CoChat
Sign Up
Cited by 72Public Relations and Crisis CommunicationSentiment Analysis and Opinion MiningComplex Network Analysis Techniques

The DOI matches its Crossref record; 96 of the 97 references it cites were found in a registry or an index.

The ripple of fear, sympathy and solidarity during the Boston bombings

Open access
Lin, Yu-Ru, Margolin, Drew·2014·EPJ Data Science
DOI 10.1140/epjds/s13688-014-0031-z
Open this paper in CoChat — free accountOpen PDF

Abstract

The Boston Marathon bombing presents a rare opportunity to study how a disruptive event can trigger inter-communal emotions and expressions - where members of one community express feelings about and support for members of a distant community. In this work, we use over 180 million geocoded tweets over an entire month to study how Twitter users from different cities expressed three different emotions: fear , sympathy and solidarity , in reaction to the bombings. We capture spikes in fear in different cities by using sentiment and time-series analyses, and track expressions of sympathy and solidarity based on the emergent use of two hashtags, #prayforboston and #bostonstrong, widely adopted after the bombings. We find first that the extent to which communities outside Boston express these emotions is correlated with their geographic proximity, social network connections, and residents’ direct, physical experiences with Boston (captured by the number of citizens who had visited Boston recently). This general effect shows interesting differences across the different kinds of emotions, however. Analyses show that the extent to which residents of a city visit Boston is the best predictor of fear and solidarity expression, as well as a strong predictor of the expression of sympathy. The expression of fear is also directly related to the expression of solidarity. Our study has theoretical implications regarding the diffusion of information and emotional contagion as well as practical implications for understanding how important information and social support can be effectively collected and distributed to populations in need.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1140/epjds/s13688-014-0031-z

