Antisocial Computing

Antisocial Computing

In online social networks, large information cascades can develop as people share content with one another. However, as these cascades develop through complex processes, prior work has argued that their future trajectory may be inherently unpredictable. My research introduces methods for studying the mechanisms of these cascades and predicting their spread. Analyzing billions of interactions by hundreds of millions of users on Facebook, I show how the future growth and structure of these cascades can be predicted, how cascades may resurface after lying dormant for months, and how diverse social protocols can produce large information cascades. Through revealing the mechanisms in which information diffuses in social media, this work explores a future where systems can better promote sharing behavior online.

Presented at the Stanford Human-Computer Interaction Seminar.

8480b47e733a040fba07c32da414b0e0?s=128

Justin Cheng

June 02, 2017
Tweet

Transcript

  1. Justin Cheng Stanford University ANTISOCIAL COMPUTING Explaining and Predicting Negative

    Behavior Online
  2. Vieweg, et al. (2010); Kittur, et al. (2013); Burke &

    Kraut (2016)
  3. None
  4. Time (2016); The Atlantic (2016); Vanity Fair (2017)

  5. 47% of online users have been harassed Data & Society

    (2017)
  6. Popular Science (2013); The Verge (2015); Chicago Sun-Times (2014)

  7. Why is bad behavior so prevalent?
 (›°□°)›ớ ᵲᴸᵲ Research Question

  8. Understanding bad behavior helps us build healthier communities Implications Systems

    Guidelines Interventions
  9. Antisocial behavior is largely due to sociopaths Prior Work Donath

    (1999); Hardaker (2010); Buckels, et al. (2014)
  10. Antisocial behavior is largely due to ordinary people This Work

  11. Antisocial Behavior & Its Spread Talk Outline What causes antisocial

    behavior? Can such cascades be predicted? Does it worsen over time? 1 2 3
  12. Data Mining + Crowdsourcing Research Approach Large-scale Analysis + Experiments

  13. Identifying principles of online behavior The Broader Picture Data +

    ML + Network Science + HCI
  14. Antisocial Behavior & Its Spread Talk Outline 1 2 3

    What causes antisocial behavior? Can such cascades be predicted? Does it worsen over time?
  15. CSCW 2017 (Best Paper); ICWSM 2015 (Honorable Mention) with M.

    Bernstein, C. Danescu-Niculescu-Mizil, J. Leskovec CAN ANYONE BECOME A TROLL? Causes of Antisocial Behavior in Online Discussions
  16. CONTENT WARNING!
 This talk contains depictions of trolling that use

    strong language. !
  17. None
  18. It also shows that Islam and Christianity teaching women to

    dress modest could be right afterall.
  19. It also shows that Islam and Christianity teaching women to

    dress modest could be right afterall. Clearly that is the only logical conclusion to this article. Now if you'll excuse me, I need to iron my tarp. I have work on Monday, and I want to appear 'modest'. fail at life. go bomb yourself. Religious nut alert
  20. We studied multiple large comment- based news communities. 470M posts

    831M votes 76M users
  21. What is trolling?

  22. What is trolling? Engaging in negatively marked online behavior Taking

    pleasure in upsetting others Not following the rules Disrupting a group while staying undercover Donath (1999); Hardaker (2010); Kirman (2012); Schwartz (2008)
  23. Trolling is behavior that occurs outside community norms. Defined using

    community guidelines Our Definition e.g., name-calling, personal attacks, profanity, threats, hate speech, ethnically/racially offensive material
  24. Are trolls just a vocal minority? Donath (1999); Hardaker (2010);

    Shachaf & Hara (2010); NYT (2008); Wired (2014); Vox (2014)
  25. How much do trolls troll? Proportion of Banned Users 0

    0.1 0.2 0.3 0.4 Proportion of Deleted Posts 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
  26. The distribution of trolls is bimodal Proportion of Banned Users

    0 0.1 0.2 0.3 0.4 Proportion of Deleted Posts 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
  27. Are there two types of trolls? Proportion of Banned Users

    0 0.1 0.2 0.3 0.4 Proportion of Deleted Posts 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Situational trolling? Lifelong trolling?
  28. What if antisocial behavior is situational?

