Hate in Context

Week 8: Community Impacts of Hate Crime

Violeta Haas

Department of Political Science

School of Social Sciences and Philosophy

Trinity College Dublin

Updates

Organisational

Weekly Comments

  • Weekly comments due Sunday 11:59 pm (unless position-paper week)
  • Keep track of used jokers (LENS: one additional joker)

Office hours

  • Slots + Zoom links are on Blackboard
  • Please check the latest version (schedule was updated)
  • Today: last chance to reach out for a slot if you forgot

Week 12 topic

  • How to write a research proposal
  • Monday: input (structure, expectations, resources, tips)
  • Wednesday: co-working session + bring questions
Causes Weeks 2–6 Demographic change
Economic conditions
(Social) media
Threatening events
Political elites
Consequences Week 8 Community impacts
Responses Weeks 9–10 Contact
Persuasion
Guest lecture Week 11 Inclusionary policy
Exclusionary policy
Workshop Week 12 How to write a
research proposal

This Week

  1. Mitts (2019)
  2. Branton & Kropf (2026)
  3. Roman & Thompson (2024)
  4. Synthesis/discussion
  5. Outlook

Mitts (2019)

Business Insider chart: number of foreign ISIS fighters who returned home from Syria or Iraq, 2016–2017

Between 2011 and 2016, about 30,000 foreign fighters traveled to Syria and Iraq to join the Islamic State (Benmelech and Klor 2018). Many from Western countries.

Overview

  • Research question: Does local anti-Muslim hostility drive pro-ISIS radicalisation in the West?
  • Y: online indicators of ISIS support
  • X: local anti-Muslim hostility
  • Theory: hostility → isolation/search for acceptance → radical messages become more attractive
  • Design/data: geo-referenced Twitter behaviour of ISIS sympathisers in FR, UK, DE, BE

Discussion: radicalisation

In groups of 3–4 (≈10–15 min):

  1. How can we define radicalisation?
  2. What stages of radicalisation are there?
  3. Across stages, what role do social media play in recruiting and which platform types matter most where?
  4. Where does Mitts’ (2019) study sit within those stages?

How can we define radicalisation?

Q1: Definition

Among the only things radicalisation experts agree on is that radicalisation is a process (Schmid, 2013: 1), which entails the gradual adoption of extremist ideas and would end, if completed, in the practice of violent extremism or terrorism.

Mitts’ (2019) Definition of radicalization

Most agree that radicalization involves a change in ideology or beliefs that support indiscriminate violence against civilians for political reasons, or a group that represents this ideology and actions (Borum 2011; Sedgwick 2010; Wilner and Dubouloz 2010).

Radicalisation as a process

  • A gradual personal and political transformation
  • A “catalyst event” can accelerate it ≈ Wiktorowicz’s (2005) cognitive opening: a shock/crisis that unsettles prior beliefs and makes new worldviews more receptive
  • Catalysts can be economic (job loss, blocked mobility), social (alienation, discrimination), political (war, conflict), or personal (death of a loved one)

Figure 1: radicalisation as a process from ideological engagement to violent extremism

Muro (2016), Fig. 1

What stages of radicalisation are there?

Borum: four-stage mindset

Borum (2003; 2011): how grievances and vulnerabilities turn into hatred and justification for violence (“terrorist mindset”).

  • Grievance (“It’s not right”): an unsatisfying event/condition is identified
  • Injustice (“It’s not fair”): that condition is framed as unjust
  • Target attribution (“It’s your fault”): blame is assigned to a broad group (e.g., the West)
  • Demonisation (“You’re evil”): the responsible party is vilified → indiscriminate aggression becomes easier to justify

Figure 2: Borum four-stage model from grievance to demonisation

Muro (2016), Fig. 2

Moghaddam: staircase to terrorism

Moghaddam (2005): radicalisation as a staircase, i.e., everyone starts on the ground floor, each floor is narrower (fewer people ascend) and almost none reach the top.

