SLIDE 1
COMPUTATIONAL TIE STRENGTH: THEORY AND APPLICATIONS Eric Gilberts - - PowerPoint PPT Presentation
COMPUTATIONAL TIE STRENGTH: THEORY AND APPLICATIONS Eric Gilberts - - PowerPoint PPT Presentation
COMPUTATIONAL TIE STRENGTH: THEORY AND APPLICATIONS Eric Gilberts Prelim kevin casey lucas leslie TIE STRENGTH concept & impact The strength of a tie is a (probably linear) combination of the amount of TIME , the emotional INTENSITY ,
SLIDE 2
SLIDE 3
casey leslie lucas kevin
SLIDE 4
The strength of a tie is a (probably linear) combination of the amount of
TIME, the emotional INTENSITY, the INTIMACY (mutual confiding), and the
reciprocal SERVICES which characterize the tie. — Granovetter
TIE STRENGTH
concept & impact
STRONG TIES are the people you really trust. WEAK TIES, conversely, are merely acquaintances.
SLIDE 5
7,000+ papers cite TSOWT firms with right mix of ties get better deals strong ties can affect mental health
TIE STRENGTH
concept & impact
SLIDE 6
USING IT
in design & analysis
2 make better friend introductions. 1 build more informed privacy controls. 3 rethink social streams. 4 analyze large-scale, networked phenomena.
MODEL TIE STRENGTH TO…
SLIDE 7
USING IT
in design & analysis
2 make better friend introductions. 1 build more informed privacy controls. 3 rethink social streams. 4 analyze large-scale, networked phenomena.
MODEL TIE STRENGTH TO…
SLIDE 8
Tie strength is a blind spot in social media.
SLIDE 9
Tie strength is a blind spot in social media.
This thesis will show how to RECONSTRUCT it from digital traces, and how to apply it as a DESIGN and ANALYTIC tool.
SLIDE 10
SLIDE 11
HEURISTICS
a sample of substitutes
communication reciprocity [19] Friedkin 1980
- ne mutual friend [56] Shi, Adamic & Strauss 2007
communication recency [41] Lin, Dayton & Greenwald 1978 interaction frequency [22] Gilbert, Karahalios & Sandvig 2008
[17] Fischer, et al. 2006
SLIDE 12
HEURISTICS
the difference
CHI 2009 MODEL 88% Adj. R2 = 0.53 MSGS → FRIEND 62% Adj. R2 = 0.09 BASELINE 52%
SLIDE 13
COMPUTATIONAL TIE STRENGTH PART ONE
SLIDE 14
DIGITAL TRACES
- ur interactions leave marks
SLIDE 15
THE MAPPING PROBLEM
SLIDE 16
RESEARCH QUESTIONS
The literature suggests seven dimensions of tie strength:
INTENSITY, INTIMACY, DURATION, RECIPROCAL SERVICES, STRUCTURAL, EMOTIONAL SUPPORT and SOCIAL DISTANCE.
As manifested in social media, can these dimensions predict tie strength? In what combination? R1. What are the limitations of a tie strength model based SOLELY
- n social media?
R2.
SLIDE 17
THE DATA
- verview
2,184 assessed friendships
from 35 university students & staff described by 70+ numeric indicators
SLIDE 18
ASSESSING TIE STRENGTH
participant interface
SLIDE 19
STATISTICAL METHODS
SLIDE 20
87.6% accuracy
χ2(1, N = 4368) = 700.9 p < 0.001
THE MODEL
performance
1 1
+ +
μ μ
SLIDE 21
DELIVERABLES
part one
multicollinearity & structure tough dimensions … new techniques the predictors vs. accuracy tradeoff five flavors of tie strength
SLIDE 22
DESIGNING AROUND TIE STRENGTH PART TWO
SLIDE 23
RESEARCH QUESTIONS
How GENERAL is the tie strength model? Does it work in a new medium, like Twitter? R1. Can automatically-inferred tie strength usefully INFORM DESIGN in the real world? Can it help the collapsed context problem? R2. Can feedback from users IMPROVE the model? If so, how does it change? R3.
