Why does twitter overload

The order of follow shows a significant effect on unfollowing, suggesting that people are less likely to unfollow the users who have connected for a relatively long time. In this way, users re-examine their recent followees and decide whether to keep them in their information repertoires. Finally, the repertoire becomes more and more stable.

An alternative explanation is that the older followees might imply strong ties, thus are less likely to be unfollowed. Figure 2 shows that even we controlled for the interaction frequency and other variables, the negative relationship between the order of follow and unfollowing probability still holds. The result implies users are intentionally stabilizing their information repertoires over time. The observed and estimated probabilities of unfollowing as a function of the order of follow.

The estimated probability was calculated based on the full model in Table 1. The major concern of the current study is to examine the role of informational factors in building personal repertoires of information sources. H1 stated that people are more likely to unfollow when they receive overloaded information.

Therefore, H1 is supported. H2 stated that users are less likely to unfollow the users sharing similar hashtags. The full model shows that hashtag similarity is negatively associated with unfollowing. Therefore, H2 is supported. H3 stated that hashtag similarity moderates the impact of information overload on unfollowing other users.

Figure 3 illustrates that the positive association between information overload and unfollowing is stronger when hashtag similarity between the ego and alter is low. From Figure 3 , we also note that the difference of unfollowing probability between high and low similarity increases exponentially as information overload increases, indicating that information overload reinforces similarity-based selection.

Therefore, H3 is supported. The interaction effect of information overload and similarity on unfollowing. H4 stated that users are more likely to unfollow the users whose tweets included many redundant hashtags. Therefore, H4 is supported. Concerning H5 and H6, the full model suggests that both interaction effects are not significant at all. It suggests that the redundancy effect is not conditional on information overload and similarity when relational factors are controlled for.

Therefore, H5 and H6 are not supported. According to the full model in Table 1 , all relational factors show significant impacts on unfollowing behavior. Ego users are less likely to unfollow users with more followers. To answer the RQ, Model 1 in Table 1 excluded all relational factors. The inconsistency between Model 1 and the full model is caused by the exclusion of relational variables. First, the redundancy effect is no longer significant in Model 1.

It indicates that relational factors are suppressors. In our data, information redundancy is positively correlated with reciprocity and the number of shared followees.

For reciprocal ties, the followees' information redundancy is 0. The Spearman's rank correlation between the number of followees and information redundancy is 0. This means that high information redundancy implies dense connections i. In addition, without considering the relational factors, the interaction effects with hashtag redundancy are significant in Model 1. The model suggests that users are less likely to unfollow the alters with redundant hashtags when information overload is high.

It implies that users prefer information redundancy, which was produced by the relational factors. This study conceptualized social media followees as information source repertoires and examined the dynamics of repertoire formation using panel data from Twitter.

First, this study suggests that users maintain relatively stable information repertoires to cope with information overload. During the 3 months of observation, only 5. Despite that, our findings suggest that some users actively and continuously adjust their information source repertoires over time.

It is consistent with previous research Kwak et al. It implies that users are intentionally stabilizing their personal repertoires for daily information other than receiving it passively. In our dataset, nearly two-thirds of users did not unfollow any users during our observations. This indicates that unfollowing actually is not a popular behavior on Twitter. However, it does not mean that unfollowing is a rare phenomenon or it lacks theoretical significance.

We tracked the unfollowing behavior in a relatively short period of time. Instead of browsing all information channels, users would like to check information from a few sources repeatedly. Second, this study extended the repertoire approach by examining the role of information overload, similarity, and redundancy in structuring information consumption patterns on a single social media platform. We found that seeking information similarity and reducing information redundancy could coexist in the process of optimizing information repertoires.

One popular argument states that users are increasingly seeking content similar sources on social media. Following this tendency, individuals would like to consume a steady diet of their preferred type of information sources.

Finally, users with similar interests will cluster together Himelboim et al. The current study indeed found that Twitter users are more inclined to keep those followees sharing similar hashtags.