  29. Challenge: how to show that antisocial behavior is situational? Observational

    data isn’t causal
  30. Challenge: how to show that antisocial behavior is situational? Experiments

    hard to generalize
  31. Simulated Discussion Experiment Large-Scale Analysis Solution: Experiment + Observational Study

  32. Anyone can become a troll Our Hypothesis

  33. “Broken windows” theory Zimbardo (1969); Wilson (1982)

  34. Unpleasant stimuli increase aggression Jones & Bogat (1978); Rotton &

    Frey (1985)
  35. Experiment: simulated discussion forum

  36. N=667, 40% female Quiz Discussion Experimental method

  37. Quiz Discussion Experimental method ×

  38. Positive/Negative Mood Positive/Negative Context Experimental method × Quiz Discussion

  39. Easy quiz (positive mood)

  40. Difficult quiz (negative mood)

  41. Positive discussion context

  42. Negative discussion context

  43. People reported being in a worse mood after the difficult

    quiz Easy: 12.2 Difficult: 40.8 (POMS mood disturbance [higher scores = worse mood], p<0.01) Manipulation Check
  44. Initial seed posts in the negative context condition perceived worse

    Positive: 90% upvoted Negative: 36% upvoted (p < 0.01) Manipulation Check
  45. How did trolling differ across conditions? Two expert raters labeled

    posts independently
  46. How did trolling differ across conditions? Positive Mood Negative Mood

    Positive Context Negative Context % Troll Posts
  47. Trolling is lowest with positive conditions… Positive Mood Negative Mood

    Positive Context 35% Negative Context % Troll Posts
  48. …increases with either negative condition… Positive Mood Negative Mood Positive

    Context 35% 49% Negative Context 47% % Troll Posts
  49. …and almost doubles in the worst case Positive Mood Negative

    Mood Positive Context 35% 49% Negative Context 47% 68% % Troll Posts (p < 0.05 using a mixed effects logistic regression model)
  50. Negative affect almost triples Positive Mood Negative Mood Positive Context

    1.1% 1.4% Negative Context 2.3% 2.9% % Negative Affect Words (LIWC) (p < 0.05)
  51. Hilary is a solid candidate. As a woman, I appreciate

    that she's a woman, but it's not the only reason I think she would do well in office. Positive Mood + Context
  52. Anyone who votes for her is a complete idiot. These

    supporters are why this country is in such bad shape now. Uneducated people. Negative Mood + Context
  53. Bad mood and negative discussion context increase trolling

  54. Simulated Discussion Experiment Large-Scale Analysis of CNN.com Online Experiment +

    Observational Study
  55. Can trolling, like mood, vary with the time of day

    and day of week? Replicating Mood Golder & Macy (2011) Neg. Affect Time of Day ?
  56. How does trolling vary with time of day? Proportion of

    Flagged Posts 0.03 0.033 0.036 0.039 0.042 Time of Day 0 6 12 18 24
  57. Trolling peaks in the evening… Proportion of Flagged Posts 0.03

    0.033 0.036 0.039 0.042 Time of Day 0 6 12 18 24 Negative Affect (Golder & Macy)
  58. …and early in the work week. Proportion of Flagged Posts

    0.03 0.032 0.034 0.036 0.038 0.04 Day of Week Mon Tue Wed Thu Fri Sat Sun
  59. Trolling peaks when moods are worse Time of Day Proportion

    of Flagged Posts Negative Affect Proportion of Downvotes Day of Week
  60. Mood spills over from prior discussions Replicating Mood

  61. … … (Discussion) Mood spills over from prior discussions

  62. … … … ? … (Discussion) (Unrelated Discussions) … ?

