Download presentation
Presentation is loading. Please wait.
Published byLoren Parks Modified over 6 years ago
1
Effects of User Similarity in Social Media Ashton Anderson Jure Leskovec Daniel Huttenlocher Jon Kleinberg Stanford University Cornell University Avia (ilouz) cohen
3
Abstract Many on-line social applications include mechanisms for users to express evaluations of one another, or of the content they create. What affects the types of such evaluations?
4
question-answering site
Dataset Description Wikipedia free encyclopedia Stack Overflow question-answering site Epinions reviewing site
5
Dataset Ways of evaluation
Direct evaluation Indirect evaluation E T E T
6
Which factors influence evaluations?
Similarity Similarity of interests Similarity of social ties Relative status User similarity in
7
Which factors influence evaluations?
Similarity Similarity of interests Similarity of social ties Relative status User similarity in
8
Similarity of interests Definition
Action User uβs binary action vector Vector of length M with 1 in the iβth position if u took action i at least once, and 0 otherwise. Alternative definition relevant on Stack overflow: uβs binary tag vector. Similarity of users u and v: π π’,π£ = π’ β π£ π’ β π£
9
Similarity of interests Example
u v uβs binary action vector: π’ = 1,0,1,0,0 vβs binary action vector: π£ =(0,0,1,1,1) u and vβs similarity: π π’,π£ = π’ β π£ π’ β π£ = β 3 = 1 6
10
Which factors influence evaluations?
Similarity Similarity of interests Similarity of social ties Relative status User similarity in
11
Similarity of social ties Definition
User can be characterized by the users he has evaluated User uβs binary evaluation vector Vector of length M with 1 in the iβth position if user u evaluated user i, and 0 otherwise. Social similarity: π π’,π£ = π’ β π£ π’ β π£
12
Similarity of social ties Example
u v uβs binary evaluation vector: π’ = 0,0,1,1,0 vβs binary evaluation vector: π£ =(0,0,1,1,0) u and vβs similarity: 1 2 3 4 5 π π’,π£ = π’ β π£ π’ β π£ = β 2 =1
13
How does similarity influence evaluations?
What do you think?
14
And the answer isβ¦
15
Similarity vs. P(+) Wikipedia
Relationship between similarity and π(+) is monotonically-increasing. π· + is the fraction of positive evaluations in a given set of evaluations.
16
Similarity vs. P(+) Stack Overflow
Strength of relationship depends on the similarity notion used: Tag similarity Evaluation similarity π· + is the fraction of positive evaluations in a given set of evaluations.
17
Similarity vs. P(+) Epinions
Similarity of interests Produces similar effects as Stack Overflow. Similarity of social ties Turns out to be too sparse to be meaningful.
18
So far. similarity is correlated with P(+). Next
So far similarity is correlated with P(+). Next how both similarity and status affect evaluations.
19
Which factors influence evaluations?
Similarity Similarity of interests Similarity of social ties Relative status User similarity in
20
Status Definition User uβs status: total number of actions u has taken
User uβs (absolute) status: π π’
21
Differential status Definition
Evaluation behavior doesn't vary with target status alone, but with the evaluator and targetβs statuses together. Differential status: Ξ= π πΈ β π π
22
Status Example u v uβs absolute status: π π’ = 1,0,1,0,0 2 =2
π π’ = 1,0,1,0,0 2 =2 vβs absolute status: π π£ = 0,0,1,1,1 2 =3 u and vβs differential status: Ξ= π π’ β π π£ =β1
23
π« vs. π·(+) Wikipedia For low-similarity pairs:
β < 0: votes are about 80% positive β > 0: votes are about 55% positive As similarity increases, votes get more positive across all values of β. Ξ= π πΈ β π π π· + is the fraction of positive evaluations in a given set of evaluations.
24
π« vs. π·(+) Stack Overflow
When β > 0 The higher similarity is, the higher P(+) is. Same as Wikipedia. When β < 0 The lower similarity is, the higher P(+) is. Opposite order of similarity curves. Why does this happen? Ξ= π πΈ β π π
25
π« vs. π·(+) Stack Overflow
Why Does this happen? Reputation system of Stack Overflow costs a small amount of reputation to βdown-voteβ. A disincentive to downvote is strongly felt by users with lower status. Ξ= π πΈ β π π
26
π« vs. π·(+) Stack Overflow
This effect disappears when removing low-status evaluators ( π πΈ <100).
