VISUALIZATION MAPPING FOR IMPROVED DIGITAL PUBLIC HEALTH - - PowerPoint PPT Presentation

APPLYING NETWORK SCIENCE AND VISUALIZATION MAPPING FOR IMPROVED DIGITAL PUBLIC HEALTH COMMUNICATION

Moderated by Brittany Seymour, DDS, MPH Public Health Consultant, MIT Faculty Associate, Berkman Klein Center for Internet and Society at Harvard Assistant Professor, Harvard School of Dental Medicine

A Story about fluoride

Dental Caries

Top 10 0 Public lic Hea ealth h Achi hievemen ements ts of the e 20th

th

Cen entur ury

• 1. Vaccines
• 2. Motor vehicle safety
• 3. Safer workplaces
• 4. Control of infectious diseases
• 5. Decline in deaths for coronary heart disease and

stroke

• 6. Safer and healthier foods
• 7. Healthier mothers and babies
• 8. Family planning
• 9. Fluoridation of drinking water
• 10. Recognition of tobacco use as a health hazard

Source: Centers for Disease Control and Prevention

A Story about Fluoride

A Story about Fluoride

A Story about Fluoride

Seymour et al, AJPH, 2015

Media Cloud

■ What other tools do: reach and impact metrics ■ What Media Cloud does: influence ■ An open platform, joint project of the Harvard Berkman Klein Center for Internet and Society and the MIT Center for Civic Media

Presenters

“Media Influence and Framing the Public Health Narrative Around Unarmed Deaths, Race, and Gun Violence: Mapping the Trayvon Martin Case” Rahul Bhargava, MS Research Scientist, MIT Media Lab MIT Center for Civic Media “Vaccine hesitancy in digital networks: trust and social proof online” Hal Roberts, ts, MS Fellow, Berkman Klein Center for Internet & Society Harvard University “Could Fragmented Communication Networks Reshape the Narrative?: Evidence from Tobacco and e-Cigarette Media Coverage” Laura Gibso son, n, PhD Research Director, Tobacco Center on Regulatory Science University of Pennsylvania

MEASURING MEDIA ATTENTION TO GUN AND POLICE VIOLENCE

Rahul Bhargava Research Scientist, MIT Media Lab MIT Center for Civic Media @rahulbot rahulb@mit.edu

Ferguson, MI 8/17/2014 Minneapolis, MI 12/3/2015 Baltimore, MD 4/29/2015 New York City, NY 11/28/2014

A PUBLIC HEALTH APPROACH

Academic Responses

NEJM, 12/2016 JAMA, 2013 Public Health, 2015

Public Health Approaches

Mozaffarian D, Hemenway D, Ludwig DS. Curbing Gun ViolenceLessons From Public Health Successes. JAMA. 2013;309(6):551-552. doi:10.1001/jama.2013.38

Two examples of how this might work in action.

We can help this effort by mapping the media narratives to identify processes and actors.

• 1. Looking for changes in media coverage of

police violence against unarmed people of color.

• 2. Understanding the fight over how the media

told the story of the Trayvon Martin case.

POLICE VIOLENCE AGAINST UNARMED VICTIMS OF COLOR

Nathan Matias, Natalie Gyenes, Allan Ko, Ethan Zuckerman, Rahul Bhargava, Hal Roberts

Data Sources

Mapping Police Violence The Guardian Washington Post

Michael Brown’s death as a key y event ent in recent history of police killings of unarmed people of color.

Sources: Media Cloud articles mentioning victims in the 5 days before, and two weeks after death (n=717,871).

Media coverage of these incidents changed after Michael Brown was killed.

Ne News ws ar arti ticles cles per unarm armed ed bl blac ack k person son killed lled by US police lice in in the days surrounding each death, including before & after the death of Michael Brown

Sources: Mapping Police Violence, Washington Post, Guardian 01/01/2013 - 06/29/2016 (n=333). Media Cloud articles in the two week period after death, normalized by total Media Cloud article count. 10

• bservations within 14 days before Michael Brown’s death are omitted.

