Online Social Dynamics and
Well-being
Munmun De Choudhury
Assistant Professor, School of Interactive Computing, Georgia Tech
March 12, 2014
Photo courtesy NPR
Three problems
Photo courtesy of NIDA.gov
(De Choudhury, Counts, Horvitz, CSCW 2013a)
Honorable Mention Award
Research Question:
Examine patterns of activity and
emotional correlates for childbirth and
postnatal course leveraging activity in
social media (Twitter)
(De Choudhury, Counts, Horvitz, CSCW 2013a)
Blue line represents approximate time of childbirth.
The red line represents mothers and the green line
represents the background cohort
(De Choudhury, Counts, Horvitz, CSCW 2013a)
(De Choudhury, Counts, Horvitz, CSCW 2013a)
mothers w/ large effects
mothers w/ small effects
background cohort
size of vocabulary
(#unigrams)
percentage of unigrams
w/ p < .01
17,117
33,785
47,214
27.65%
19.74%
3.86%
percent of all unigrams in the language vocabulary used that changed significantly in usage frequency
after childbirth, i.e. with p < .01 based on paired sample t-tests
background cohort
mothers w/ small effects
mothers w/ large effects
now (↓), shit (↑), back (↑), that (↑), day
(↓), life (↑), time (↓), them (↑), me (↑),
you (↑), fuck (↑), today (↓), sleep (↑),
tonight (↓), love (↓), good (↓), here(↓),
her (↓), morning (↑), tomorrow (↑), go
(↑), know (↑), him (↓), people (↓)
#past (↑), duh (↑), people (↓),
photo (↑), post (↑), decision (↓),
reunite (↓), women (↑), story (↑),
time (↑), asap (↓), do (↑), life (↓),
wait (↑), fired (↑), days (↑), happy
(↓)
haha (↓), blessed (↑), lol (↓),
#lifecangetbetter (↑), awesome
(↓), monthly (↑), fantastic (↓),
cuddle (↑), home (↑), love (↓), sick
(↑), aww (↑), scary (↑)
top unigrams showing the most change (in usage frequency) in the postnatal period, compared to the
prenatal phase
(De Choudhury, Counts, Horvitz, CSCW 2013a)
(De Choudhury, Counts, Horvitz, Hoff, CSCW 2014a)
Facebook & Postpartum Depression
Web survey to recruit mothers who were Facebook users and
who gave birth to a child within the last nine months or less
To incentivize participation, mothers were entered into the
random drawing of four $500 Amazon gift cards
Survey was active between mid-July and mid-September,
2012
Advertised through:
•
•
•
•
•
•
•
•
•
•
Mailing list of new mothers at Microsoft
Neighborhood based mommy blogs in the Seattle metro area
Postings from our organization’s official Twitter and Facebook accounts
Our personal Twitter, Facebook and Google+ accounts
Paid Facebook ads targeting mothers in the age group 20-39 years
Sponsored posts on BabyCenter (babycenter.com)
(De Choudhury, Counts, Horvitz, Hoff, CSCW 2014a)
Patient Health Questionnaire (PHQ-9)
• the 9-item questionnaire seeks responses over the past two week period
• based directly on the nine diagnostic criteria for major depressive disorder in the DSM-IV
(Diagnostic and Statistical Manual Fourth Edition)
• scores on the PHQ-9 range from zero to 27; individuals with scores 15 or above are
considered to be moderately severe to severely depressed (Kroenke et al. 2001)
Not at
all
Several
days
More than half the
days
Nearly
everyday
Little interest or pleasure in doing things
0
1
2
3
Feeling down, depressed, or hopeless
0
1
2
3
Trouble falling or staying asleep, or
sleeping too much
0
1
2
3
Feeling tired or having little energy
0
1
2
3
Poor appetite or overeating
0
1
2
3
Feeling bad about yourself — or that you
are a failure or have let yourself or your
family down
0
1
2
3
Trouble concentrating on things, such as
reading the newspaper or watching
television
0
1
2
3
Moving or speaking so slowly that other
people could have noticed? Or the
opposite — being so fidgety or restless
that you have been moving around a lot
more than usual
0
1
2
3
(De Choudhury, Counts, Horvitz, Hoff, CSCW 2014a)
Prediction Task
Dev.
