An ideological asymmetry in moral
contagion among political leaders
Ü @jayvanbavel | New York University
psyarxiv.com/43n5e/
What factors influence the diffusion of
moralized content in social networks?
Stage Theory Social-Intuitionist
Model
Dyadic Morality
Model
Moral Conviction
Model
Dual-Process
Model
Moral Contagion Hypothesis: moral emotion increases diffusion of
moralized content in online social networks.
Brady et al. (2017), PNAS
Moral emotions: emotions that reliably signal, either to others or to the
self, that something morally relevant has occurred (e.g., Keltner &
Haidt, 1999)
Moral contagion: the spread of moralized content as a result of people
incorporating others’ moral-emotional expressions as informational
input into their own evaluation of a situation that can inform their own
emotional state and also guide their decisions to share the content.
Moral contagion: the spread of moralized content as a result of:
• People incorporating others’ moral-emotional expressions as
information that informs their evaluation of a situation (e.g,.
Manstead & Fischer, 2001)
• This can inform their own emotional state and/or guide their
decisions to share the content.
.
Method
• Large social media sample from Twitter (N = 563,217)
• 3 studies, collecting Twitter messages on 3 different moral topics
• Gun control
• Same-sex marriage
• Climate change
Brady et al. (2017), PNAS
Measuring morality and emotion
• Morality and Emotion defined by their expression through language
in a message (Tausczik & Pennebaker, 2010)
Brady et al. (2017), PNAS
• Morality and Emotion defined by their expression through language
in a message (Tausczik & Pennebaker, 2010)
• Moral dictionary: Moral foundations lexicon (Graham, Haidt,
& Nosek, 2009) + piloted moral word list (Gantman & Van Bavel,
2014)
• Emotion dictionary: emotion words from Linguistic Inquiry and
Word Count software (LIWC) (Tausczik & Pennebaker, 2010)
Method
Brady et al. (2017), PNAS
Distinctly Moral Distinctly EmotionalMoral-Emotional
• Morality and Emotion defined by their expression through language
in a message (Tausczik & Pennebaker, 2010)
• Moral dictionary: Moral foundations lexicon (Graham, Haidt,
& Nosek, 2009) + piloted moral word list (Gantman & Van Bavel,
2014)
• Emotion dictionary: emotion words from Linguistic Inquiry and
Word Count software (LIWC) (Tausczik & Pennebaker, 2010)
Method
Brady et al. (2017), PNAS
Distinctly Moral Distinctly EmotionalMoral-Emotional
• Morality and Emotion defined by their expression through language
in a message (Tausczik & Pennebaker, 2010)
• Moral dictionary: Moral foundations lexicon (Graham, Haidt,
& Nosek, 2009) + piloted moral word list (Gantman & Van Bavel,
2014)
• Emotion dictionary: emotion words from Linguistic Inquiry and
Word Count software (LIWC) (Tausczik & Pennebaker, 2010)
Tweets with moral-emotional language
Brady et al. (2017), PNAS
Moral Word Emotion WordMoral-Emotional Word
Method
• Diffusion / Contagion defined by retweet count
• People typically retweet content because
1. they endorse the content
2. they trust the source (Metaxas et al., 2015)
Brady et al. (2017), PNAS
Moral-emotional language associated with significant
increases in diffusion
• Model: RT Count = distinctly moral language + distinctly emotional language +
moral-emotional language + covariates
Brady et al. (2017), PNAS
Gun Control Same-Sex Marriage Climate Change
Moral emotional language associated with significant
increases in diffusion
• Model: RT Count = distinctly moral language + distinctly emotional language +
moral-emotional language
Mean IRR = 1.20; 20% increase in retweets per ME word
Brady et al. (2017), PNAS
Gun Control Same-Sex Marriage Climate Change
Moral emotional language associated with significant
increases in diffusion
• Model: RT Count = distinctly moral language + distinctly emotional language +
moral-emotional language + covariates
Mean IRR = 1.20; 20% increase in retweets per ME word
Brady et al. (2017), PNAS
Gun Control Same-Sex Marriage Climate Change
Exploring boundary conditions
• Measured user ideology based on well-validated algorithm that
tracks political accounts each user follows (Barbera et al., 2015)
Moral-emotional
content
Information
Diffusion
Political Identity:
Receiver
Brady et al. (2017)
Moral emotional language associated with increases in
“echo-chamber” discourse
Liberals Conservatives
Node = user
Edge = retweet
Brady et al. (2017), PNAS
Results: Group Identity
Gun control
Same-sex
marriage
Climate change
Mean interaction IRR = 1.21 Brady et al. (2017), PNAS
Summary
• Moral emotion expression plays a key role in the diffusion of
moralized content
• Political identity of the receiver may be an important moderator
of this effect
Twitter study 2:
Moral contagion in the 2016 election and
partisanship
Moral-emotional
content
Information
Diffusion
Election Study: 2016 U.S. Presidential Election
Partisan
Source
• How do source cues interact with properties of a
message to affect diffusion?