An unarmed black person killed by US police received 10.5x .5x th the inci cide dence nce rat ate

• f Ne

News ws Ar Arti ticl cles es after Michael Brown’s death from those killed before

Sources: Mapping Police Violence, Washington Post, Guardian 01/01/2013 - 06/29/2016 (n=333). MediaCloud articles in the two week period after death, normalized by total MediaCloud article count. NB Model controls for age, gender, and regional population. p=7.84e^-14

Among unarmed black people killed by US police, the number of news articles per person for June 2016 was not significantly different from Jan 2013

Sources: Mapping Police Violence, Washington Post, Guardian 01/01/2013 - 06/29/2016 (n=333). MediaCloud articles in the two week period after death, normalized by total MediaCloud article count. NB Model controls for age, gender, and regional population.

Articles about an unarmed black person killed by US police received 49.8x 9.8x th the incidence cidence rat ate e of Total tal Facebook acebook Sha hares es after Michael Brown’s death

Sources: Mapping Police Violence, Washington Post, Guardian 01/01/2013 - 06/29/2016 (n=333). Facebook Shares to articles in the two week period after death, normalized by total Media Cloud article

• count. NB Model controls for age, gender, and regional population. p=2.56e^-12

Articles about an unarmed black person killed by US police received 49.8x 9.8x th the incidence cidence rat ate e of Total tal Facebook acebook Sha hares es after Michael Brown’s death

Sources: Mapping Police Violence, Washington Post, Guardian 01/01/2013 - 06/29/2016 (n=333). Facebook shares of articles in the two week period after death, normalized by total MediaCloud article

• count. NB Model controls for age, gender, regional population, and article count.

A Useful Approach for a Public Health Response

1.

• 1. Identif

entification cation: key event 2.

• 2. Awaren

areness ess: sustained social media interest 3.

• 3. Pr

Prevent ntion ion: easier to talk about

THE TRAYVON MARTIN CASE

Erhardt Graeff, Matt Stempeck, Ethan Zuckerman

A fight over the media narrative about this boy’s death.

Data Sources

Media Cloud n=8,643 Bit.ly clicks n=1,233,899 PageOneX Newspaper front pages n=1.91 daily avg The Internet Archive n=2,764 Google Trends n=25k daily avg Tweets n=374,690 Change.org Petition n=2,038,557

A media story best told in 5 acts

Act 1: Not a Story Act 2: Building Pressure Act 3: National Exposure Act 4: Political Agenda War Act 5: Tabloid Court Case

Act ct 1: Not a Story

Feb 26 – Mar 6

Act ct 2: 2: Building Pressure

Mar 7 – 15

Act ct 3: 3: National Exposure

Mar 6 – 22

Act ct 4: 4: Political Agenda War

Mar 23 – Apr 10

Act ct 4: 4: Plot 1: The Left

The Left goes after the “American Legislative Exchange Council”.

Act ct 4: 4: Plot 2: The Right

The Right goes after Trayvon Martin’s image; labelling him a “drug dealer”.

Act ct 5: 5: Tabloid Court Case

Apr 11 – 30

A Useful Approach for a Public Health Response

1.

• 1. Traj

ajec ector

• ry: from a clip to national story

2.

• 2. Pi

Piggy-ba back ck: activists shifting to their cause 3.

• 3. Tran

ansmedia smedia: analyzing across media

CONCLUSION

Public Health Responses

• Changing media depictions
• Media education campaigns
• Publicizing support options
• Nonprofit advocacy

VACCINE HESITANCY IN DIGITAL NETWORKS: TRUST AND SOCIAL PROOF ONLINE

Hal Roberts Fellow, Berkman Klein Center for Internet & Society Harvard University

