Source-linked AI summary
Evolution of Online User Behavior During a Social Upheaval
Onur Varol, Emilio Ferrara, Christine L. Ogan, Filippo Menczer, Alessandro Flammini
TL;DR
The paper asks how the Gezi Park protest conversation on Twitter developed across locations and time, and how users’ roles and behavior changed as events unfolded. It analyzes conversation trends, user roles, and behavior alongside real-world events. The discussion spread internationally, became more democratic as influence redistributed, and users changed screen names in response to political events.
Problem
The study asks how the Gezi Park Twitter conversation varied across space and time, what roles and influence users had, and how behavior changed in response to real-world events.
Source: §1 · pp. 1–2Method
The authors analyze Twitter conversation trends across locations and time, user roles and influence, and screen-name changes alongside protest events.
Source: §4.1 · p. 5, §4.2 · p. 6, §4.3 · p. 7Results
The discussion spread worldwide, local trends formed geographic clusters, influence became more distributed as the conversation grew more democratic, and users changed screen names in response to political events.
Source: §6 · p. 9Takeaways & Limitations
The findings show that online user behavior and discussion changed alongside events in the offline protest.
Source: §6 · p. 9Whether universal communication patterns emerge across social and political topics versus sports, news, or entertainment remains open.
Source: §6 · p. 9
Abstract
from arXiv · showhide
Social media represent powerful tools of mass communication and information diffusion. They played a pivotal role during recent social uprisings and political mobilizations across the world. Here we present a study of the Gezi Park movement in Turkey through the lens of Twitter. We analyze over 2.3 million tweets produced during the 25 days of protest occurred between May and June 2013. We first characterize the spatio-temporal nature of the conversation about the Gezi Park demonstrations, showing that similarity in trends of discussion mirrors geographic cues. We then describe the characteristics of the users involved in this conversation and what roles they played. We study how roles and individual influence evolved during the period of the upheaval. This analysis reveals that the conversation becomes more democratic as events unfold, with a redistribution of influence over time in the user population. We conclude by observing how the online and offline worlds are tightly intertwined, showing that exogenous events, such as political speeches or police actions, affect social media conversations and trigger changes in individual behavior.
1. INTRODUCTION
Social media support peer-to-peer communication and have featured in political uprisings, but this study focuses on how the Gezi Park Twitter conversation and its users changed over time. It uses data-driven analysis to examine the conversation’s spatial and temporal patterns, user roles and influence, and responses to real-world events.
Social media and communication
- Social media shift information distribution from broadcasting to peer-to-peer exchange and have altered how people seek information about societal events.They also affect how people interact with peers.Source: §1 · p. 1
Social media and protest
Research aim
- The study examines Gezi Park discussion on Twitter through its spatio-temporal patterns, user roles and influence, and changes in behavior in response to real-world events.The authors describe this as the first study to explore the temporal evolution of online user roles and behaviors as a reflection of events on the ground during a social upheaval.Source: §1 · pp. 1–2
Contribution and outline
The paper contributes methods for extracting Gezi Park conversations and analyses their geographic patterns, user roles, and behavior as the protest unfolded.
Contribution and outline
- The paper presents methods for extracting topically focused Twitter conversations about the Gezi Park uprising and related trending topics.Source: Contribution and outline · p. 2
- Similarities in discussion trends across locations yielded city clusters mostly consistent with Turkey’s geopolitics.The study examines where tweets originated and which locations shared similar topics and trends.Source: Contribution and outline · p. 2
- The study finds that influence was redistributed over time, making the conversation more democratic as user roles evolved during the protest.The analysis distinguishes common users, rebroadcasters, influentials, and hidden influentials.Source: Contribution and outline · p. 2
- Online user behavior was affected by external factors, including political leaders’ speeches and police action against the protests.Source: Contribution and outline · p. 2
2. BACKGROUND ABOUT THE PROTEST
The Gezi Park protests began as a sit-in against the park’s planned redevelopment and grew into a broader movement amid police action and wider political grievances.
Protest origins and escalation
- A sit-in by about 50–100 activists began on May 28, 2013, to oppose plans to replace a central Istanbul green space with a replica barracks for luxury residences and a mall.The peaceful encampment resisted demolition until police attacked protesters with tear gas and water cannons on May 30 and again the following morning.Source: §2 · p. 2
- The protests expanded from opposition to Gezi Park’s destruction into criticism of authoritarian government practices and intrusions into private life.By the occupation’s end on June 15, demonstrations had grown to about 10,000 people on both sides of the Bosphorus and thousands more in major cities.Source: §2 · p. 2
Communication and motivations
- Protesters used social media to share plans, recruit participants, warn of police danger, locate makeshift medical help, and announce goals.A poll found that more than 90% of about 3,000 respondents said they protested because of Erdogan’s authoritarian attitude.Source: §2 · pp. 2–3
3. DATA COLLECTION
The study assembled a hashtag-defined Twitter sample and used tweet, user, location, and trend data to examine the Gezi Park conversation.