Brady et al., (2018), JEP:G
Clinton vs. Trump
• Scraped twitter messages from both candidates posted in the
last 1 year leading up to the 2016 U.S. Presidential Election
• N = 9,505
Brady et al., (2018), JEP:G
Results: Clinton vs. Trump
Logretweetcount
# Moral-emotional words
Clinton Trump
Brady et al., (2018), JEP:G
Results: Clinton vs. Trump
Predictedretweetcount
# Moral-emotional words
Clinton Trump
IRR = 1.27, p <.001
Brady et al., (2018), JEP:G
Results: Clinton vs. Trump
Moral-emotional words associated with most viral messages
Brady et al., (2018), JEP:G
Why the asymmetry in moral contagion?
1. Ideology effects
Brady et al., (2018), JEP:G
Why the asymmetry in moral contagion?
1. Ideology effects
• Conservatism associated with moral dogmatism and
moralization of action (Jost et al., 2003)
• Conservatives more reactive to high arousal emotions
(e.g., anger) when moral views are threatened
(Tomkins, 1995; Jost et al., 2003)
• Prediction: Conservatives exhibit a relatively greater
moral contagion effect than Liberals
Brady et al., (2018), JEP:G
Why the asymmetry in moral contagion?
1. Other source cues?
• Gender, race and age (Brewer, 1988; Fiske & Neuberg,
1990)
Brady et al., (2018), JEP:G
U.S. Congress
• Scraped tweets from all U.S. Senators + House one year leading up
the 2016 U.S. Presidential Election
• Test for political ideology and other source effects
• N = 276,750
Brady et al., (2018), JEP:G
Results: U.S. Congress
# Moral-Emotional Words
LogRetweetCount
1 – Replicating moral contagion effect in political elites
Brady et al., (2018), JEP:G
Results: U.S. Congress
ConservativesLiberals
2 – Political ideology analysis
Cory Booker
Jeff Sessions
Brady et al., (2018), JEP:G
Results: U.S. Congress
ConservativesLiberals
2 – Political ideology analysis
Cory Booker
Jeff Sessions
Brady et al., (2018), JEP:G
Results: U.S. Congress
3 – Political ideology analysis: follow-up analyses
1. Gender effects
• No significant gender effects on any language category
2. Age effects
• Younger elites exhibited significantly greater moral contagion effect than
older elites, IRR = 0.96, p <.001, 95% CI = [0.95, 0.98]
3. Race effects
• Non-white elites exhibited significantly greater moral contagion effect
than white elites, IRR = 0.93, p = .019, 95% CI = [0.87, 0.99]
Brady et al., (2018), JEP:G
Ideology associated with the spread of
moralized content
Brady et al., (2018), JEP:G
Moral-emotional
content
Information
Diffusion
Partisan Source
GRATITUDE
Billy Brady Julian Wills John Jost
Research Assistants:
Miaohan Wang Jino Kwon Stephanie Leung
Dominic Burkart
https://psyarxiv.com/43n5e/

Moral contagion and political leaders

  • 1.
    An ideological asymmetryin moral contagion among political leaders Ü @jayvanbavel | New York University psyarxiv.com/43n5e/
  • 2.