27
So far similarity controls the extent to which differential status influences evaluations.
Next relationship between similarity and status.
28
Similarity vs. Ξ Wikipedia elections
Selection effect Users with higher status tend to vote on targets who are active in the same areas as they are.
29
Similarity vs. Ξ Stack Overflow elections
Evaluator-target pairs are more similar as Ξ moves away from 0. No selection effect
30
Similarity vs. Ξ Wikipedia and Stack Overflow
Hypothesis: This happens due to the different contexts that evaluations occur on Stack Overflow and Wikipedia. Ranges of similarity in both graphs are very small, thus differences between plots are not as big as it seems.
31
Is π« an appropriate way to compare statuses?
32
Absolute status vs. π·(+) Stack Overflow
The P(+) curves are approximately flat and donβt depend on π π . P(+) depends on Ξ, the higher the β range is, the lower P(+) is. π π
33
Summary Main insights P(+) is positively correlated with similarity.
β plays a major rule for low-similarity pairs. Selection effect on Wikipedia, but not on Stack Overflow. β is a correct measure for relative status.
34
How are similarity and status useful?
36
Ballot-blind prediction
Knowledge of similarity and status of first voters alone provides enough information to successfully predict the final result.
37
Ballot-Blind prediction Definition
The task Predicting administrator promotion on Wikipedia from early voters The method By looking at properties of first few voters and their similarity with the candidate. Without looking at the sign of the votes.
38
Ballot-Blind prediction Experimental setup
Election (π, β¦, π
) Target Set of votes Result β 1, β1 Vote π£ π β β¦ ( πΈ π , π π , π‘ π ) Evaluator Vote β 1, β1 Time of vote
39
Ballot-Blind prediction Experimental setup
English Wikipedia 7,600 voters 120,000 votes 2,953 candidates 3,422 elections k = 5: looking at the first 5 voters
40
Ballot-Blind prediction Classes of features
Simple Summary Statistics (S) πππ ( π π ) π = π=1 π π πΈ π , π π π₯ = π=1 π ( π πΈ π β π π ) π β-s quadrants Partition β-s space into four quadrants (s=0.025 in Wikipedia).
41
Ballot-Blind prediction Baselines
Baseline B1: logistic regression classifier S statistics 4 features of β-s quadrants Baseline B2: probability π π π=π π π π threshold Voterβs positivity π π - his historical fraction of positive votes Baseline GS βgold-standardβ: best possible performance π π π=π π π π
42
Ballot-Blind prediction Main assumptions
User positivity π π indicates about the default voting behavior. Default behaviorβs changes depend on π₯ π , π π . π π₯ π , π π is the average deviation of π π in the β π , π π bucket, compared to the overall fraction of positive votes across the entire dataset.
43
Ballot-Blind prediction First method M1
Adjusting π π according to average deviation of π π in the β π , π π bucket, across the entire population. Model π πΈ π =1 = π π +π( π₯ π , π π ) π π π=π π π πΈ π =1 Threshold
44
Ballot-Blind prediction Second method M2
Adjusting π π β π , π π based on effects of s and β on π π β π , π π for both individual and global levels. Model π πΈ π =1 =πΌΒ· π π β π , π π + 1βπΌ Β·π β π , π π π π π=π π π πΈ π =1 Threshold π π β π , π π is πΈ π βs positivity in the β π , π π bucket.
45
Ballot-Blind prediction Results on English Wikipedia
B2 without S does as well as B1 with S. Models M1 and M2 outperform B1 and B2.
46
Ballot-Blind prediction Results on German Wikipedia
Baseline B2 performs about as well as B1. Models M1 and M2 outperform B1 and B2. Model M2 with S features gives performance that is halfway to even the gold standard.
47
Ballot-Blind prediction Conclusions
How evaluations depend on where they fall in the β-s space? Low evaluator-target similarity with evaluatorβs higher status Low evaluator-target similarity with targetβs higher status High evaluator-target similarity with evaluatorβs higher status High evaluator-target similarity with targetβs higher status More predictive Less predictive
48
Conclusions Similarity is positively correlated with on-line users evaluation. Similarity controls the effect of Ξ on user evaluations. Similarity and status have a predictive power (ballot blind-prediction).
49
Questions?
50
for your attention!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.