LL
psd. R2
sfit
N
Error df
Prenatal
+10d +20d
+30d
Postnatal
61.32 57.39
52.58
46.61
58.15
-30.668 -26.3 -20.34 -17.24
-28.84
0.355 0.383
0.439
0.484
0.372
0.72
0.68
0.62
0.57
0.703
156
156
156
156
156
100
100
100
100
100
• The prenatal period does provide PPD-predictive information together with a
brief period of postnatal observations.
• Aligns with findings in the clinical literature where prepartum depression is
known to be a good indicator of PPD (Beck 2001).
(De Choudhury, Counts, Horvitz, CSCW 2013a)
Honorable Mention Award
Low-cost, privacy-preserving
mechanisms to identify new mothers’
behavior can improve social support
and encourage postpartum wellness
(De Choudhury, Gamon, Counts, Horvitz, ICWSM 2013)
(De Choudhury, Gamon, Counts, Horvitz, ICWSM 2013)
(De Choudhury, Gamon, Counts, Horvitz, ICWSM 2013)
Theme
Symptoms
Unigrams
anxiety, withdrawal, severe, delusions, adhd, weight, insomnia,
drowsiness, suicidal, appetite, dizziness, nausea, episodes, attacks,
sleep, seizures, addictive, weaned, swings, dysfunction, blurred,
irritability, headache, fatigue, imbalance, nervousness, psychosis,
drowsy
Disclosure
fun, play, helped, god, answer, wants, leave, beautiful, suffer, sorry,
tolerance, agree, hate, helpful, haha, enjoy, social, talk, save, win,
care, love, like, hold, cope, amazing, discuss
Treatment
medication, side-effects, doctor, doses, effective, prescribed, therapy,
inhibitor, stimulant, antidepressant, patients, neurotransmitters,
prescriptions, psychotherapy, diagnosis, clinical, pills, chemical,
counteract, toxicity, hospitalization, sedative, 150mg, 40mg, drugs
Relationships, life
home, woman, she, him, girl, game, men, friends, sexual, boy,
someone, movie, favorite, jesus, house, music, religion, her, songs,
party, bible, relationship, hell, young, style, church, lord, father,
season, heaven, dating
(De Choudhury, Gamon, Counts, Horvitz, ICWSM 2013)
 mean frequency—the average
measure of the time series of a
feature at any given day:
µi=(1/N)∑tXi(t).
 variance—the variation in the
time series: (1/N)∑t(Xi(t) −µi)2.
 mean
momentum—relative
trend of a time series,
compared to a period before:
(1/N)∑t(Xi(t)-(1/(t-M))∑(M≤k≤t1)Xi(k)).
 entropy—the
measure
of
uncertainty in a time series:
−∑tXi(t)log(Xi(t)).
(De Choudhury, Counts, Horvitz, WebSci 2013)
SMDI (t ) 
nd (t )  d ns (t )  s

d
s
(De Choudhury, Counts, Horvitz, WebSci 2013)
(De Choudhury, Monroy-Hernandez, Mark, CHI 2014)
Best Paper Award
The Atlantic
(De Choudhury, Monroy-Hernandez, Mark, CHI 2014)
(De Choudhury, Monroy-Hernandez, Mark, CHI 2014)
(De Choudhury, Monroy-Hernandez, Mark, CHI 2014)
Goal:
Study affective responses in social
media and how they might indicate
desensitization to violence
experienced in communities embroiled
in an armed conflict
(De Choudhury, Monroy-Hernandez, Mark, CHI 2014)
Towards the beginning of the time period of analysis, i.e., early on in 2010 or early 2011, the peaks in
number of homicides are actually correlated with those in NA. However over time, especially in
2012, that ceases to be the case
(De Choudhury, Monroy-Hernandez, Mark, CHI 2014)
Over time, from the trends of activation, we observe a general increase, implying that Twitter users
mentioning the four cities in their postings, were increasingly using higher intensity emotions.
Dominance shows a rise with persistent violence (ref. the slope of the linear fit), indicating that users
are increasingly using dominating and aggressive emotions
borderlandbeat.com