    What factors influencethe diffusion of moralized content in social networks?
  • 3.
    Stage Theory Social-Intuitionist Model DyadicMorality Model Moral Conviction Model Dual-Process Model
  • 4.
    Moral Contagion Hypothesis:moral emotion increases diffusion of moralized content in online social networks. Brady et al. (2017), PNAS
  • 5.
    Moral emotions: emotionsthat reliably signal, either to others or to the self, that something morally relevant has occurred (e.g., Keltner & Haidt, 1999) Moral contagion: the spread of moralized content as a result of people incorporating others’ moral-emotional expressions as informational input into their own evaluation of a situation that can inform their own emotional state and also guide their decisions to share the content.
  • 6.
    Moral contagion: thespread of moralized content as a result of: • People incorporating others’ moral-emotional expressions as information that informs their evaluation of a situation (e.g,. Manstead & Fischer, 2001) • This can inform their own emotional state and/or guide their decisions to share the content. .
  • 7.
    Method • Large socialmedia sample from Twitter (N = 563,217) • 3 studies, collecting Twitter messages on 3 different moral topics • Gun control • Same-sex marriage • Climate change Brady et al. (2017), PNAS
  • 8.
    Measuring morality andemotion • Morality and Emotion defined by their expression through language in a message (Tausczik & Pennebaker, 2010) Brady et al. (2017), PNAS
  • 9.
    • Morality andEmotion defined by their expression through language in a message (Tausczik & Pennebaker, 2010) • Moral dictionary: Moral foundations lexicon (Graham, Haidt, & Nosek, 2009) + piloted moral word list (Gantman & Van Bavel, 2014) • Emotion dictionary: emotion words from Linguistic Inquiry and Word Count software (LIWC) (Tausczik & Pennebaker, 2010) Method Brady et al. (2017), PNAS Distinctly Moral Distinctly EmotionalMoral-Emotional
  • 10.
    • Morality andEmotion defined by their expression through language in a message (Tausczik & Pennebaker, 2010) • Moral dictionary: Moral foundations lexicon (Graham, Haidt, & Nosek, 2009) + piloted moral word list (Gantman & Van Bavel, 2014) • Emotion dictionary: emotion words from Linguistic Inquiry and Word Count software (LIWC) (Tausczik & Pennebaker, 2010) Method Brady et al. (2017), PNAS Distinctly Moral Distinctly EmotionalMoral-Emotional
  • 11.
    • Morality andEmotion defined by their expression through language in a message (Tausczik & Pennebaker, 2010) • Moral dictionary: Moral foundations lexicon (Graham, Haidt, & Nosek, 2009) + piloted moral word list (Gantman & Van Bavel, 2014) • Emotion dictionary: emotion words from Linguistic Inquiry and Word Count software (LIWC) (Tausczik & Pennebaker, 2010) Tweets with moral-emotional language Brady et al. (2017), PNAS Moral Word Emotion WordMoral-Emotional Word
  • 12.
    Method • Diffusion /Contagion defined by retweet count • People typically retweet content because 1. they endorse the content 2. they trust the source (Metaxas et al., 2015) Brady et al. (2017), PNAS
  • 13.
    Moral-emotional language associatedwith significant increases in diffusion • Model: RT Count = distinctly moral language + distinctly emotional language + moral-emotional language + covariates Brady et al. (2017), PNAS Gun Control Same-Sex Marriage Climate Change
  • 14.
    Moral emotional languageassociated with significant increases in diffusion • Model: RT Count = distinctly moral language + distinctly emotional language + moral-emotional language Mean IRR = 1.20; 20% increase in retweets per ME word Brady et al. (2017), PNAS Gun Control Same-Sex Marriage Climate Change
  • 15.
    Moral emotional languageassociated with significant increases in diffusion • Model: RT Count = distinctly moral language + distinctly emotional language + moral-emotional language + covariates Mean IRR = 1.20; 20% increase in retweets per ME word Brady et al. (2017), PNAS Gun Control Same-Sex Marriage Climate Change
  • 16.
