1
Understanding Users and
Relationships in Social Networks
Jen Golbeck iSchool
2
Predicting Personality from
Facebook with Cristina Robles, Karen Turner
Jennifer Golbeck, Cristina Robles, and Karen Turner.
2011. Predicting personality with social media. In
Proceedings of the 2011 annual conference extended abstracts on Human factors in computing systems (CHI
'11). ACM, New York, NY, USA, 253-262.
The Big Five Personality
Traits
3
Agreeableness
4
The Test
• You can take the test – and help our next project if you use Twitter!
• http://tinyurl.com/TwitterPersonality
• Test is a series of statements
• Subjects rate how much they believe each statement applies to themselves
5
Features Considered
Language
Features
Personal
Info
Activities &
Preferences
Internal FB
Stats
Composite
Features
Structural
Linguistic Features
Swear Words
Social Processes (e.g. Mate, talk, they, child)
Human Words (e.g. baby, man)
Affective Processes (e.g. Happy, cried, abandon)
Positive Emotions (e.g. Love, nice, sweet)
Anxiety Words (e.g. Worried, fearful, nervous)
Perceptual Processes (e.g. Observing, heard, feeling)
Seeing Words (e.g. View, saw, seen)
Biological Processes (e.g. Eat, blood, pain)
Ingestion Words (e.g. Dish, eat, pizza)
Work Words (e.g. Job, majors, xerox)
Money Words(e.g. Audit, cash, owe)
Open Consc.
Extra Agree Neuro
0.006
-0.171
0.032
-0.084
0.01
0.078
0.105
0.264
0.203
-0.009
0.091
0.07
0.136
-0.022
-0.05
0.203
-0.12
-0.142
-0.062
0.038
0.052
0.044
0.045
0.117
0.167
-0.013
-0.15
0.008
0.101
0.192
-0.04
-0.195 -0.163
-0.027
0.06
-0.227
-0.112
0.013
0.096
0.067
-0.014
-0.098
0.134
-0.161
0.042
0.038
0.154
0.067
-0.05
0.029
0.031
0.207
0.096
0.154
0.048
-0.044
0.024
0.012
-0.006
0.029
Structural Features
Number of Friends
Egocentric Network Density
Activities and Preferences
Activities (char length)
Favorite Books (char length)
6
Personal Information
Relationship Status ( none listed,single, not single)
Last Name length in characters
-0.094
-0.152
0.115
0.158
0.093
0.012
-0.078
0.186
0.013
-0.069
0.05 -0.224 0.059
0.032
0.095
0.188
0.066
-0.145
-0.093
0.019
0.082
0.028
0.071
0.194
-0.111
0
0.04
-0.044
-0.036
0.184
7
0.006
-0.171
0.032
-0.084
0.264
0.091
-0.022
-0.142
0.07
-0.05
-0.062
-0.009
0.136
0.203
0.045
0.117
0.167
-0.013
-0.15
0.008
0.101
0.192
-0.04
-0.195 -0.163
-0.027
0.06
-0.227
-0.112
0.013
0.042
0.038
0.154
-0.05
0.029
0.031
0.207
0.096
0.154
0.048
-0.044
0.024
0.012
-0.006
-0.078
0.186
0.013
-0.069
0.05 -0.224 0.059
0.095
0.188
0.066
-0.145
-0.093
0.019
0.082
0.071
0.194
0.04
-0.036
0 -0.044
0.184
8
Results
Openness
Conscientiousness
Extraversion
Agreeableness
Neuroticism
Error
M5'Rules
0.099
0.104
0.138
Gaussian
Process
0.117
0.117
0.124
0.109
0.127
0.117
0.117
9
Computing Political
Preference
Among Twitter Followers with Derek Hansen
Jennifer Golbeck and Derek Hansen. 2011. Computing political preference among twitter followers. In
Proceedings of the 2011 annual conference on Human factors in computing systems (CHI '11). ACM, New
York, NY, USA, 1105-1108.
10
11
12
4/18/2020
– Fourth level
» Fifth level
Messages
• What user traits relate to your problem or goal? There are ways to profile that from social media.
• More info at http://cs.umd.edu/~golbeck
• golbeck@cs.umd.edu
14
Personality research funded by Army Research Lab Network Science CTA under Cooperative Agreement Number W911NF-09-2-0053