  • Ground: discontent and perceived injustice/deprivation
  • 1st floor: some try to do something (seek options/solutions)
  • 2nd floor: if no solution → displace aggression onto enemy
  • 3rd floor: moral engagement with group legitimising the cause
  • 4th floor: recruitment into a terrorist organisation
  • 5th floor: trained to sidestep inhibitory mechanisms → violence/terrorist act

Figure 3: Moghaddam staircase to terrorism

Muro (2016), Fig. 3 — built around suicide bombing; stages may not generalise everywhere

McCauley & Moskalenko: Pyramid model

McCauley & Moskalenko (2008): Radicalisation = the gradient from a broad base of sympathisers to a tiny terrorist apex (higher on the pyramid → more commitment/fewer people).

  • Sympathisers (base): agree with the goals terrorists claim to fight for and identify with the community they claim to represent but not violent means
  • Supporters: justify the goals and the violent means
  • Activists: do not commit the violence themselves but may provide support (recruitment, political/financial aid,…)
  • Radicals: few active terrorists/fighters (indiscriminate violence against civilians)

Figure 4: pyramid model of radicalisation

Muro (2016), Fig. 4

Mitts’ three stages

Mitts (2019, p. 174): while models vary, most involve three common stages.

Three common stages: attraction, commitment, possible violence

Across stages, what role do social media play in recruiting and which platform types matter most where? Draw a schematic figure.

Q3: Role of social media

Three stages with platform types: mainstream targeting in attraction, mainstream belonging in commitment, fringe migration in violence

Where does Mitts’ (2019) study sit within those stages?

Q4: Mitts concentrates on second stage

Full stages and social media figure with stage 2 commitment and its social media box framed as Mitts focus

Discussion: Mitts’ design

Turn to a neighbor (≈4 min):

  1. How are ISIS sympathisers identified? How are they located geographically?
  2. How is radicalisation (Y) measured? How is anti-Muslim hostility (X) measured?
  3. Any problems you see with these measures?

How are ISIS sympathisers identified? How are they located geographically?

Identifying and locating sympathisers

Identified

  • ISIS activists (~15,000): accounts publicly flagged by anti-ISIS hacktivist groups; scraped in real time (Dec 2015–Jan 2017) before suspension
  • Followers (~1.6 million): everyone following those activist accounts

Located

  • Most users do not self-report a usable location
  • Spatial label propagation (Jurgens 2013): predict a user’s offline place from the known locations of friends/followers
  • Matched predicted locations to localities in FR, UK, DE, BE

How is radicalisation (Y) measured? How is anti-Muslim hostility (X) measured?

Measuring Y and X

Radicalisation (Y)

User-level indicators:

  • Flagged as an ISIS activist
  • Account suspended for ISIS links
  • Number of ISIS accounts followed

Tweet content (supervised ML; human-labelled training data; EN/AR/FR/DE):

  • Sympathy with ISIS
  • Life in ISIS territories/travel to Syria/foreign fighters
  • Anti-West rhetoric

Anti-Muslim hostility (X)

  • Main measure: local far-right/anti-Muslim party vote share (proxy for anti-Muslim animosity)
  • Additional (UK): police-recorded hate crimes at force-area level — as a second X and as a moderator of the vote-share link

Any problems you see with these measures?

Problems with these measures

  • Accounts ≠ people: sockpuppets/multiple handles can inflate counts
  • Location prediction is imperfect/noisy
  • Far-right vote share is an imperfect proxy for hostility (also picks up elite cues). Hate-crime is a more direct measure but limited to UK.

What are the main findings?

Findings

Mitts 2019 Table 5: far-right vote share and support for ISIS on Twitter across multiple outcomes

Higher local far-right vote share is associated more online radicalisation across several radicalization measures.

Hate crimes and ISIS support (UK)

Mitts 2019 Figure 5: hate crimes and far-right vote share linked to pro-ISIS tweets in the UK

In the UK, more hate crime are associated with more pro-ISIS tweeting (A). The link between far-right vote-shares and pro-ISIS tweets is also stronger where hate crime is higher (B)

Event responses by far-right climate

Mitts 2019 Figure 7: larger rise in pro-ISIS rhetoric after events in areas with higher far-right vote share

After terrorist attacks, ISIS propaganda, and PEGIDA marches, pro-ISIS rhetoric rises more where far-right vote share is higher

A threshold for hostile climate?

