thin slices of online profiles kristin stecher and scott counts
TRANSCRIPT
Thin Slices of Online Profiles
Kristin Stecher and Scott Counts
Realistic Accuracy Model (Funder, 1995)
Impression Formation Process
PerceiverInformation(Behavior/
Profile)
Target Trait
Thin Slices of Behaviors
“…the ability of our unconscious to find patterns in situations and people based on very narrow ‘slices’ of experience” - Gladwell (2005)
Teaching assistants (Ambady & Rosenthal, 1993)
• 6 seconds• No sound.
http://www.wjh.harvard.edu/~na/
Online
Hypotheses
H1: Users can make predictive inferences using condensed profiles.
H2: Certain attributes contribute more meaningfully to these profiles than others.
Part I, Brief Profiles“Your task is to find a friend in this social network.”
Compare sets of 3 profiles.
1. Choose an important attribute.2. View your selected attributes.3. Rank the profiles.
44 Participants
Part I, Brief Profiles
After 5 attribute rankings.
•Rate 1-100.
Part II, Full Profiles
View sets of profile comparisons again.
•Rank•Rate (1-100)
H1: Users can make predictive inferences using brief profiles.
Relationship between Part I (5 attributes) & Part II (full profile) ratings, r=0.404, p≤0.01.
Offline meta-analysis, r=0.39. (Ambady & Rosenthal, 1992)
H1: Users can make predictive inferences using brief profiles.
Participant ratings were intercorrelated after Part I, r=0.35, p<0.001.
Offline meta-analysis, r=0.20, 0.27. (Ambady et al., 1992)
H2: Some attributes contribute more meaningfully to brief profiles.
1. Perceived Utility: What attributes did participants choose to view?
2. Predictiveness: What attributes made the correlation between Part I- Part II ratings higher or lower?
3. Perceived Diagnosticity: What attributes caused participants to change their rankings?
Attribute Perceived Utility Attribute Perceived Utility
Photo 31.8% Gender 15.7%
Interests 29.1% Relationship Status 13.9%
Interested In 28.9% College 13.0%
About Me 26.4% Full Name 13.0%
Political Views 25.7% Concentration 12.7%
Activities 24.5% Status 12.7%
Looking For 22.5% Current Town 11.1%
Movies 21.4% Hometown 10.7%
Religious Views 21.4% Number of Friends 10.5%
Quotes 19.8% Last Updated 9.5%
Employer 18.9% Name 9.3%
Books 18.2% Networks 9.3%
TV Shows 16.6% Birthday 8.0%
Music 16.4% High School 6.6%
Number of Groups 6.1%
Attribute Predictiveness Attribute Predictiveness
Name 14.9 Interested In 17.6
High School 15.0 Full Name 17.7
Gender 15.7 Looking For 17.7
Photo 15.9 Political Views 17.7
Status 15.9 Movies 17.9
Number of Groups 16.1 Books 18.2
Activities 16.6 Concentration 18.6
About Me 16.7 Relationship Status 18.7
Interests 16.7 Music 18.8
Number of Friends 16.8 College 18.9
Hometown 17.0 Religious Views 18.9
Birthday 17.5 Networks 19.0
Current Town 17.5 Quotes 19.7
Employer 17.6 TV Shows 20.6
Last Updated 21.1
Attribute Diagnosticity Attribute Diagnosticity
College 2.75 Quotes 2.35
TV Shows 2.70 Interested In 2.34
Music 2.66 Books 2.33
Looking For 2.61 Relationship Status 2.27
Networks 2.58 Current Town 2.15
Movies 2.54 Activities 2.14
Gender 2.47 Photo 2.12
Religious Views 2.41 Last Updated 2.09
Employer 2.40 Full Name 2.07
Hometown 2.40 Status 2.02
Political Views 2.40 Concentration 2.00
Birthday 2.39 High School 1.98
Interests 2.37 About Me 1.92
Number of Friends 2.37 Number of Groups 1.89
Name 1.71
Predictiveness x Diagnosticity Trade-Off’s
y = 95.77x4 - 804.0x3 + 2490.x2 - 3360.x + 1709
1.5 2 2.5 30
10
20R² = 0.344334699301562
Attributes
Diagnosticity
Pred
ictiv
enes
sH
iLo
w
Name
Last Updated
High School
TV Shows
Gender
Manipulate Goal
Imagine you are entering a blogging community. Your task is to choose a blog within this community that you would like to subscribe to. You will choose a blog by viewing blogger profiles.
73 Participants
H1: Users can make predictive inferences using brief profiles.
Relationship between Part I (5 attributes) & Part II (full profile) ratings, r=0.30, p≤0.01.
H1: Users can make predictive inferences using brief profiles.
Participant ratings were intercorrelated after Part I, r=0.40, p<0.001.
Attribute Perceived Utility Attribute Perceived Utility
Photo 55.6% College 14.1%
Interests 35.5% Looking For 12.5%
About Me 34.9% Gender 12.3%
TV Shows 26.2% Relationship Status 12.2%
Activities 25.6% Networks 11.5%
Movies 24.1% Current Town 11.1%
Music 23.6% Birthday 9.0%
Interested In 22.9% Full Name 9.0%
Books 21.1% Status 8.6%
Quotes 21.0% Last Updated 8.5%
Political Views 18.2% Hometown 8.1%
Employer 17.4% Name 7.4%
Concentration 15.1% High School 7.3%
Religious Views 14.5% Number of Groups 7.1%
Number of Friends 5.6%
Attribute Predictiveness Attribute Predictiveness
Photo 17.8 Last Updated 20.8
Employer 18 Interested In 21
Hometown 18.3 College 21.1
Interests 18.3 Concentration 21.1
Current Town 18.7 Political Views 21.1
Movies 18.7 Quotes 21.1
Activities 18.8 Relationship Status 21.2
About Me 18.9 Birthday 21.4
Books 19 Number of Friends 21.5
TV Shows 19.4 Status 22.1
Networks 20 High School 22.4
Full Name 20.2 Name 22.8
Music 20.4 Religious Views 23
Gender 20.8 Number of Groups 23.1
Looking For 23.9
Attribute Diagnosticity Attribute Diagnosticity
Religious Views 2.78 Activities 2.16
Employer 2.62 Movies 2.09
Photo 2.55 About Me 2.08
Current Town 2.46 Full Name 2.05
Number of Groups 2.37 Interests 2.04
TV Shows 2.35 Music 2.04
Political Views 2.34 Interested In 2.02
Concentration 2.33 Name 1.99
High School 2.32 Networks 1.99
Last Updated 2.31 Gender 1.89
Birthday 2.3 Status 1.86
Books 2.27 Looking For 1.85
Number of Friends 2.26 Relationship Status 1.75
Hometown 2.19 Activities 2.16
Movies 2.09
H2: Some attributes contribute more meaningfully to brief profiles.
Relationship between domains– Perceived Utility, r=0.83, p<0.001.• Photo, Interests, About Me, Activities
– Diagnosticity, r=-0.02, p=ns– Predictiveness, r=0.05, p=ns• About Me, Activities, Interests, Photo
Suggestion
Thin Slices of Online Profiles
Kristin Stecher and Scott Counts