Under the information overload situations, the tendency of selecting content similar alters is reinforced see Figure 3. The reason is that, as Table 1 suggests, people intentionally unfollowed the users with redundant information, even though they kept the similar alters at the same time.

As a balance, their information repertoires contain the messages they are interested in and with very little redundancy. This also implies that people do have diverse interests and try to sample a diverse range of sources to build their information repertoires. Third, we note that the formation process is significantly constrained by relational factors i.

In addition to their direct effects on unfollowing, the relational factors can alter the impacts of informational factors. We found that relational variables are suppressors of the redundancy effect.

This implies that some users received unexpected and redundant information from their networked users. In addition, we hypothesized that the informational effects are conditional on each other. However, our results suggest that the redundancy effect is not dependent on information overload and similarity when the relational factors are controlled for. Furthermore, although we focused on the information variables in building information repertoires, it does not mean that alternative explanations are impossible.

On the contrary, our study is consistent with previous repertoire studies that the structural factors are more important that other factors see Webster, The structural factors in the present study include the relational variables that characterize the online social networks and the control variables. For example, the low ratio of the number of followers to the number of followees popularity indicates that the users are inclined to keep more information sources.

Following many sources may suggest the users' availability in viewing new messages. However, these variables are at the microlevel or mesolevel in general. Future studies can explore the impacts of more macrolevel variables on the unfollowing behavior.

As suggested by Webster , the aggregate network level analysis would be beneficial to understand the bounded rationality of online user behaviors. Several limitations can be associated with this study. First, when considering followees as information repertoire, we assume that users actually only read the messages posted by their followees.

This assumption might not be accurate. Users can simply ignore the messages that they are not interested in to reduce information overload Savolainen, In addition, users can receive messages beyond their immediate following networks. Social networks are not the only mechanisms through which users are directed to media.

The recommender system and search engine are commonly used for direct audience attention on social media platforms Webster, However, the following relationships do indicate awareness of the presence of the followees Himelboim et al. Future studies can track users' browsing history on social websites to examine patterns of consuming specific messages other than sources. Second, Twitter provides researchers with the unique opportunity to track patterns of individual selection of information sources.

Although, the unobtrusive approach provides more objective measures, it lacks information on both demographic and psychological variables. Previous studies have found that demographic variables, such as gender and age, show significant impact on the composition of media repertoires e. Future studies need to further control these variables and examine the interaction effects between the self-reported and objective measures employed to build information repertoires.

In addition, the unobtrusive approach can cause potential measurement errors. For example, using tweeting frequency to measure information overload might be problematic. Even receiving the same amount of messages, some users may perceive more overload than would other users. We could not measure this subjective feeling directly. Instead, we employed the multilevel framework to control this individual heterogeneity carefully.

First, the impact of tweeting frequency on the probability of being unfollowed by egos was considered separately for each ego. Furthermore, we included the potential compounding variables to control the individual differences. For example, Table 1 suggests that egos with more followees actually are less likely to unfollow other users, indicating that those users may have a higher threshold of information overload. We measured information similarity and redundancy based on hashtags. This kind of operationalization was based on the repertoire approach to studying the user-defined channel types.

Another way to measure information similarity and redundancy is to calculate the variables based on the raw text. However, we think they are conceptually different things. The purpose of the current study is to demonstrate that the user's choice of information sources is based on content topics other than using similar or unique words.

As a robustness check, we conducted a post hoc analysis based on the raw text measures see the last column in Table 1 : Raw Text Measures. We found that the main effects are similar, whereas the interaction effects are slightly different.

Future studies should use more advanced techniques to detect user-defined categories, such as the topic modeling approach, which is similar to factor analysis in media repertoire studies see Weng et al. Finally, social media platforms emphasize different technological characteristics. Our results rely on Twitter, which puts a greater emphasis on news sharing.