Research team

Brittany Seymour

Harvard School of Dental Medicine

Rebekah Getman

Harvard School of Dental Medicine

Mohammad Helmi

Harvard School of Dental Medicine

Alfa Yansane

Harvard School of Dental Medicine

Hal Roberts

Berkman Klein Center, Harvard University

David Cutler

Department of Economics, Harvard College

Vaccine Hesitancy

Vaccine Support

Media Cloud Topic Mapper Spider

1. Search Media Cloud archive for ‘vaccine*’ in U.S. sources from 2014-06-01 – 2015-03-01. 2. Parse all links from those ~14,000 stories. 3. Download all unrecognized urls and test for the presence of ‘vaccin*’ in the text of each. 4. Add any matching stories to the topic. 5. Repeats steps 2. – 4. fifteen times. 6. Analyze resulting set: 49,144 stories, 53,092 story links, 4,817 media sources, 20,979 links between media sources

Link Network

Influential Sources

Source rank nk Overall ll Pro vaccin ine Vaccine ne hesit itant nt Healt lth h and Scienc nce Mains nstream eam Media ia 1 cdc.gov scienceblogs.com ncbi.nlm.nih.gov sciencedirect.com nbcnews.com 2 ncbi.nlm.nih.gov sciencebasedmedicine.org nvic.org chemport.cas.org msnbc.com 3 Wikipedia.com Wikipedia.com mercola.com cdc.gov livescience.com 4 ageofautism.com leftbrainrightbrain.co.uk naturalnews.com apps.weofknowledge.co m nytimes.com 5 scienceblogs.com rationalwiki.org whale.to springer.com washingtonpost.com 6 youtube.com braindeer.com youtube.com who.int npr.org 7 nytimes.com pediatrics.aappublications.org greenmedinfo.com jid.oxfordjournals.org latimes.com 8 sciencedirect.com feeds.feedburner.com medalerts.org nature.com huffingtonpost.com 9 naturalnews.com

• racknows.blogspot.com

healthimpactnews.co m jama.jamanetwork.com twitter.com 10 10 fiercevaccines.com theness.com sanevax.org tandfonline.com cnn.com

Primary Science Authority

Community Authority

Vaccine Sentiment

PRO

• A story should be categorized as having PRO-vaccine sentiment if any of the following are true:
• the author presents a pro-vaccine argument in her own voice;
• the story describes only pro-vaccine arguments or scientific research;
• the story mentions vaccines in a way that presumes vaccines in general are a public good;
• the story describes efforts by the government or other institutions to vaccinate without critiquing the effectiveness of the

vaccines;

• the story presents pro-vaccine arguments as coming from the scientific establishment but presents anti-vaccine arguments

as coming only from anti-vaccine advocates who are not scientists or doctors;

• the story presents a critical accounting of only anti-vaccine advocates or their tactics without any counter- balancing anti-

vaccine arguments. Some examples of PRO-vaccine stories:

• Large New Study Confirms That Childhood Vaccines Are Perfectly Safe

http://thinkprogress.org/health/2014/07/01/3455131/childhood-vaccine-study-safe/ [ pro-vaccine argument ]

• Has human progress stalled? And if so, what can we do about it? http://www.aei.org/publication/progress-stalled-can/ [

presumes vaccine as public good ]

• US approves first bird flu vaccine for people http://www.newscientist.com/article/dn11626-us-approves-first-bird-flu-

vaccine-for-people.html [ describes government vaccination ]

• "The way God intended" - California parents are having "measles parties" instead of vaccinating their kids

http://www.salon.com/2015/02/10/the_way_god_intended_california_parents_are_having_measles_parties_instead_of_ vaccinating_their_kids/ [ critical of anti-vaccine advocates ]

Overall Sentiment

ANTI 15% NA NA 16% NONE 10% PRIMA MARY 13 13% PRO RO 46% 46%

70,21% of all ANTI stories located in Anti Vax link community 66.67% of stories in MSM link community are PRO

Conclusions

• The Vaccine controversy is strongly segregated into vaccine

hesitant, pro vaccine, public health, and mainstream media content.

• ANTI vaccine content is a small part of the overall debate.
• Almost all ANTI vaccine content is located within the vaccine

hesitant link community.

• Mainstream media publishes mostly PRO and little ANTI

vaccine content.

• The vaccine hesitant community relies mostly on subverting

the traditional mode of scientific authority by citing primary research, whereas the pro vaccine community relies mostly

• n Wikipedia’s community mode of authority.

