Source-linked AI summary
Who creates trends in online social media: The crowd or opinion leaders?
Leihan Zhang, Jichang Zhao, Ke Xu
TL;DR
The paper investigates whether the crowd or opinion leaders create trends in online social media. The study focuses on long-term analysis of how the popularities of words from 2013 vary over time. The crowd plays a decisive role early in trend creation, while early opinion-leader dominance produces small-scale coverage and ordinary-user participation can boost spread to broad coverage.
Problem
The paper investigates whether the crowd or opinion leaders create trends in online social media.
Source: §III · p. 11Method
The study focuses on long-term analysis of how the popularities of words from 2013 vary over time.
Source: §I · p. 4, §III · p. 11Results
The crowd plays a decisive role early in trend creation, while early opinion-leader dominance produces small-scale coverage and ordinary-user participation can boost spread to broad coverage.
Source: §III · p. 11Takeaways & Limitations
These findings challenge the influentials hypothesis by distinguishing local diffusion from broad trend formation.
Source: §III · p. 11The authors identify comparison of different diffusions as a limitation, but the specific constraint is not stated.
Source: §III · p. 12
Abstract
from arXiv · showhide
Trends in online social media always reflect the collective attention of a vast number of individuals across the network. For example, Internet slang words can be ubiquitous because of social memes and online contagions in an extremely short period. From Weibo, a Twitter-like service in China, we find that the adoption of popular Internet slang words experiences two peaks in its temporal evolution, in which the former is relatively much lower than the latter. This interesting phenomenon in fact provides a decent window to disclose essential factors that drive the massive diffusion underlying trends in online social media. Specifically, the in-depth comparison between diffusions represented by different peaks suggests that more attention from the crowd at early stage of the propagation produces large-scale coverage, while the dominant participation of opinion leaders at the early stage just leads to popularity of small scope. Our results quantificationally challenge the conventional hypothesis of influentials. And the implications of these novel findings for marketing practice and influence maximization in social networks are also discussed.
I. INTRODUCTION
Prior work studies temporal or spatial trend dynamics, but leaves the roles of different users in popularity formation underexamined. The authors connect user roles to popularity’s life cycle and compare the crowd with opinion leaders as trend creators.
I. INTRODUCTION
- Earlier work focused on temporal or spatial dynamics, leaving user roles in popularity formation and their connection to long-term popularity underexamined.Source: §I · pp. 3–4
- Whether traditional two-step flow applies in online social networks remains unresolved, and empirical validation of the influentials assumption is missing.Source: §I · pp. 3–4
- The 42 selected 2013 Weibo slang words each show two adoption peaks, with the first much lower than the second.The study tracks their popularity over time.Source: §I · p. 4
- The two peaks represent separate opportunities for collective attention and trend formation, making their diffusion differences a way to examine trend creation.The authors propose comparing the diffusion represented by each peak to investigate why one forms a trend.Source: §I · p. 4
II. RESULTS
The results examine early participation in slang-word diffusion and its association with the scope of online trends. They compare patterns among early adopters, including follower counts, verification status, retweeting, and geography.
II. RESULTS
- In successful p2, users with fewer than 200 followers make up a higher share of early adopters than in failed p1; leader-dominated starts do not ensure broad diffusion.At the peaks themselves ([−0.1, 0.1]), participant follower-count distributions do not differ between p1 and p2.Source: Figure 2 · p. 6, §II · p. 6
- 232 followers is the median critical threshold across 42 slang words, with a corresponding largest CDF deviation of 0.17; below it, ordinary users are more prevalent in p2.The reported proportion of ordinary users below the threshold is 0.17 higher in successful p2 than in failed p1.Source: §II · p. 7, Figure 4 · p. 8
- In early-stage diffusion, verified-to-non-verified users were about 3:7 for small-scale p1 and below 2:8 for successful p2, suggesting non-verified users play a decisive role in attracting collective attention.Source: §II · p. 8, Figure 5 · p. 9
- Two-time retweeters occupy a higher share in successful p2's early stage, while one-time retweet dominance in p1 does not yield high popularity; repeated ordinary-user participation facilitates spread.The passages note that influential users' forwarding can reach more followers initially, but repeat participation by ordinary users helps information spread further.Source: §II · p. 9
- For early adoption in [−0.1, −0.5], the geographic comparison reports more users in hot regions including Beijing, Guangdong, and Shanghai.Source: §II · p. 9
- 4.26 for p1 versus 4.46 for p2 is the geographical-distribution entropy of early adopters, indicating that p2’s participants are more uniformly distributed across the country.The comparison characterizes p1 as small-scale and p2 as large-scale diffusion.Source: §II · p. 9
- For p1, slang words remain local popular phrases, as opinion leaders from hot regions produce diffusion with limited coverage.This contrasts with p2’s large-scale diffusion.Source: §II · p. 9
- 4.10 for p2 versus 3.72 for p1 is the geographical-distribution entropy during the peak period [−0.1, 0.1], with p2 more uniform.Figure 7 shows geographical distributions of adopters at different stages.Source: §II · p. 10, Figure 7 · p. 10
- The comparison finds that ordinary users’ participation at the early stage can help create trends, challenging the previous focus on social-network influentials.The paper describes this as an unexpected role for the crowd in creating trends.Source: §II · pp. 10–11
III. DISCUSSION
Comparing two adoption peaks, the paper argues that early crowd participation accompanies broad trend coverage, whereas opinion-leader-dominated adoption remains small. The discussion notes that the study examines participation patterns, not the intrinsic factors that bring the crowd into slang diffusion.
III. DISCUSSION
- The two temporal peaks represent different diffusions: the first reaches small scope, whereas the later spreads globally across the network.Long-term analysis of 2013 Internet slang-word popularity identifies two life-cycle peaks.Source: §III · p. 11
- Early crowd participation is presented as decisive to broad trend coverage, while opinion-leader-dominated early adoption produces only small-scale coverage.The authors contrast the limited reach of opinion leaders with the crowd's potential for global diffusion.Source: §III · pp. 11–12
- The findings qualify influentials-based seeding for viral marketing and influence maximization, calling for attention to ordinary-user participation early.The discussion contrasts this with the typical approach of mining influentials to seed diffusion.Source: §III · p. 11
- The discussion recommends attention to early ordinary users with fewer than 232 followers, rather than relying solely on influential users to seed diffusion.The authors connect this group to the definition of Dunbar's number.Source: §III · p. 11
- Opinion leaders, described as overloaded hubs, are contrasted with ordinary users whose interactions are called more reliable and diffusive.The discussion links focusing only on influential users with local, limited coverage.Source: §III · p. 12
- The analysis focuses on crowd and leader occupations, leaving intrinsic participation drivers unresolved, including slang features, exogenous events, and living environment.The authors note that about one third of first peaks occurred on the May Day holiday.Source: §III · p. 12
- The authors call for further exploration of crowd-participation factors to inform strategies for attracting early adoption.They identify appropriate real-world strategies for attracting crowd adoption as a future need.Source: §III · p. 12
IV. MATERIALS AND METHODS
The study analyzes Weibo tweets and popular Internet slang words from 2013, then identifies two adoption peaks using a popularity-based rule.
IV. MATERIALS AND METHODS
- The slang-word set comprised 42 popular terms listed by the Chinese encyclopedia baike.com for 2013.The dataset was made publicly available to the research community.Source: §IV · p. 12
- The highest-popularity day defines the second peak, and the first is the highest-popularity day found in the specified earlier search range.That range extends leftward from the second peak to the date with popularity 0.01P2, starting from the observed period’s beginning.Source: §IV · pp. 12–13