SLIDE 1
Eric Gilbert | Tony Bergstrom | Karrie Karahalios | University of Illinois
Are blogs echo chambers?
Do blogs cut readers off from dissenting opinions?
BLOGS ARE ECHO CHAMBERS: BLOGS ARE ECHO CHAMBERS Eric Gilbert | Tony - - PowerPoint PPT Presentation
BLOGS ARE ECHO CHAMBERS: BLOGS ARE ECHO CHAMBERS Eric Gilbert | Tony - - PowerPoint PPT Presentation
BLOGS ARE ECHO CHAMBERS: BLOGS ARE ECHO CHAMBERS Eric Gilbert | Tony Bergstrom | Karrie Karahalios | University of Illinois Are blogs echo chambers? Do blogs cut readers off from dissenting opinions? Negroponte 1996 Sunstein Sunstein 2002
SLIDE 2
SLIDE 3
Negroponte 1996
SLIDE 4
Sunstein 2002 Sunstein 2008
SLIDE 5
“ It’s hardly possible to overstate the value, in the present state of
human improvement, of placing human beings in contact with
- ther persons dissimilar to themselves, and with modes of
thought and action unlike those with which they are familiar.
John Stuart Mill Principles of Political Economy, 1848
”
SLIDE 6
112 M blogs
as measured by Technorati, Jun 2009
SLIDE 7
Are blogs echo chambers?
Do blogs cut readers off from dissenting opinions?
SLIDE 8
Political bloggers link to like-minded bloggers. Adamic & Glance 2005 Hargittai & Gallo 2008 Groups often adopt extreme viewpoints. Baron & Hoppe 1996 Sechrist & Stangor 2001
RELATED WORK
SLIDE 9
5 blog genres 33 top blogs 1,094 blog comments
OUR DATASET
SLIDE 10
Technology
2 TechCrunch 3 Gizmodo 4 Engadget 23 Kotaku 30 Scobelizer 35 Gigaom 37 TUAW 44 Joystiq 45 Threat Level # Technorati rank Apr 08
Political
1 Huffington Post 11 Daily Kos 26 Think Progress 41 Crooks & Liars 58 NewsBusters
Entertainment
5 Boing Boing 13 Gawker 20 Perez Hilton 31 Valleywag 36 Neatorama 42 Slashfilm
Lifestyle
6 Life Hacker 29 Consumerist 32 uthink 45 Zenhabits 38 Dooce 53 Sartorialist
Meta
10 ReadWriteWeb 18 Dosh Dosh 21 ProBlogger 27 Copyblogger 34 ShoeMoney 43 Daily Blog Tips 71 Matt Cutts
SLIDE 11
Great post and I really like the video. is is extremely similar to the approach I use in writing almost anything …
“
Just wait until hackers exploit the print layer to this mesh stuff enough to grab root and start injecting python code …
“
ProBlogger Scobelizer
SLIDE 12
Cohen’s κ = 0.71 pointwise = 0.74
Inter-rater reliability
agree neither disagree 85 103 22 46 26 2
SLIDE 13
Wald method p < 0.05
neither agree 39.2%
Proportions of agreement
11.1% disagree
49.4%
SLIDE 14
χ2(8, N=979) = 86.3 p < 0.001
Agreement proportions by genre
tech enter. lifestyle politics meta
63% neither 24% agree 13% 63% 24% 13% 44% 51% 5 40% 47% 13% 33% 55% 12%
SLIDE 15
ALGORITHMIC PREDICTION OF AGREEMENT
SLIDE 16
AGREE/DISAGREE/NEITHER
LEXICAL
uni/bi/trigams TFIDF
POS
raw tags combo lexical
SENTIMENT
congressional floor rotten tomatoes LIWC
SEMANTIC
sim to post ESA
NAMED ENTITY
- rganizations
people
SLIDE 17
Algorithmic prediction: Lexical features
This feels like an echo chamber within an echo chamber!
SLIDE 18
Algorithmic prediction: Lexical features
This feels like an echo chamber within an echo chamber!
SLIDE 19
Algorithmic prediction: POS features
This feels like an echo chamber within an echo chamber!
DT VBZ IN DT JJ NN IN DT JJ NN
SLIDE 20
Algorithmic prediction: Semantic features
Cosine similarity to post WordNet similarity to post Explicit Semantic Analysis
SLIDE 21
Algorithmic prediction: Sentiment features
LIWC positive, negative, anger, … Rotten Tomatoes +/– classifier Congressional floor +/– classifier
SLIDE 22
Algorithmic prediction: The actual algorithm
Bagged Complement Naive Bayes
Rennie & Shih. Proc. ICML, 2003.
SLIDE 23
AGREE/DISAGREE/NEITHER
LEXICAL
uni/bi/trigams TFIDF
POS
raw tags combo lexical
SENTIMENT
congressional floor rotten tomatoes LIWC
SEMANTIC
sim to post ESA
NAMED ENTITY
- rganizations
people
bagged comp. naive bayes
SLIDE 24
Learning an NLP-based model of commenter agreement
67.4%
χ2(1, N=196) = 7.6 p = 0.006
baseline model 49.4%
Model accuracy
Rotten Tomatoes classifier: 90% Congressional floor classifier: 71%
SLIDE 25
LIWC pos. emotion words agree LIWC affect words agree exclamations agree adjectives agree @ neither ellipsis !disagree great agree is tech blog neither cosine similarity to post !disagree great [noun] agree personal pronouns !disagree present tense verbs neither [prepos] [poss pronoun] agree tf-idf dot product with post !neither coordinating conjunctions agree
Features + Info Gain
0.079 0.049 0.043 0.041 0.041 0.038 0.035 0.034 0.034 0.03 0.028 0.026 0.026 0.026 0.026
SLIDE 26
Blogs are echo chambers.
77.9% of opinionated commenters agree with the blog author.
SLIDE 27
An algorithmic approach could help.
Consider an echo index.
SLIDE 28
Eric Gilbert | Tony Bergstrom | Karrie Karahalios University of Illinois at Urbana-Champaign
http://social.cs.uiuc.edu/echo.zip
all data & code