    … Mood spills over from prior discussions
  63. … … … … (Discussion) … … Mood spills over

    from prior discussions
  64. A user who trolled in a previous discussion is twice

    as likely to troll in a later, unrelated discussion (p < 0.01) Replicating Mood
  65. The initial post affects subsequent trolling Replicating Context

  66. (Separate discussions of same article) … ? ? ? …

    ? ? ? The initial post affects subsequent trolling
  67. (Separate discussions of same article) … … The initial post

    affects subsequent trolling
  68. An initial troll post increases the subsequent trolling by 63%

    (p < 0.01) Replicating Context
  69. Can we predict trolling before it happens? Balanced dataset of

    120K posts Logistic regression
  70. Mood Context The User What factors affect trolling? Trolling is

    situational Trolling is innate
  71. How predictable are troll posts? User-specific Mood Discussion Context Combined

    AUC 0.00 0.20 0.40 0.60 0.80 0.78 0.74 0.60 0.66
  72. How predictable are troll posts? User-specific Mood Discussion Context Combined

    AUC 0.00 0.20 0.40 0.60 0.80 0.78 0.74 0.60 0.66
  73. How predictable are troll posts? User-specific Mood Discussion Context Combined

    AUC 0.00 0.20 0.40 0.60 0.80 0.78 0.74 0.60 0.66
  74. Troll or not? User

  75. Troll or not? User Mood

  76. Troll or not? User Mood Other users { }

  77. Because trolling is situational, ordinary people can end up trolling

  78. None
  79. Can voting mitigate bad behavior?

  80. Downvoting causes negative behavior to worsen Our Hypothesis

  81. Antisocial Behavior & Its Spread Talk Outline 1 2 3

    What causes antisocial behavior? Can such cascades be predicted? Does it worsen over time?
  82. How Antisocial Behavior Worsens ICWSM 2014 with C. Danescu-Niculescu-Mizil, J.

    Leskovec CAN ANTISOCIAL BEHAVIOR SPI R A L?
  83. Downvoting causes negative behavior to worsen Our Hypothesis

  84. What effects do evaluations have? Positively evaluated ? ? Negatively

    evaluated
  85. What is a positive or negative evaluation?

  86. Defining positive and negative evaluations : 9 1 2 8

    N↑
 N↑ + N↓ 2
 2+8 = = 0.2 p↓ ≤ : N↑
 N↑ + N↓ 9
 9+1 = = 0.9 p↑ ≥ Positive Evaluation Negative Evaluation (validated using a crowdsourcing experiment)
  87. What effects do evaluations have? Positively evaluated ? ? Negatively

    evaluated
  88. Does feedback encourage better behavior? Skinner (1938)

  89. Or is bad stronger than good? Brinko (1993); Baumeister, et

    al. (2001)
  90. Four large comment-based news communities

  91. What effects do evaluations have? Positively evaluated Negatively evaluated

  92. What effects do evaluations have? … … … …

  93. What effects do evaluations have? … … … … Before

    After vs. Before After vs.
  94. Challenge: how to compare different users and posts? Aren’t downvoted

    users/posts inherently worse?
  95. Solution: propensity score matching PSM: Rosenbaum (1983); CEM: Iacus, et

    al. (2012) Positively evaluated Negatively evaluated
  96. Match on text quality Similar text quality q } ≈

  97. Computing text quality Learn p with bigrams
 (binomial regression) 1

    3 lorem ipsum… q = ? Lorem… ? ? 9 2 lorem ipsum… … Text quality q is predicted p
  98. Validating text quality Manually label subset (n=171) using crowdsourcing lorem

    ipsum… Good Bad Good Good # Good
 # Total q’=
  99. Validating text quality Manually label subset (n=171) using crowdsourcing lorem

    ipsum… Good Bad Good Good # Good
 # Total q’= ? ? corr(q’, p) = corr(q’, q) =
  100. Validating text quality (n.s.) (p < 0.01) p: actual proportion

    of upvotes q: text quality (predicted proportion) q’: crowd guess corr(q’, q) = corr(q’, p) = 0.11 0.25
  101. Validating text quality |Residuals| 0 0.2 0.4 0.6 0.8 1

    0.6 0.7 0.8 0.9 1 |q − q’| (n.s. using a Breusch-Pagan test) Text Quality q
  102. Match on text quality Similar text quality q(c↑ )=q(c↓ )

    } ≈
  103. …as well as other covariates Similar history (# posts, overall

    proportion of upvotes, etc.) { ≈ … … ≈
  104. …as well as other covariates ≈ … … … …

  105. How are subsequent posts evaluated? ≈ … … … …

  106. How much are evaluations due to textual or community effects?

  107. How much are evaluations due to textual effects (i.e., people

    writing worse)? f***ing a****** i.e., downvoting because of post content
  108. How much are evaluations due to community effects (i.e., inherent

    bias)? We dislike you. i.e., downvoting because of community dislikes author We dislike you.
  109. Do people write better/worse after a positive/negative evaluation? Textual Effects

  110. ≈ … … … … Better/Worse? Do people write better/worse

    after a positive/negative evaluation? ≈
  111. Text quality drops significantly after a negative evaluation… (p <

    0.05, mean effect size r = 0.18) … … Negativity bias
  112. …but doesn’t change after a positive evaluation … … (n.s.)