Thanks!

• This study was funded by the Pershing Square Fund for

Research on the Foundations of Human Behavior

• Media Cloud project led Ethan Zuckerman at the Center for

Civic Media at the MIT Media Lab and Yochai Benkler at the Berkman Klein Center for Internet & Society at Harvard University

• Funding by the Ford Foundation, the Open Society

Foundations, the John D. and Catherine T. MacArthur Foundation, Bill and Melinda Gates Foundation and the Robert Wood Johnson Foundation.

COULD FRAGMENTED COMMUNICATION NETWORKS RESHAPE THE NARRATIVE?: EVIDENCE FROM TOBACCO AND E- CIGARETTE MEDIA COVERAGE

Laura Gibson Research Director, TCORS University of Pennsylvania

Research team

Robert Hornik Laura Gibson Matt O’Donnell Sharon Alajajian Sherry Emery

Health Media Collaboratory NORC at UChicago

Faculty Consultants

Joseph Cappella, Paul Allison, Catherine Maclean, Janet Audrain-McGovern

Post-doctoral fellows

Kirsten Lochbuehler, Alisa Padon, LeeAnn Sangalang, Candy Yang

Doctoral students

Michelle Jeong, Kwanho Kim, Elissa Kranzler, Stella Lee, Jiaying Liu, Danielle Naugle, Leeann Siegel, Allyson Volinsky

More tobacco news coverage, less teen smoking 2001-2003

Proliferation of sources

The idea of the Public Communication Environment

Scanning Measures

PCE

Broadcast News Major newspapers AP

Specific media use by individuals

Websites Entertain- ment TV Tobacco/ e-cig ads Facebook fan pages YouTube comments Twitter

Exposure

Personal conversation Cognitions/ Behaviors

Changing tobacco environment

This study

■ Do all of the sources move together? – [Test: Does coverage in websites popular among youth and young adults vary together with coverage in traditional media sources over time?] ■ Do they move together regardless of topic? – [Test: Is it the same for tobacco and e-cigarettes?]

Sources included

AP National State/Local Top 50 Newspapers 8 Broadcast TV/Radio News Channels

Top 50 Popular Websites each Quarter

• 1. Finding texts

smoking, smoker(s), smokes tobacco, nicotine hookah(s) cig(s), cigar(s), cigarette(s), cigarillo(s), ecig(s), ecigar(s), ecigarillo(s), ecigarette(s), vaping, vape*

• 2. Finding tobacco & e-cig texts

Sample 1 Sample 2

Content in sources May 2014-Jun 2016

Source ce

All Texts xts Tob Tob NPM E-cig ig E-cig ig NPM AP 8,454 6,053 (72%) 803 (9%) News 53,630 42,693 (80%) 2,731 (5%) BTN 5,417 3,611 (67%) 223 (4%) Web 78,061 51,427 (66%) 3,709 (5%) Total 145,562 103,784 7,466

Passing mentions

■ Scott Disick Parties in Mexico With Three Models, Is No Longer Interested in Recovery | E! Online – “…While Kourtney Kardashian's former partner was busy on the phone and drinking water in this particular photo, his vacation has not been entirely sober. Earlier in the week, Disick was photographed drinking what appeared to be a beer and a mixed drink while also smoking a cigar arette…” ■ More food, less hibernation mean bigger NC bears – “…Farmers shifting from inedible crops like tobacco cco and cotton to edible items like corn and soybeans give the bears plenty to eat so they no longer have to expend so much energy and lose weight hibernating, state Wildlife Resources Commission black bear biologist Colleen Olfenbuttel said…” ■ Not passing mentions – At least 3 tobacco-related words in a paragraph & in more than

• ne sentence

– Or 1 tobacco-related word in the title

Content in sources May 2014-Jun 2016

Source ce All Texts xts Tob Tob NPM E-cig ig E-cig ig NPM AP 8,454 6,053 1,319 (22%) 803 624 (78%) News 53,630 42,693 4,831 (11%) 2,731 1,772 (65%) BTN 5,417 3,611 416 (12%) 223 118 (53%) Web 78,061 51,427 7,619 (15%) 3,709 2,516 (68%) Total 145,562 103,784 14,185 7,466 5,030