Mitts 2019 Figure 7 with a vertical line at 15 percent far-right vote share in each panel

Sparse data/uncertainty in the middle of the distribution; different party systems across countries; when does far-right support become visible enough to shape norms and everyday hostility; something else?

Any surprising findings?

A common elite/media frame

Jihadi radicalisation often framed as an “imported” problem by (far right) political elites:

Portrait of Georg Pazderski (AfD)

Georg Pazderski (AfD)

“Not all refugees are terrorists, but most terrorists nowadays are refugees.” (2016)

Portrait of Marine Le Pen (FN)

Marine Le Pen (FN)

“Multiculturalism is the soft weapon of Islamic fundamentalists…” (2017)

Portrait of Charles Michel (Belgian PM)

Charles Michel (Belgian PM)

“Almost every time, there is a link to Molenbeek.” (2015)

Guardian 2015: Molenbeek as Europe's jihadi central

The Guardian (Nov 2015)

Data on Sympathizers and Activists does not support this!

Asylum seekers (FR/DE)

Mitts 2019 Table 7: asylum seekers negatively related to ISIS indicators in France and Germany

Asylum-seeker share is negatively related to pro-ISIS indicators, while far-right vote share stays positive

Muslim share and foreigners (UK)

Mitts 2019 Table 8: Muslim share negatively related to pro-ISIS content in the UK; foreigners weak

UK: % Muslim negatively correlated with pro-ISIS content; % foreigners is weak; far-right vote share remains positive and robust

Discussion: beyond ISIS

Turn to a neighbor (≈4 min):

How applicable do you think these dynamics of recruitment are for other extremist groups — say, white supremacists?

Where are similarities? What might differ?

Beyond ISIS: similarities

Recruitment/platforms

  • Strong parallels once social media are in the picture
  • Some far-right milieus borrow ISIS-style propaganda, aesthetics, and tactics
  • Spectacularisation of violence (livestreams)

Online belonging

  • Offline isolation/failed belonging → solace in an online community
  • Messages of inclusion, camaraderie

Narratives/pathways

  • Shared narrative structure: hate + peril + urgency; heroism; apocalyptic “restore the ideal order”
  • Similar pathways into violence even when ideologies differ
  • Reciprocal radicalisation: each side uses the other as proof of threat

Mitts on ISIS’s online community

“… [ISIS] was particularly effective at disseminating messages of inclusion … a lot of camaraderie and really love … And if you put that in a context of isolation in the physical world … when somebody just feels like they do not belong in their offline world, they might find solace and happiness online.”

— Tamar Mitts, with Tarik Abou-Chadi, Transformation of European Politics, Ep. 15

Beyond ISIS: differences

Mechanism

  • Mitts (2019): experienced hostility/discrimination → isolation → support
  • Far-right (e.g. US): often not personal hate victimhood, but a story that “the white race is under threat” (great replacement conspiracy) from diversity → join/fight or “go extinct”
  • “threat” ≈ “isolation”

Who is recruited?

  • Far-right targets are often in positions of power (sometimes police/military); jihadi recruits more often from marginalised positions

Ideology

  • Different enemies and frames: religious/jihadist vs racist/anti-Muslim/ethnonationalist
  • Overlapping hate–peril–urgency rhetoric, but different targets, imagery, and emphasis (e.g. religion/promised future change more central in jihadi texts)

Mitts & Abou-Chadi, Transformation of European Politics, Episode 15

Critique/Open questions?

Open question: what about stage 1 and 3?

Full stages and social media figure with stage 2 commitment and its social media box framed as Mitts focus

Sample concentrates on stage 2: activists + followers of ISIS accounts. We don’t know how discrimination plays out for individuals getting attracted but not yet following + some individuals might have already moved activity on to (unobserved) fringe platforms.

Recommendations

Cover art for Transformation of European Politics Podcast
Poster for Radicals Rising (The Naked Truth, 2016)
1. Podcast: Transformation of European Politics
2. Documentary: Radicals Rising
Episode 15 with Tamar Mitts (by Tarik Abou-Chadi)
Radicals Rising (2016): The Naked Truth, S1E3 (by Mariana van Zeller)

Branton & Kropf (2026)

Research Question: Does exposure to hate crime shape how African Americans vote?