    Negativity bias
  113. How does community bias change after an evaluation? Community Effects

  114. … … … … How does community bias change after

    an evaluation?
  115. Measuring community bias N↑ q N↓ N↑
 N↑ + N↓

    = 0.5 p(c) = = 0.8 q(c) p(c) q(c) = 0.3 Prop. Upvotes Text Quality Community Bias − −
  116. Community bias increase more after a negative than positive evaluation

    (p < 0.01, mean effect size r = 0.13) … … Halo effect
  117. More positive Negative Eval. Positive Eval.

  118. More positive Negative Eval. Positive Eval. Similar text quality
 q(c↑

    )=q(c↓ )
  119. More positive Before Negative Eval. Positive Eval. Similar history Similar

    text quality
 q(c↑ )=q(c↓ )
  120. More positive Before After Negative Eval. Positive Eval. Similar history

    Similar text quality
 q(c↑ )=q(c↓ ) Worse text quality
 q(c↑(1,3) ) > q(c↓(1,3) ) *
  121. More positive Before After Negative Eval. Positive Eval. Similar history

    Similar text quality
 q(c↑ )=q(c↓ ) Worse text quality
 q(c↑(1,3) ) > q(c↓(1,3) ) Worse perception
 q(c↓(1,3) ) - p(c↓(1,3) ) > q(c↑(1,3) ) - p(c↑(1,3) ) * *
  122. What happens to negatively-evaluated users?

  123. They post worse content Perceptions of them become worse They

    post more frequently* They evaluate others more negatively* * More details in our ICWSM 2014 paper (http://bit.ly/feedback-paper) What happens to negatively-evaluated users?
  124. Troll or not? User Mood Other users { }

  125. Troll or not? User Mood Other users { }

  126. Troll or not? User Mood Other users { }

  127. Trolls may start out normal, but tip into a spiral

    and never recover
  128. Do communities worsen over time?

  129. Communities may worsen over time (?) Proportion of Upvotes 0.6

    0.65 0.7 0.75 0.8 Time December 2012 February 2013 April 2013 June 2013 August 2013
  130. Antisocial Behavior & Its Spread Talk Outline 1 2 3

    What causes antisocial behavior? Can such cascades be predicted? Does it worsen over time?
  131. The Predictability of Information Cascades in Social Networks CAN C

    S BE PREDICTED? AS A C DE WWW 2014; WWW 2016 with L. Adamic, P. A. Dow, J. Kleinberg, J. Leskovec
  132. Rumors on Facebook ICWSM 2014 (with A. Friggeri, L. Adamic,

    and D. Eckles)
  133. Same rumor, different popularity

  134. Are these cascades predictable?

  135. None
  136. None
  137. None
  138. None
  139. None
  140. None
  141. Are cascades unpredictable?

  142. Large cascades are rare Empirical CCDF 0.00 0.20 0.40 0.60

    0.80 1.00 Cascade size 0 200 400 600 800 1000 0.09 100
  143. “Increasing the strength of social influence increased both inequality and

    unpredictability of success.” Salganik, Dodds & Watts (2006)
  144. Cascades can recur after long periods

  145. Cascades are predictable Our Hypothesis size, structure, content even if

    they recur
  146. How do we begin to predict cascade growth?