Note: NPM = Not passing mentions only

Results over time – Tobacco

5 10 15 20 25 Web: Average daily proportion - Tobacco only 2 4 6 8 10 01jul2014 01jan2015 01jul2015 01jan2016 01jul2016 AP BTN News Web

Results over time – E-cig

5 10 15 20 25 Web: Average daily proportion - E-cig 2 4 6 8 10 01jul2014 01jan2015 01jul2015 01jan2016 01jul2016 AP BTN News Web

Removing linear and weekly cyclical trends

■ Weekly trends – AP & Websites have less coverage on the weekend – Newspapers have more coverage on Sundays ■ Linear time trends – For example, some decrease in newspaper coverage when San Diego Union Tribune bought by LA Times

Example: De-trended & aggregating 3-days

# News ws Stories

• ries

Pre redi dict cted Residual siduals 3-da day y aver erage ge Sun, 6/22/14 16 10.1 5.9 2.4 Mon, 6/23/14 4 5.6

• 1.6

Tue, 6/24/14 10 7.1 2.9 Wed, 6/25/14 7 8.6

• 1.6

2.1 Thu, 6/26/14 14 9.3 4.7 Fri, 6/27/14 11 7.8 3.2

Do sources move together?

Tobacco r 95% CI E-cigare arett tte r 95% CI Topic c differe rence nce

Web-AP .36 [.25, .46] .35 [.24, .45]

Same

Web-News .31 [.19, .41] .60 [.52, .67] Ecig > Tob Web-BTN .14 [.02, .25] .59 [.50, .66] Ecig > Tob News-AP .30 [.19, .41] .48 [.38, .57] Ecig > Tob News-BTN .17 [.05, .29] .57 [.48, .65] Ecig > Tob AP-BTN .06 [-.07, .18] .30 [.18, .41] Ecig > Tob Note: Correlations among residuals of topic mentions (> passing mention) in media sources with 3-days as the unit of analysis (N=258)

E-cig Web–News (r = .60)

• 5

5 10 15 20 01jul2014 01jan2015 01jul2015 01jan2016 01jul2016 Web Newspapers

Usefulness of websites

■ Chosen because popular among our target population ■ Include non-news sites ■ Even among news websites, contributing unique information relative to traditional sources – Average inter-correlation with corresponding specific News or BTN source, Tobacco = .24, E-cigarette = .47

Top 50 Popular Websites each Quarter

Summary

■ Coordination across sources (not different for Web) ■ Coordination stronger for e-cigarettes across all sources – More tobacco coverage (mostly in passing) – More variation across time in e-cigarette coverage ■ Still value in adding online sources popular among youth and young adults

Next steps

■ Valence of coverage – Don’t know if sources are coordinated in framing of e-cigs ■ Themes (addiction, health, policy, youth) – Perhaps sources emphasize different themes ■ Connect to survey outcomes – Test if content coverage is associated with youth reported exposure & outcomes over time

Thank you!

Laura Gibson lgibson@asc.upenn.edu

Research reported in this presentation was supported by the National Cancer Institute (NCI) of the National Institutes of Health (NIH) and FDA Center for Tobacco Products (CTP) under Award Number P50CA179546. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the Food and Drug Administration (FDA).

Exploratory Theories

■ Decentralization of information systems/ dismantling of traditional information authorities ■ Key events lead to peaks of attention and

• pportunities for framing/reframing an

issue ■ Language usage and linking behaviors create online “communities”

Take Home Message:

General public is not merely a target audience, but also a strong contributing player in shaping online narratives

Image source: Goel, et al. The structural virality of online diffusion. 2013

Broadcast Social Diffusion

Public Health Communication?

(Scientific proof) (Social Proof)

“Community water fluoridation:

• pen discussions strengthen public

health.”

• Alfred

edo

• Morabia

ia, MD, PhD, Editor-in-Chief of AJPH, Washington, DC.

■ Questions? ■ Thank you!