1. Using a Model of Social Dynamics to Predict Popularity of News Kristina Lerman Tad Hogg USC Information Sciences Institute HP Labs WWW 2010

2. Outline Introduction Introduction Social News Portal Digg Social News Portal Digg Social Dynamics of Digg Social Dynamics of Digg Model-Based Prediction Model-Based Prediction Conclusions Conclusions

3. Outline Introduction Introduction Social News Portal Digg Social News Portal Digg Social Dynamics of Digg Social Dynamics of Digg Model-Based Prediction Model-Based Prediction Conclusions Conclusions

4. Introduction Popularity of content in social media is unequally distributed. Popularity of content in social media is unequally distributed. –16,000 new stories submitted to Digg everyday, while only a handful subset becomes popular Importance of popularity prediction Importance of popularity prediction –Provide users with tools to indentify interesting items –Enable social media companies to maximize revenue Studies of past researches Studies of past researches –Content quality weakly correlates with eventual popularity –Social influence is responsible for the unpredictability of popularity

5. Outline Introduction Introduction Social News Portal Digg Social News Portal Digg Social Dynamics of Digg Social Dynamics of Digg Model-Based Prediction Model-Based Prediction Conclusions Conclusions

6. User Interface of Digg Popular list Popular list –The front page –Promoted news Upcoming list Upcoming list Friends ’ Activity Friends ’ Activity

7. Inequality of Popularity Figure: Dynamics of social voting. (a) Evolution of the number of votes received by two front page stories in June 2006. (b) Distribution of popularity of 201 front page stories submitted in June 2006.

8. Outline Introduction Introduction Social News Portal Digg Social News Portal Digg Social Dynamics of Digg Social Dynamics of Digg Model-Based Prediction Model-Based Prediction Conclusions Conclusions

9. Story Data Sets May May –Submitted to Digg between May 25-27, 2006 –2152 stories, 1212 distinct users –510 stories by 239 users are promoted to the front page June June –Promoted (popular) subset 201 stories promoted between June 27-30, 2006 201 stories promoted between June 27-30, 2006 User name and time stamp of the first 216 votes for each story User name and time stamp of the first 216 votes for each story –Upcoming subset Submitted between June 30, 2006 and July 1, 2006 Submitted between June 30, 2006 and July 1, 2006 159 stories received at least 10 votes 159 stories received at least 10 votes

10. Snapshot of Social Network in Digg June June –1020 top-ranked users with their friends and fans –Augment the network in February, 2008 –Eliminate users who joined Digg after June 30, 2006 May May –Retain only the top 1020 users and their fans –Assume other users had zero fans

11. Stochastic Model of Social Dynamics in Digg Hogg and Lerman (ICWSM ’ 09) Hogg and Lerman (ICWSM ’ 09) –The stochastic processes framework relates users ’ individual choices to their aggregate behavior. –Represent user behavior in Digg as transitions between a small number of states Explanatory power Explanatory power –Why some stories accumulate many more votes than others? Predictive power Predictive power

12. Dynamical Model of Social Voting Rate equation for the number of users who vote for a story: Rate equation for the number of users who vote for a story: (vote_rate = interest * visibility) (vote_rate = interest * visibility) s(0) = S (the number of fans of the story’s submitter) N vote (0)=1

13. Model Parameters Some parameters are measured directly from the May data set. Some parameters are measured directly from the May data set. Story specific parameters Story specific parameters –r: estimated as the value that minimizes the root-mean-square (RMS) difference between the observed votes and the model predictions. –S = the number of fans of the story’s submitter

14. Observations on the Model The correlation between S and r = -0.13 The correlation between S and r = -0.13 General observations reproduced by the model General observations reproduced by the model –Slow initial growth in votes while the story is on the upcoming list –More interesting stories are promoted faster and receive more votes –A story submitted by a poorly connected user tends to need high interest to be promoted (Lerman, 2007)

15. Outline Introduction Introduction Social News Portal Digg Social News Portal Digg Social Dynamics of Digg Social Dynamics of Digg Model-Based Prediction Model-Based Prediction Conclusions Conclusions

16. Applications of the Model Estimating inherent story quality from the evolution of its observed popularity Estimating inherent story quality from the evolution of its observed popularity Predicting a story ’ s eventual popularity based on the early reaction of users to the story Predicting a story ’ s eventual popularity based on the early reaction of users to the story

17. Story Quality Estimation A wide range of interestingness to users A wide range of interestingness to users Well fit lognormal distribution Well fit lognormal distribution

18. Examples

19. Predicting Final Popularity of Stories Correlations are 0.87 and 0.49, respectively. Correlations are 0.87 and 0.49, respectively. Strong prediction in popularity rating Strong prediction in popularity rating

20. Comparison with Social Influence only Prediction Decision tree classifier based on social influence Decision tree classifier based on social influence –Two Features: 1. number of fan votes received within the first 10 votes; 2. number of submitter ’ s fans Model-based prediction outperforms the decision tree classifier Model-based prediction outperforms the decision tree classifier

21. Outline Introduction Introduction Social News Portal Digg Social News Portal Digg Social Dynamics of Digg Social Dynamics of Digg Model-based Prediction Model-based Prediction Conclusions Conclusions

22. Conclusions Research has shown that popularity is weakly related to inherent content quality, and that social influence leads to an uneven distribution of popularity, and makes it difficult to predict. Research has shown that popularity is weakly related to inherent content quality, and that social influence leads to an uneven distribution of popularity, and makes it difficult to predict. We claim that the model of social dynamics, which is developed in an earlier work, can quantitatively characterize evolution of popularity of items in Digg. We claim that the model of social dynamics, which is developed in an earlier work, can quantitatively characterize evolution of popularity of items in Digg. How interesting a story is and how connected the submitter is fully determines the evolution of the number of received votes. How interesting a story is and how connected the submitter is fully determines the evolution of the number of received votes.