Theory and mechanisms

Who else is affected, and how?

Hate crimes are “message crimes” also affecting other minority group members

Which emotions are most likely?
Joy / happiness
Sadness
Surprise
Disgust
Fear / anxiety
Anger
What behaviors do these emotions promote?
Fear / anxiety
risk aversion + information seeking
Anger
risk-taking
What should we expect for minority participation, given each emotion?
↓
(de)mobilization
(e.g., abstaining)
⚡
mobilization
(e.g., voting)
How to participate while avoiding risk?
↓
Branton & Kropf’s resolution: vote-by-mail
participate and avoid risky in-person spaces
Why is that still costly?
↓
Why shifting to VBM is costly
  • In-person voting is often social: e.g. church groups go together to vote early in person (before Election Day)
  • Mail ballots from Black (and Latino) voters are more often rejected
    • Hispanic-surname ballots more likely rejected for signature mismatch than for lateness (Herndon et al. 2026)
    • Election officials less likely to reply (and reply less accurately) to Latino-sounding email inquiries about voting (White et al. 2015)
  • More paperwork/rules: signatures, envelopes, postage, strict deadlines, i.e., easier to make a mistake
  • Less help from get-out-the-vote efforts organized around early-vote sites and Election Day
  • Greater distrust of newer / less familiar voting methods relative to established in-person voting

Hypothesis: higher county hate crime targeting African Americans → higher probability of VBM (vs in-person)

Vote method as a function of hate crimes

Branton and Kropf 2026 Figure 2: predicted probability of vote-by-mail rises with hate crimes

Predicted probability of VBM rises with both the number and per capita rate of local hate crimes

Impact of hate crime on VBM

Branton and Kropf 2026 Table 1: logistic models of hate crime and vote-by-mail

Higher county hate crime (raw count or per capita) raises the probability that African Americans vote by mail

Critique / open questions

  • Only voters: Y is vote mode among people who already voted. Demobilization into non-voting may still happen. Different sub-groups of minority groups may demobilize (not turn out at all), mobilize safely (VBM), or mobilize (turnout in person + other risk-taking actions)
  • Mechanism unobserved: no direct measures of fear vs anger, perceived safety, or awareness of local hate crimes
  • Open question: What would we expect in contexts where VBM is not an option?

Roman & Thompson (2024)

Research question: Does indirect exposure to violence against LGBTQ+ people make the mass public more prosocial toward LGBTQ+ people?

Exercise: stylized expectations

How does the public react in response to violent events against marginalized groups?

Roman and Thompson 2024 Figure 1: stylized expectations for no shift, durable shift, and FPVR model after violence against marginalized groups

Split into three groups (≈5 min):

  1. Each group takes one panel (a / b / c)
  2. Formulate 2 reasons why that expectation should materialize

Stylized expectations

Roman and Thompson 2024 Figure 1: stylized expectations for no shift, durable shift, and FPVR model after violence against marginalized groups

(a) No shift
  • Weak agenda-setting if coverage is thin (Birkland 1998)
  • Dominant-group esteem from minority marginalization (Tajfel & Turner 1982; Cikara et al. 2014)
  • Empathy gap / no shared experience (Sirin et al. 2021; Cikara et al. 2014)
  • Episodic frames: troubled individual, not societal queerphobia (Iyengar 1994; Ott & Aoki 2002; Zahzah 2019)
(b) Durable shift
  • Focusing events mobilize attitudes (Birkland 1998)
  • Sympathetic elite/media messaging (Zaller 1992; Iyengar 1994; Harrison & Michelson 2017)
  • Illegitimacy → sympathy/empathy → prosocial attitudes (Harth et al. 2008; Branscombe & Miron 2004)
  • Prior durable shifts after analogous stimuli (Broockman & Kalla 2016; Reny & Newman 2021; Oskooii et al. 2021)
(c) FPVR (fickle)
  • Issue-attention cycle: salience fades (Downs 1972)
  • Need sustained sympathetic messaging (Zaller 1992)
  • Short-term impression management, not genuine change (Harth et al. 2008; Nguyen et al. 2021; Sears 1993)
  • Countervailing queerphobic norms + episodic frames (Vuletich & Payne 2019; Ott & Aoki 2002)