  147. Challenge: how to predict cascade growth? ? k=5 reshares
 observed

  148. Challenge: how to predict cascade growth? ? Will a cascade

    get 100 reshares? Exactly how big will a small cascade get? Only consider the largest cascades?
  149. Challenge: how to predict cascade growth? ? Will a cascade

    get 100 reshares? Exactly how big will a small cascade get? Only consider the largest cascades? class imbalance outliers skew results selection bias
  150. Solution: will a cascade reach the median? ? ≤ the

    median f(k) ≥ the median f(k) k=5 reshares
 observed
  151. Solution: will a cascade double in size? ? ≤ the

    median f(k) ≥ the median f(k) k=5 reshares
 observed
  152. Given that a cascade has obtained k reshares, will it

    double in size? balanced track growth over time Cascade Growth Prediction Problem
  153. Reshare cascades on Facebook 70M cascades 5B reshares Activity over

    28 days
  154. Content
 has overlaid text captions … User
 friend count
 gender

    … Structural
 tree depth outdegree … Temporal
 time between shares change in time … What factors affect predictability?
  155. How predictable is cascade doubling? All Temporal All but temporal

    Structural User Content AUC (k=5) 0.00 0.23 0.45 0.68 0.90 0.58 0.71 0.74 0.79 0.87 0.88 All but temporal
  156. Given that a cascade has obtained k reshares, will it

    double in size? Cascade Growth Prediction Problem
  157. Given that a cascade has obtained 5 reshares, will it

    double in size? Cascade Growth Prediction Problem
  158. Given that a cascade has obtained 100 reshares, will it

    double in size? Cascade Growth Prediction Problem
  159. How does performance change with k? k = 5 k

    > 10 k = 10 k > 20
  160. How does performance change with k? k = 5 k

    > 10 k = 10 k > 20 Less data More data
  161. How does performance change with k? k = 5 k

    > 10 k = 10 k > 20 Shorter-term Longer-term
  162. Easier to predict larger cascades doubling Accuracy 0.78 0.79 0.8

    0.81 0.82 Number of reshares observed, k 0 25 50 75 100
  163. Cascade growth is predictable

  164. Cascade structure is predictable AUC = 0.80 for predicting structural

    virality * More details in our WWW 2014 paper (http://bit.ly/memes-paper) vs.
  165. Cascade recurrence is predictable AUC = 0.89 for predicting a

    subsequent burst * More details in our WWW 2016 paper (http://bit.ly/cascades-paper) vs.
  166. Troll or not? User Mood Other users { }

  167. None
  168. What we now know What we thought Trolls are a

    vocal minority Trolls can be ordinary people Trolling is innate Trolling can spiral from a single bad post Cascades can be predicted Cascades are unpredictable ANTISOCIAL COMPUTING
  169. Predicting the demise of communities Proportion of Upvotes 0.6 0.65

    0.7 0.75 0.8 Time December February April June August Future Directions
  170. Designing prosocial discussion platforms Future Directions

  171. Designing prosocial discussion platforms Future Directions

  172. Munger (2016) Introducing conversation mediators Future Directions Don’t be a

    n****r. Hey man, just remember that there are real people who are hurt when you harass them with that kind of language. (e.g., bots)
  173. WWW 2017 (with S. Kumar, J. Leskovec, and V.S. Subrahmanian)

    Identifying different types of trolling Future Directions Possibly the best blog I’ve ever read major props to you Thanks. I knew Marvel fans would try to flame me, but they have nothing other than “oh that’s your opinion” Quit talking to yourself […] (e.g., sockpuppets)
  174. Addressing polarization Future Directions Measuring algorithmic impact Tracking cascades at

    scale
  175. Holistic approaches for analyzing and building social systems Research Approach

  176. Holistic approaches for analyzing and building social systems Large-scale Analysis

    Experimentation + Macro-scale Micro-scale + Understand Build + Research Approach
  177. Multi-methods analyses identify patterns in data, verify hypotheses, make predictions,

    and develop social systems. Multi-methods analyses identify patterns in data, verify hypotheses, make predictions, and inform the design of better social systems.
  178. Jure Leskovec Michael Bernstein Jon Kleinberg Lada Adamic Thank you!

  179. James Landay Jeff Hancock Cristian Danescu-Niculescu-Mizil Dan Cosley Thank you!

  180. Thank you!

  181. Thank you!

  182. Thank you!

  183. Thank you! Stanford HCI Group SNAP Group Stanford VPGE Microsoft

    Research Facebook Pinterest Disqus
  184. Justin Cheng / @jcccf / clr3.com Stanford University More resources

    and credits: http://bit.ly/jobtalkcredits ANTISOCIAL COMPUTING Explaining and Predicting Negative Behavior Online