Cites

96 of 96 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Daily number of geocoded tweets (volume) in dataset. Tweets without geocodes are removed. From top to bottom: the total daily volume, the daily volume within the Boston area, and the daily volume within the direct affected region (DAR). The vertical dashed lines indicate the day of Boston bombings (April 15; black line) and the day of manhunt (April 19; gray line). The tweets posted on April 1, 10 and one half of the day on April 11 were missing due to data collection process errors. There are over 180 million geocoded tweets over an entire month and only tweets posted within the geographical borders of the cities (over 38 million tweets in total) are used in further analysis.
Figure 1 Daily number of geocoded tweets (volume) in dataset. Tweets without geocodes are removed. From top to bottom: the total daily volume, the daily volume within the Boston area, and the daily volume within the direct affected region (DAR). The vertical dashed lines indicate the day of Boston bombings (April 15; black line) and the day of manhunt (April 19; gray line). The tweets posted on April 1, 10 and one half of the day on April 11 were missing due to data collection process errors. There are over 180 million geocoded tweets over an entire month and only tweets posted within the geographical borders of the cities (over 38 million tweets in total) are used in further analysis.CC-BY · From the PDF
Figure 2 Sentiment indices over time in (a) the direct affected region (DAR) and (b) the City of Boston. The hourly sentiment indices of fear (and joy) before, during and after the Boston bombings are shown in smooth fitted curves over time (in UTC) with shaded area indicating a 95% confidence region. The vertical dashed lines around April 15 and 19 indicate the times of bombings and manhunt, respectively. The spikes in the level of fear correspond to the Boston bombings and the subsequent events, while the level of joy does not exhibit a sudden increase and is relatively stable over the period.
Figure 2 Sentiment indices over time in (a) the direct affected region (DAR) and (b) the City of Boston. The hourly sentiment indices of fear (and joy) before, during and after the Boston bombings are shown in smooth fitted curves over time (in UTC) with shaded area indicating a 95% confidence region. The vertical dashed lines around April 15 and 19 indicate the times of bombings and manhunt, respectively. The spikes in the level of fear correspond to the Boston bombings and the subsequent events, while the level of joy does not exhibit a sudden increase and is relatively stable over the period.CC-BY · From the PDF
Figure 3 Sentiment indices over time in (a) Boston, compared with those in (b) New York City, (c) Washington and (d) Chicago. Similar but slightly weaker spike patterns were observed in New York City, Washington and Chicago over the same time (in UTC).
Figure 3 Sentiment indices over time in (a) Boston, compared with those in (b) New York City, (c) Washington and (d) Chicago. Similar but slightly weaker spike patterns were observed in New York City, Washington and Chicago over the same time (in UTC).CC-BY · From the PDF
Figure 4 The expression of fear, sympathy and solidarity in relation to geo-distance, social ties and personal visits. ( a-c ) are scatterplots of the fear correlation (with square root transformation for normality) with Boston against geographic proximity (inverse distance in kilometers), number of replies, and number of personal visit, respectively. ( d-f ) are scatterplots of #prayforboston volume (total number of tweets containing the hashtag) against the three factors. ( g-i ) are scatterplots of #bostonstrong against the three factors. In all plots, non-US and US cities are colored in red and blue, respectively. The correlation coefficients are reported on top of the scatterplots, where (a-c) are Pearson correlation coefficients between the variables x and log( y ), and (d-i) are Spearman's rank correlation coefficients between the variables x and y . All coefficients are significant with p -value < 0.001. Among the 35 non-US cities, #prayforboston appeared in 25 cities, while #bostonstrong only appeared in 3 cities (Toronto, Montreal, and Melbourne). Cities with zero #prayforboston or #bostonstrong hashtags are not shown.
Figure 4 The expression of fear, sympathy and solidarity in relation to geo-distance, social ties and personal visits. ( a-c ) are scatterplots of the fear correlation (with square root transformation for normality) with Boston against geographic proximity (inverse distance in kilometers), number of replies, and number of personal visit, respectively. ( d-f ) are scatterplots of #prayforboston volume (total number of tweets containing the hashtag) against the three factors. ( g-i ) are scatterplots of #bostonstrong against the three factors. In all plots, non-US and US cities are colored in red and blue, respectively. The correlation coefficients are reported on top of the scatterplots, where (a-c) are Pearson correlation coefficients between the variables x and log( y ), and (d-i) are Spearman's rank correlation coefficients between the variables x and y . All coefficients are significant with p -value < 0.001. Among the 35 non-US cities, #prayforboston appeared in 25 cities, while #bostonstrong only appeared in 3 cities (Toronto, Montreal, and Melbourne). Cities with zero #prayforboston or #bostonstrong hashtags are not shown.CC-BY · From the PDF
Figure 5 Frequency of categories associated with the two hashtags. Most (83%) of the sampled tweets with #bostonstrong were given the code A1, A2 and A3. The majority of the sampled tweets with #prayforboston were given the code A4, and some were coded as A5, A6 and A7. In rare cases, a tweet cannot be assigned to one of the seven codes, which include: (Mu) Multiple codes : the coders agreed that more than one code can apply to the tweet; (No) Noagreement : the coders did not reach an agreement for the code(s) of the tweet; (Un) Uncodable : the coders did not understand the content of the tweet.
Figure 5 Frequency of categories associated with the two hashtags. Most (83%) of the sampled tweets with #bostonstrong were given the code A1, A2 and A3. The majority of the sampled tweets with #prayforboston were given the code A4, and some were coded as A5, A6 and A7. In rare cases, a tweet cannot be assigned to one of the seven codes, which include: (Mu) Multiple codes : the coders agreed that more than one code can apply to the tweet; (No) Noagreement : the coders did not reach an agreement for the code(s) of the tweet; (Un) Uncodable : the coders did not understand the content of the tweet.CC-BY · From the PDF
Figure 6 Out-of-sample R 2 for models of predicting shared fear. Models including the 'visit' predictor perform significantly better than others. Error bars show the bootstrapped 95% confidence intervals of the out-of-sample R 2 .
Figure 6 Out-of-sample R 2 for models of predicting shared fear. Models including the 'visit' predictor perform significantly better than others. Error bars show the bootstrapped 95% confidence intervals of the out-of-sample R 2 .CC-BY · From the PDF
Figure 7 Out-of-sample R 2 for models of predicting #bostonstrong. ( a ) Overall, the geo-social model performs significantly better than all other models. ( b ) The predictor 'visit' adds more predictability than 'geo' in predicting the #bostongstrong occurrences in farther cities (top 50% in terms of geo-distance). Error bars show the bootstrapped 95% confidence intervals of the out-of-sample R 2 .
Figure 7 Out-of-sample R 2 for models of predicting #bostonstrong. ( a ) Overall, the geo-social model performs significantly better than all other models. ( b ) The predictor 'visit' adds more predictability than 'geo' in predicting the #bostongstrong occurrences in farther cities (top 50% in terms of geo-distance). Error bars show the bootstrapped 95% confidence intervals of the out-of-sample R 2 .CC-BY · From the PDF
Figure 8 Out-of-sample R 2 for models of predicting #prayforboston. Models including the 'visit' or 'social' predictor perform slightly better than the baseline and geo models. However, the predictability of the two predictors is not differentiable. Error bars show the bootstrapped 95% confidence intervals of the out-of-sample R 2 .
Figure 8 Out-of-sample R 2 for models of predicting #prayforboston. Models including the 'visit' or 'social' predictor perform slightly better than the baseline and geo models. However, the predictability of the two predictors is not differentiable. Error bars show the bootstrapped 95% confidence intervals of the out-of-sample R 2 .CC-BY · From the PDF
Image 9
CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 97 references

