Virality Prediction and Community Structure in Social Networks

Transcription

1 Virality Prediction and Community Structure in Social Networks Lilian Weng, Filippo Menczer, Yong-Yeol Ahn Center for Complex Networks and Systems Research (CNetS) School of Informatics and Computing

2 Uploaded to Youtube on 2012 July 15 More than 1.38 billions views now!

3 Uploaded to Youtube on 2012 July 15 More than 1.38 billions views now! World population 7.07 billions

4 We are pleased to inform you that your paper has been accepted! As a PhD student: 4 months 2.4k FB likes 71 notes As a post doc: As a professor: by Nikolaj & Jilles

5 Corporations Government Political Campaigns

6

7 Virality Attention

8 Viral information spread through social networks

9 Epidemic Spreading: Germs and viruses spread through the social network

10 Ideas and behaviors also spread

11 Are they same?

12 Aha! They are Different

13 Social Reinforcement Multiple exposures

14 Social Reinforcement Multiple exposures

15 Social Reinforcement Multiple exposures

16 Social Reinforcement Multiple exposures

17 Social Reinforcement Multiple exposures

18 Which network is better at spreading information quickly? Large world Small world D. Centola, Science 2010

19 D. Centola, Science 2010 Information spread more quickly on large world network

20 D. Centola, Science 2010 Information spread more quickly on large world network Easier to have Multiple exposure

21 D. Centola, Science 2010 Information spread more quickly on large world network

22 D. Centola, Science 2010 Information spread more quickly on large world network Harder to have Multiple exposure

23 Epidemic Spreading: Germs spread through the social network Simple Contagions

24 Ideas and behaviors also spread Complex Contagions D. Centola and M. Macy. Complex Contagions and the Weakness of Long Ties. AJS, 2007.

25 (Node) Users

26 (Edge) Social Relationship

27 (Community) Social Circles

28 (Community) Social Circles

29 How do communities affect information diffusion?

30 Structural Trapping

31 Traps for random walkers

32 Multiple Exposures Multiple Exposures A A B A B High Clustering Low Clustering Social Reinforcement

33 Homophily

34 (1) Structural Trapping (2) Social Reinforcement (3) Homophily More communication within than across communities; Communities trap information.

35

36 500 million users 340 million tweets per day

37 Gardenhose (10%)

38 Tweet Short messages Follow Subscribe users Retweet Spread messages Hashtag Topic identifier, i.e. #ows

39 [1] Richard Dawkins. The Selfish Gene Hashtag ~ Meme [1]

40 [1] Richard Dawkins. The Selfish Gene Hashtag ~ Meme [1] truthy.indiana.edu

41 Two community detection methods Disjoint Communities Infomap (Rosvall & Bergstrom, 2008) Overlapping Communities Link clustering (Ahn, Bagrow, Lehmann, 2010)

42 English #usa BBC News #bbc Arabic #usa Fox News #foxnews Retweet Network Follower Network Figure 1: The importance of community structure in the spreading of social contagions. (A) Structural trapping: dense communities with few outgoing links naturally trap information flow. (B) Social reinforcement: people who have adopted a meme (black nodes) trigger multiple exposures to others (red nodes). In the presence of high clustering, any additional adoption is likely to produce more multiple exposures than in the case of low clustering, inducing cascades of additional adoptions. (C) Homophily: people in the same community (same

43 Do the edges inside communities transmit more information?

44 For each community, we measure the average edge weights of intra- and inter-community links. (A) w RT w RT

45 For each user, we measure fraction of activity that is directed to each neighbor in the same or different community. (B) f RT f RT

46 Is the communication concentrated inside communities?

47 No concentration Randomly distributed

48 Weak concentration Randomly diffusion

49 Strong Concentration Distributed more in a few communities

50 Null models M 1 Random selection M 2 Random diffusion (structural trapping)

51 Proportion of #tweets in dominant community Relative Usage Dominance (C) Real Total #tweets

52 Entropy of #tweets distributed in different communities Relative Usage Entropy (D) Total #tweets

53 Viral memes are less trapped by communities, spreading like diseases.

54 (A) Structural Trapping Multiple Exposures (B) Social Reinforcement Multiple Exposures (C) Homophily A High Clustering Low Clustering Just this Other memes Viral memes

55 (A) Structural Trapping Multiple Exposures (B) Social Reinforcement Multiple Exposures (C) Homophily A High Clustering Low Clustering Just this Other memes Viral memes Complex Contagions Simple Contagions

56 Avg. #exposures required for each adopters Total #tweets

57 No enough data Complex Contagions Simple Contagions

58 Maybe...We can do virality prediction by qualifying concentration?

59 Old New Less dominant More dominant (A) #ThoughtsDuringSchool 30 tweets Early Stage Late Stage (B) #ProperBand 30 tweets Early Stage Late Stage

60 Old New Less dominant More dominant (A) #ThoughtsDuringSchool 200 tweets 30 tweets Early Stage Late Stage (B) #ProperBand 30 tweets 65 tweets Early Stage Late Stage

61 Random guess Community-blind prediction Community-based prediction 0.8 (A) (C) Precision Recall T T Viral memes T tweets.

62 Summary Communities give us invaluable information about spreading patterns of memes. We can predict viral memes by looking at communities Non-viral memes seems to act like complex contagions, strongly affected by social reinforcement and homophily, while viral memes are not. Viral memes spread like (literally) epidemics.

63 Thanks! Questions? Lilian Weng Fil Menczer YY Ahn {weng, fil,