SLIDE 24
collision of social circles which normally remain separate
THE COLLAPSED CONTEXT PROBLEM
SLIDE 25
We Meddle demo
SLIDE 26
RT @Francesca_Elle teens don't tweet is a trending topic WTF ! Teens DO Tweet ... 2m ago
from R E D L O G
The Logo Evolution of Coca-Cola and Pepsi. 5m ago
from D A R I N G F I R E B A L L
Motorola Unveils Cliq: http://df.ws/dqx 14m ago
from D A R I N G F I R E B A L L
good morning yall! just got done eating breakfast and now im off too the gym! 8m ago
from J E R M A I N E O N E A L
Just read that Edward Hall died recently. His work on territoriality briefly mentioned in obit. Wow. http://bit.ly/ltUNL 45m ago
from J E N N T H O M
we meddle
BLOG | FAQ | LOGOUT
TWEET!
about us
WE MEDDLE arranges things just for you. You can emphasize either your inner circle
- r your outer circle. If we put somebody in
the wrong one, you can fix it: try hovering
- ver a profile icon. Also, we hope you enjoy
- ur filters above. We're in private alpha.
5
invite someone
twitter screenname, then press enter
New Typekit JS is live - just 1/10 the size! (about 2k for sites w/o badge. All sites soon). Republish your kits to upgrade. 57m ago
from J A S O N S A N TA M A R I A
If the bands opening the NFL season can use auto-tune, can the QB's use auto-pass? 100% completions, but no style at all? 1h ago
from G S L I S
The airport is opening its eyes. 1.5h ago
from YA P C H A I A N
more meddling
show only P OS I T I V E T W E E T S 7 show only N E G AT I V E T W E E T S 2 show only F R E Q U E N T P O S T E R S 9 show only I N F R E Q U E N T P O S T E R S 4 show only L IN K S 1 5
Jer Thorp (@blprnt): Visualizations of arts & culture budget cuts in British Columbia http://is.gd/37rfS 1h ago
from G E N E R ATO R X
SLIDE 27
[17] Fischer, et al. 2006 [62] Whittaker, et al. 2004.
RELATED SYSTEMS
socially-rendered interfaces
SLIDE 28
METHODS
how general is the model? can feedback improve it?
SLIDE 29
METHODS
can tie strength usefully inform design?
QUALITATIVE EVALUATION with We Meddle’s users
10 – 20 semi-structured interviews centered on usage and utility
SLIDE 30
TIE STRENGTH & DIFFUSION PART THREE
SLIDE 31
Dichotomous distinctions [edge exists or not] can sometimes be misleading … a weighted graph representation will frequently be more appropriate. More studies that assess the effectiveness of such approximations — and provide concrete, empirically validated guidelines for practice within particular problem domains — would be a welcome addition to the literature.
— 2009 Science article “Revisiting the Foundations of Network Analysis”
TIE STRENGTH
its potential in diffusion studies
SLIDE 32
DIFFUSION
every link is equal
Fowler & Christakis. 2008. Dynamic Spread of Happiness … BMJ. Ivkovic & Weisbenner. 2007. Information Diffusion Effects in Individual Investors' Common Stock Purchases … Rev. Fin. Kossinets, Kleinberg & Watts. 2008. The Structure of Information Pathways in a Social Communication Network. KDD. Onnela, et al. 2007. Structure and Tie Strengths in Mobile Communication Networks. PNAS.
SLIDE 33
WEIGHTED DIFFUSION
from Onnela, et al.
SLIDE 34
RESEARCH QUESTIONS
Does tie strength modulate the flow of information through a network? R1. Does tie strength interact with content as it flows through a network? R2. If tie strength regulates information flow through a network, how does this affect macroscopic properties of diffusion? R3.
SLIDE 35
METHOD
the retweet
SLIDE 36
METHOD
value-laden content
1 link per 700 tweets ⇒ 1000–1500 users
SLIDE 37