Research works printed but not linked: 7

  1. Sutton J, Palen L, Shklovski I: Backchannels on the front lines: emergent uses of social media in the 2007 southern California wildfires. ISCRAM2008 2008, 624–632.
  2. Globe Staff (2013) 102 hours in pursuit of Marathon bombing suspects - The Boston Globe. Accessed 2014-04-10, [ https://www.bostonglobe.com/metro/2013/04/28/bombreconstruct/VbSZhzHm35yR88EVmVdbDM/story.html ]
  3. Markon J, Horwitz S, Johnson J (2013) Dzhokhar Tsarnaev charged with using ‘weapon of mass destruction’11. The Washington Post. Chap. National. Accessed 2013–09–15
  4. Feinsand M (2013) Boston Marathon bombings: Yankees play Neil Diamond’s ’Sweet Caroline,’ a staple at Fenway. Accessed 2014-04-10, [ http://www.nydailynews.com/sports/baseball/yankees-honor-boston-play-sweet-caroline-wake-bombings-article-1.1318805 ]
  5. Dudek M (2013) Chicagoans run to show solidarity with Boston Marathon victims - Chicago Sun-Times. Accessed 2014-04-10, [ http://www.suntimes.com/news/metro/19530547–418/chicagoans-run-to-show-solidarity-with-boston-marathon-victims.html ]
  6. Chew C, Eysenbach G: Pandemics in the age of Twitter: content analysis of Tweets during the 2009 H1N1 outbreak. PLoS ONE 2010., 5(11): 10.1371/journal.pone.0014118 · DOI
  7. Zimmer B (2013) ‘Boston Strong,’ the phrase that rallied a city. Boston Globe Zimmer B (2013) ‘Boston Strong,’ the phrase that rallied a city. Boston Globe

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Lin, Y.-R., & Margolin, D.
Affiliations
  • University of Pittsburgh (US)
  • Cornell University (US)
Corresponding author
Lin, Yu-Ru
ORCID
Venue
EPJ Data Science
Publisher
Springer Science and Business Media LLC
Published
2014-11-05
Volume
3
Issue
1
ISSN
2193-1127 (ISSN-L)
DOI
10.1140/epjds/s13688-014-0031-z
OpenAlex
W2149307575
Cited by
72
References
97
Citations by year
2026: 1 · 2025: 2 · 2024: 3 · 2023: 5 · 2022: 9 · 2021: 13 · 2020: 7 · 2019: 7 · 2018: 11 · 2017: 7 · 2016: 4 · 2015: 1 · 2014: 1
FWCI
12.77
Type
journal article
Language
en
Open access
Licence
Locations
  • EPJ Data Science (publishedVersion, cc-by)
  • D-Scholarship@Pitt (University of Pittsburgh) (acceptedVersion)
Funders
National Science Foundation (1423697)
Topics
  • Social Sciences › Communication › Public Relations and Crisis Communication
  • Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
  • Physical Sciences › Physics and Astronomy › Statistical and Nonlinear Physics › Complex Network Analysis Techniques
Keywords
Boston Marathon bombing, fear, sympathy, solidarity, emotional contagion, hashtags, Twitter, geographic proximity, disaster response, social support

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review