produced. Each user also contains an attribute field with a BotScore that is produced using a bot detector approach named BotOrNot [ 11 ]. This assigns a score in the range 0,1 as the probability that a user is a bot or not. The values closer to 1 are more supportive of the probability that the user is a bot. The stance classification task is performed using stance-indicative SI hashtags, and machine learning approaches for the users that lack stance-indicative hashtags in their tweets [ 6 ].
From the complete dataset, users were chosen, the stance was cross examined by 2 human curators independently on the topic of Brexit. The users were chosen based on the following criteria: 1 they have been active during the Brexit referendum and the UK snap election period.
The corresponding tweets were retrieved from the Twitter API. Of the chosen accounts, users were created before and the rest were created before In the end, all the tweets in the dataset which were created by these labelled users were then retrieved by Twitter API.
The corresponding dataset created includes , tweet objects over the span of 45 months. The dataset is then analyzed with the metrics discussed in Sect. Each day contains a number of active unique users. Figure 1 shows daily number of active users during Brexit referendum subfigure a and UK snap election periods subfigure b. There are dashed lines for the exact day of the election. Both periods observe an increase in the number of unique users participating in the discussion right after the events.
As discussed in Sects. Since between-group events members of opposite stance labels interacting are found to be sparse, this particular investigation shown in Fig. The dataset used for this investigation differs from the previous one in which users stance labels were manually validated after the automated procedure.
This new dataset arises from the same source [ 5 ], but the human validation is not included to verify the correctness of the automated approach. The stance attributed to each user was performed automatically and is present in the dataset, as explained in [ 6 ]. Daily number of active unique users in each day. Several metrics are utilized for analyzing the Twitter dataset described in Sect. This is discussed in more detail in Sect.
The measures described here are those used in Sect. The users in the dataset are first sorted based on their followers count number, and are subsequently divided into two tiers, with each containing an equal number of unique users. The Gini coefficient as a measure of statistical dispersion has been applied in economics to capture and measure income inequality [ 14 ].
This coefficient is calculated daily for all tweets, with respect to the number of times each tweet has collected retweets. Figure 3 compares the Gini coefficient calculated for tweets considering their retweet number wealth.
Here t is the time point for which the coefficient is calculated, e. A lower number is indicative of a more uniform distribution in comparison to a larger number representative of some disproportionate number of retweets. Here is presented the measure used for measuring the disparity between the number of retweets received by the users in different tiers in the dataset. In Fig. These two quantities are divided by each other, and the retweet ratio corresponding to that specific day is calculated.
In the end, a 3-day centered moving average is then taken. Note that in subfigures c and d of Fig. The favorite ratio is then obtained by dividing the first quantity by the second quantity:. This part explains the measure used for creating Fig. An interaction is defined as an event is which a user does one of the following: 1 quote, 2 mention, or 3 reply. In the Twitter network, these events always take place between at least two users.
Also, as explained in Sect. The second type of interaction is called cross-group interaction and is one in which the participating users belong to both sides of Brexit discussion, some belong to the remain side and some belong to leave side.
This section presents the results of applying the metrics and measures described in Sect. The overarching goal is to explore whether a key political event can induce a change in the distribution of social media exchanges between active users. Particularly to investigate if users of higher rank can benefit from times of polarized discussion in terms of being able to spread content with greater relative value than their peers and having their content favorited with larger relative values.
The data used are taken from Twitter for exchanges relevant to the Brexit vote events; the UK referendum of and the UK snap election. It was hypothesized in previous work that the events can offer an opportunity for influencers to regain relative broadcast rankings which are diluted during non-polarized exchanges.
Here the findings show increased influence during and surrounding these events. Key findings are that more users participate in the discussion during key events as hypothesized from modeling considerations , and as well during these events that there is a larger disparity between the upper tiers of influencers and lower tiers during critical political events.
The increase in activity does not maintain the same distribution of as before, but that the influencers gain not just in the quantity but also the relative impact. This alludes to a possible anticipation and capitalization of the process through dialogue choices. The results show that the capitalization does not involve influencers producing more content but receiving activity on their content by other users and those in a lower tier increasing the relative social capitol.
Figure 2 looks at the amount of relative activity of the 2 tiers of the users participating in the Twitter Brexit discussion in the timelines of the Brexit referendum and the UK snap election subfigure a and b , respectively, with the dashed lines for the voting day events. The timelines show that there is a decrease in the ratios around the event dates.
This means that there is an increase of activity from the bottom tier in relation to the group influencers and can allude to the activity being uniform, but as shown in Fig. As a result this shows that the lower tier activity is there to support and propagate the influencer content with more activity than without the polarizing discussion as discussed in [ 16 ]. Looking at the relative activity of users of the top and bottom tiers. Here the ratio of the number of tweets that users of the top tier create divided by that of the users of bottom tier is shown, for each day, measured independent of the previous days.
It is showing that the relative activity of the users of two tiers decreases during the events, meaning that the subsequent plots and results of this work are not due to higher activity of users of top tier.
0コメント