The Structure of Political Discussion Networks
- 1. The Structure and Dynamics of Th S dD i f Online Political Discussions Online Political Discussions Sandra Gonzalez‐Bailon sandra.gonzalezbailon@oii.ox.ac.uk d l b il ii k
- 2. The Questions We Will Consider Today: • Do online political discussions promote deliberation? (focus: structure of discussion threads) deliberation? (focus: structure of discussion threads) Gonzalez‐Bailon, Kaltenbrunner and Banchs (2010) “The Structure of Political Discussion Networks: A Model for the Analysis of Online Deliberation“, Di i N t k A M d l f th A l i f O li D lib ti “ Journal of Information Technology, 25 (2): 230‐243 • Can we use online political discussions to track p public opinion? (focus: content of discussions) p (f f ) Gonzalez‐Bailon, Kaltenbrunner and Banchs (2010) “Emotional Reactions and the Pulse of Public Opinion: Measuring the Impact of Political Events on the the Pulse of Public Opinion: Measuring the Impact of Political Events on the Sentiment of Online Discussions”, under review
- 4. The Logic of Deliberation • Based on social interactions, not on private decisions • Superior form of collective decision making: () (i) Decisions are more informed (plurality of views) (p y ) (ii) Decisions are more legitimate (all voices are represented)
- 5. The Logic of Deliberation • Based on social interactions, not on private decisions • Superior form of collective decision making: () (i) Decisions are more informed (plurality of views) (p y ) (ii) Decisions are more legitimate (all voices are represented) Deliberation requires: D lib i i Multiple points of view and conflictive arguments Multiple points of view and conflictive arguments Do political discussions involve more points of view and p p f more argumentation?
- 7. Data: September 2005 to August 2006 # posts (N=10016) # comments (N=2075085) yro yro science science politics politics linux linux it it interviews interviews hardware hardware games games features features developers developers bsd bsd books books backslash backslash ask ask apple apple Thousands Hundreds apache apache 0 50 100 150 200 250 300 350 0 5 10 15 20 25 # users (N=93636) yro science politics linux it interviews hardware games features developers bsd books backslash ask apple Thousands apache 0 10 20 30 40 50
- 8. Discussion Threads as Hierarchical Networks Post Layer 1 Comment Comment Layer 2 Comment C t Comment C t Comment C t Layer 3 L Comment C Comment C Comment C Comment C
- 9. Discussion Threads as Hierarchical Networks Post Layer 1 Comment Comment Layer 2 Comment C t Comment C t Comment C t depth d th Layer 3 L Comment C Comment C Comment C Comment C width
- 10. Four Types of Discussion max Type II Type I T tion) depth (argumentat Type III Type IV min max width (representation) ( i )
- 12. The Empirical Measurement of Deliberation p max Type II T Type I T tation) dept (argument Type III Type IV th min max width (representation)
- 13. Classification of Discussions in Possibility Space max Type II Type I ation) dep (argumenta Type III Type IV pth min max width (representation) ( t ti )
- 14. Additional Measures of Width and Depth Width Depth Max # comments # nested levels Total # comments H‐index users Total # use s # users H‐index comments H i dex co e ts
- 15. Classification of Discussions in Possibility Space max Type II Type I ation) dep (argumenta Type III Type IV pth min max width (representation) ( t ti )
- 16. Conclusions Do political discussions involve more points of view and more argumentation? and more argumentation? Yes. They reproduce important pre‐conditions for Yes They reproduce important pre conditions for deliberation to take place. Future research Comparative analysis: do other online forums promote the same type of dynamics? promote the same type of dynamics?
- 17. Paper 2: Analysing Public Opinion p y g p
- 18. The Importance of Public Opinion p p • Public opinion: proxy to the way citizens u i opi io p o y o e ay i i e perceive political issues and react to current affairs •S Surveys and Polls – mechanism behind d P ll h i b hi d opinion tides? (focus: emotions) What explains polarisation and volatility? p p y
- 19. Measuring the Emotional Content of Discussions g Valence Arousal A l Dominance Bradley and Lang. 1999. ʺAffective norms for English words (ANEW) Technical report C‐1.ʺ The Center for Research in Psychophysiology, University of Florida, Gainesville, FL. y p y gy, y , , Dodds and Danforth. 2009. ʺMeasuring the Happiness of Large‐Scale Written Expression: Songs, Blogs, and Presidents.ʺ Journal of Happiness Studies DOI: 10.1007/s10902‐009‐9150‐9.
- 20. Data: Usenet Political Discussions (1999‐2005) x 10000 8 6 4 2 0 09/1999 09/2000 09/2001 09/2002 09/2003 09/2004 gun white war war world war war white black gun white news world gun white people news white news time world good time news war news terrorist peace gun death world time people o d people people l news time people hate house dead peoplegood black terrorist party hate dead hatewhite socialdead hate world war house good f ree man truth lie man house good man party free black death hate party death house party deathgood mancrime time man dead black free black lie god death truth gun torture fraud win free house time house peace truth free letter god terrorist t i t gun bl k black money boy abortion flag party world good cut death power hate man fraud free truthcrime
- 27. Conclusions • Emotions help predict public opinion E o io e p p e i pu i opi io (politics is as much about passions as about reasons) •A Arousal is the most significant emotional l i th t i ifi t ti l dimension – does this also apply to other time windows?
- 28. Future Research • Combine these two approaches: Do emotions o i e e e o app oa e o e o io help explain the growth of discussion threads? • Validate approach across platforms and linguistic communities li i ti iti