    Exploring boundary conditions •Measured user ideology based on well-validated algorithm that tracks political accounts each user follows (Barbera et al., 2015) Moral-emotional content Information Diffusion Political Identity: Receiver Brady et al. (2017)
  • 17.
    Moral emotional languageassociated with increases in “echo-chamber” discourse Liberals Conservatives Node = user Edge = retweet Brady et al. (2017), PNAS
  • 18.
    Results: Group Identity Guncontrol Same-sex marriage Climate change Mean interaction IRR = 1.21 Brady et al. (2017), PNAS
  • 19.
    Summary • Moral emotionexpression plays a key role in the diffusion of moralized content • Political identity of the receiver may be an important moderator of this effect
  • 20.
    Twitter study 2: Moralcontagion in the 2016 election and partisanship
  • 21.
    Moral-emotional content Information Diffusion Election Study: 2016U.S. Presidential Election Partisan Source • How do source cues interact with properties of a message to affect diffusion? Brady et al., (2018), JEP:G
  • 22.
    Clinton vs. Trump •Scraped twitter messages from both candidates posted in the last 1 year leading up to the 2016 U.S. Presidential Election • N = 9,505 Brady et al., (2018), JEP:G
  • 23.
    Results: Clinton vs.Trump Logretweetcount # Moral-emotional words Clinton Trump Brady et al., (2018), JEP:G
  • 24.
    Results: Clinton vs.Trump Predictedretweetcount # Moral-emotional words Clinton Trump IRR = 1.27, p <.001 Brady et al., (2018), JEP:G
  • 25.
    Results: Clinton vs.Trump Moral-emotional words associated with most viral messages Brady et al., (2018), JEP:G
  • 26.
    Why the asymmetryin moral contagion? 1. Ideology effects Brady et al., (2018), JEP:G
  • 27.
    Why the asymmetryin moral contagion? 1. Ideology effects • Conservatism associated with moral dogmatism and moralization of action (Jost et al., 2003) • Conservatives more reactive to high arousal emotions (e.g., anger) when moral views are threatened (Tomkins, 1995; Jost et al., 2003) • Prediction: Conservatives exhibit a relatively greater moral contagion effect than Liberals Brady et al., (2018), JEP:G
  • 28.
    Why the asymmetryin moral contagion? 1. Other source cues? • Gender, race and age (Brewer, 1988; Fiske & Neuberg, 1990) Brady et al., (2018), JEP:G
  • 29.
    U.S. Congress • Scrapedtweets from all U.S. Senators + House one year leading up the 2016 U.S. Presidential Election • Test for political ideology and other source effects • N = 276,750 Brady et al., (2018), JEP:G
  • 30.
    Results: U.S. Congress #Moral-Emotional Words LogRetweetCount 1 – Replicating moral contagion effect in political elites Brady et al., (2018), JEP:G
  • 31.
    Results: U.S. Congress ConservativesLiberals 2– Political ideology analysis Cory Booker Jeff Sessions Brady et al., (2018), JEP:G
  • 32.
    Results: U.S. Congress ConservativesLiberals 2– Political ideology analysis Cory Booker Jeff Sessions Brady et al., (2018), JEP:G
  • 33.
    Results: U.S. Congress 3– Political ideology analysis: follow-up analyses 1. Gender effects • No significant gender effects on any language category 2. Age effects • Younger elites exhibited significantly greater moral contagion effect than older elites, IRR = 0.96, p <.001, 95% CI = [0.95, 0.98] 3. Race effects • Non-white elites exhibited significantly greater moral contagion effect than white elites, IRR = 0.93, p = .019, 95% CI = [0.87, 0.99] Brady et al., (2018), JEP:G
  • 34.
    Ideology associated withthe spread of moralized content Brady et al., (2018), JEP:G Moral-emotional content Information Diffusion Partisan Source
  • 35.
    GRATITUDE Billy Brady JulianWills John Jost Research Assistants: Miaohan Wang Jino Kwon Stephanie Leung Dominic Burkart https://psyarxiv.com/43n5e/