Fickle Prosocial Violence Response model

Roman and Thompson 2024 Figure 2: Fickle Prosocial Violence Response model linking salient illegitimate violence to temporary prosocial attitudes

SSM support increases post-Pulse

Pulse: June 2016 massacre at Pulse, an LGBTQ+ nightclub in Orlando

Roman and Thompson 2024 Figure 3: balance, Pulse effect on same-sex marriage support, and falsification tests, with panel b framed

Post-Pulse respondents more supportive of same-sex marriage

Pulse effect attenuates over time

Roman and Thompson 2024 Figure 4: influence of Pulse on same-sex marriage support attenuates over time

The prosocial shift is short-lived, consistent with the “fickle” FPVR prediction

Anti-gay attitudes after Pulse

Roman and Thompson 2024 Figure 5: post-Pulse reduces anti-gay attitudes across bandwidths

Roman and Thompson 2024 Figure 6: post-Pulse reduction in anti-gay attitudes attenuates over time

Top: post-Pulse cuts implicit and explicit anti-gay bias (D-score, straight bias, heterocentrism) across bandwidths.
Bottom: that reduction holds for ~50 days, then fades. Same “fickle” pattern as SSM support

Critique/Open questions?

Local, low-salience crimes by out-group members can spark majority retaliation (Riaz et al. 2024), while majority prosociality toward minorities seems to need highly salient events (Roman and Thompson 2024).

Synthesis

Exercise: sort the readings

In pairs (≈5 min): place today’s readings. Any earlier required readings that also fit?

whose reaction is observed
Minority group
Majority group
who discriminates/enacts violence
Bottom-up
hostility
Bottom-up discrimination → minority outcomes
Bottom-up discrimination → majority outcomes
Top-down
hostility
Top-down discrimination → minority outcomes
Top-down discrimination → majority outcomes

Exercise: sort the readings

whose reaction is observed
Minority group
Majority group
who discriminates/enacts violence
Bottom-up
hostility
Bottom-up discrimination → minority outcomes
  • Mitts (2019): anti-Muslim hostility → online ISIS radicalisation
  • Branton & Kropf (2026): anti-African American hate crime → vote-by-mail
Bottom-up discrimination → majority outcomes
  • Roman & Thompson (2024): salient civilian anti-LGBTQ+ violence → short-term prosocial attitudes
Top-down
hostility
Top-down discrimination → minority outcomes
  • Mitts (2019): far-right vote share (>15%), i.e., far-right elites in power → online ISIS radicalisation
Top-down discrimination → majority outcomes
  • Romarri (2020): Italian far-right mayors → hate crimes against immigrants
  • Müller & Schwarz (2023): Trump’s anti-Muslim tweets → anti-Muslim hate crime
  • Allinger (n.d.): Italy’s nomad-camp policy → hate crimes against Roma
  • Levy & Levy (2017): US state gay/lesbian rights policy → anti-LGBT hate crime

Outlook

Next week: reducing hate crime I — contact

Week 9 (Mon 9/Wed 11 Nov): intergroup contact as a remedy?

Required readings

  • Mousa, S. (2020). Building social cohesion between Christians and Muslims through soccer in post-ISIS Iraq. Science, 369(6505), 866–870.
    • Scope Conditions. (n.d.). The promise and limits of intergroup contact, with Salma Mousa [Audio podcast episode]. Scope Conditions Podcast.
  • Alrababa’h, A., Marble, W., Mousa, S., & Siegel, A. A. (2021). Can exposure to celebrities reduce prejudice? The effect of Mohamed Salah on Islamophobic behaviors and attitudes. American Political Science Review, 115(4), 1111–1128.
  • Scacco, A., & Warren, S. S. (2018). Can social contact reduce prejudice and discrimination? Evidence from a field experiment in Nigeria. American Political Science Review, 112(3), 654–677.

Weekly comments due Sunday 8 Nov, 11:59 pm.

Questions?

Fossil Fuel Treaty advocacy slide: TCD endorsement and petition QR code