For other social media platforms, like Facebook, studies may emphasize social networking. In this sense, users might be less susceptible to the information variables than was the case in our study. Furthermore, previous repertoire studies have demonstrated that people could build their personal repertoires across media platforms or rely on one of them. The choice of different repertoires is associated with user background characteristics e.

For studies based on a single platform, it is difficult to capture more general media use patterns. For example, watching TV news intensively may cause information overload or redundancy on Twitter. Therefore, future studies are encouraged to test our hypotheses across different social media platforms.

Barabasi , A. Emergence of scaling in random networks. Science , , — Google Scholar. Chaffee , S. Mass media and interpersonal channels: Competitive, convergent, or complementary. Cathcart Eds. Google Preview. Dutta-Bergman , M. Complementarity in consumption of news types across traditional and new media. Farhoomand , A. Managerial information overload. Communications of the ACM , 45 10 , — Ferguson , D.

The World Wide Web as a functional alternative to television. Franz , H. The impact of computer mediated communication on information overload in distributed teams. Golder , S. Structural predictors of tie formation in twitter: Transitivity and mutuality. Harrigan , N. Influentials, novelty, and social contagion: The viral power of average friends, close communities, and old news.

Social Networks , 34 4 , — Hermida , A. Share, like, recommend: Decoding the social media news consumer. Journalism Studies , 13 5—6 , — Himelboim , I. Playing the same Twitter network: Political information seeking in the US gubernatorial elections. Holton , A. News and the overloaded consumer: Factors influencing information overload among news consumers.

Cyberpsychology Behavior and Social Networking , 15 11 , — Jenkins , H. Convergence culture: Where old and new media collide. Kim , S. A repertoire approach to cross-platform media use behavior. Kivran-Swaine , F. The impact of network structure on breaking ties in online social networks: Unfollowing on Twitter.

Kwak , H. Fragile online relationship: A first look at unfollow dynamics in twitter. What is Twitter, a social network or a news media? Paper presented at the Proceedings of the 19th international conference on World wide web. More of a receiver than a giver: Why do people unfollow in Twitter?

Liang , H. Testing propositions derived from Twitter studies: Generalization and replication in computational social science. Plos One. Marwick , A. Using the Twitter. You can choose from many different apps to track your tweets. Good list experiences are a little slimmer on the Mac. Muting Twitter users or topics is a great way to get a temporary reprieve from a chatty tweeter or a topic you want to avoid.

Whatever the reason, some Twitter clients offer the option to mute things like users, hashtags, or even clients. Generally, mute options are offered in increments of time such as a day, a week, a month, or forever. So yes: if you really need to, you can give in to that one really persistent friend begging for a follow, but never have to actually see his or her tweets—as long as you use an app that supports the mute feature.

Who says you have to follow anyone at all? Plenty of people use Twitter as nothing but a research tool in addition to, or in lieu of, their personal aspirations for tweetness.

Enter a keyword like iPhone in the search box at Twitter. Then, click the gear menu next to the search field to save that query for easy access later. Depending on what version of Twitter. Being online, or worse, Very Online, can often feel indistinguishable from descending into madness.

Our brains simply cannot have been designed to withstand such a constant onslaught of conflicting information at once. I thought about this after reading a particularly devastating story about an Iraqi teenager whose parents were killed a decade ago by U. It was one of those stories that soaks into your mind like spilled ink, and the way she spoke of vengeance haunted me.

But immediately after reading, I went back to Twitter and came across a prominent young Republican dork being roasted for a selfie. Naturally, I joined in. Sexual assault, football highlight, dog picture, mass shooting, around and around in an unending circle. But for the next hour or so, those two contrasting concepts jostled back and forth in my brain for primacy.

War horror. Goofy college conservative. This is, of course, the nature of social media—stacking the incongruous on top of one another in theoretically manageable information nuggets. And to make things worse, many of us actually do this to ourselves, by choice , every single day. Who knows what long term effects that sort of see-saw of emotional overload will have? Turns out, nobody does, as a number of psychologists I spoke to explained.