Data collection
- User metadata included screen names, follower and followee counts, self-reported locations, and GPS coordinates for tweets posted from enabled devices.Source: §3 · p. 3
- A hashtag seed-expansion procedure began with #diregeziparki, selected the top 100 co-occurring hashtags, then collected tweets containing any hashtag in the expanded set [18].The selected hashtags were divided into general-interest, local-protest, and government-support categories.Source: §3 · pp. 3–4
- 2,361,335 tweets from 855,616 users formed the Gezi Park dataset, including 1,475,494 retweets, 47,163 replies, and 43,646 geolocated tweets.The data were collected from a 10% random sample of tweets streamed in real time over 27 days, May 25–June 20, 2013; the four pre-protest days provided a baseline for user activity and interests.Source: §3 · pp. 3–4
- Among 2,361,335 tweets associated with the Gezi Park movement, 43,646 had latitude/longitude coordinates, and the study adopted this subset to examine the protest’s spatio-temporal nature.Figure 1 maps the tweets’ geographic distribution and latitude–longitude volumes, showing content production beyond Turkey in Europe, North and South America.Source: §3 · p. 4, Figure 1 · p. 4
- The researchers monitored country-level trends in Turkey and city-level trends in 12 metropolitan areas, retrieving the top 10 hashtags and phrases every 10 minutes.They also monitored worldwide trends to assess when the discussion gained global visibility.Source: §3 · p. 4
4. RESULTS
The Gezi Park Twitter conversation spread worldwide while city-level trends reflected geographic and geopolitical divisions. As events unfolded, influence shifted toward a broader user population, and users’ screen-name changes reflected reactions to political events.
4.1 Spatio-temporal cues of the conversation
- Gezi Park discussion reached users worldwide, and after Turkish, English, Spanish, and Portuguese were the most popular languages.The global map uses geolocated tweets; the language distribution is based on tweet metadata.Source: Figure 2 · p. 4, §4.1 · p. 5
- Trend similarities grouped Turkish cities into three clusters that corresponded to geographic areas and broadly matched their geopolitical profiles.The authors note that further work is needed to explain Konya’s placement among the seacoast cities.Source: §4.1 · p. 5, Figure 3 · p. 5
- Tweet, retweet, and reply volumes surged around major on-the-ground events; about 60% of observed users joined during the first few days.Hashtags emerged more steadily over a longer period.Source: §4.1 · pp. 5–6, Figure 4 · p. 6
4.2 User roles and their evolution
- Influence shifted over time as common users and rebroadcasters gained visibility, making information production more broadly distributed.Information producers remained relatively stable and comprised more than 50% of users each day.Source: §4.2 · p. 7
4.3 Online behavior and exogenous factors
- Users changed screen names in response to political leaders’ remarks, illustrating how offline events and online behavior were tightly interconnected.Examples include adopting “TC” and using nicknames cited by Erdogan for protesters.Source: §4.3 · p. 7
5. RELATED WORK
Prior studies link social media to participation, information diffusion, and collective identity during uprisings. Work on protest conversations also documents global reach and influential users, while this study emphasizes that user roles can evolve over time.
5. RELATED WORK
- Social media adoption has been associated with lower participation barriers, easier aggregation of small acts, rapid diffusion of logistics and narratives, and stronger community and collective identity.These benefits were discussed across studies of Arab Spring movements, Occupy Wall Street, and the Spanish Indignados uprisings [11] [19] [20].Source: §5 · p. 7
- Twitter discussion of the Spanish Indignados was linked to recruiting new participants and disseminating information about mass mobilization activities [25].Source: §5 · p. 8
- During the Egyptian revolution, tweet sentiment fluctuations correlated with influential users’ expressed sentiment, and users tweeting about the revolution were located both inside and outside Egypt.The Gezi Park analysis likewise reports worldwide visibility beyond the protest’s country of origin.Source: §5 · p. 8
- Prior work identified hidden influentials who foster large cascades, while this analysis shows that user roles evolve as the protest unfolds.Source: §5 · p. 8
- Occupy Wall Street discussions included localized coordination and interstate channels for collective framing, while users’ connectivity, interests, and attention did not shift from baseline activity during the protest [19] [20].That work left open whether Occupy had a lasting effect on its online participants.Source: §5 · pp. 8–9
6. CONCLUSIONS
The study analyzes Twitter discussion of the Gezi Park protest from May 25 to June 20, 2013. It finds global discussion with geographically patterned local trends, changing user roles, and collective screen-name changes in response to external provocations.
6. CONCLUSIONS
- The dataset covers May 25 to June 20, 2013, during the main events of the protests.Source: §6 · p. 9
- Discussion spread worldwide, with sustained tweeting outside Turkey and local trends forming three geographic clusters among 12 monitored cities; attention bursts largely correlated with on-the-ground events.Source: §6 · p. 9
- As the protest unfolded, four user roles—common users, rebroadcasters, influentials, and hidden influentials—evolved, and the discussion became more democratic with more influential users.Source: §6 · p. 9
- Individuals responded to external provocations such as political speeches by collectively changing screen names to reflect sobriquets attributed to them by political leaders.Source: §6 · p. 9
- Whether communication patterns are universal across social and political discussions and topics such as sports, news, and entertainment remains to be investigated.Such comparisons could help separate general communication dynamics from topic-specific patterns.Source: §6 · p. 9