Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
June 7, 20111
A research project funded by the European Commission’s7th Framework Programme
PrimeLife SUMMITJune 7, 2011
the Social Web: Privacy and Information Quality
Ronald LeenesTilburg University
June 7, 2011
agenda
1. privacy in social networks
2. quality of information on web 2.0
2
June 7, 20115
PrimeLife inception
access control in web 2.0
???
project start
June 7, 2011
social factors
6
June 7, 2011
it’s the architecture, stupid
7
June 7, 20118
PrimeLife inception Clique
access control in web 2.0
???
project start
June 7, 2011
our challenge
11
June 7, 201112
privacy
data disclosure
data protection
sociability
orand
June 7, 2011
Teen sacked for 'boring' job Facebook comment
15
TheRegister, 26 Feb 2009
June 7, 201116
June 7, 2011
Murder 'followed Facebook change'
17
Hayley Jones had described herself as 'married' to Brian Lewis
A mother-of-four was murdered by her partner days after she changed her Facebook status from "married" to "single", a court has heard.
http://news.bbc.co.uk/2/hi/uk_news/wales/8232250.stm
June 7, 2011
context matters™
18
June 7, 201137
‘social’
Ronald
colleages
family
runnerscollection
contactsfamily
mom grandpgrandmaClair
colleagues
Dave SeanClaire Eric
runners
…
June 7, 201138
June 7, 2011
To support users of social networking sites we created the first tool called Scramble! that allows encryption of texts on users’ profile sites. Clique, an extension of the popular social networking site software Elgg is the second tool. It enables users to segregate between different audiences for their data. For users and providers of phpBB forums we developed a phpBB extension as the third tool. The extension upgrades the access control features of the forum software so that users instead of administrators can define who should have access to their own contributions.
close friends
Dave SeanClaire Eric
June 7, 201144
June 7, 201149
fake medical wiki
task: answer questions about 5-HT
using information in wikisearch term ->
[search results <-> read info]
-> answer questions
complete survey
June 7, 201150
June 7, 201151
June 7, 2011
selected result by position in list
52
n=1440
15,0
30,0
45,0
60,0
1 2 3 4 5 6
percent
June 7, 2011
selected search result by indicator
53
0
10,0
20,0
30,0
40,0
0 1 2 3 4 5 6
occu
patio
nal
title titl
e +
reputa
tion
name
reputa
tion m
easu
re
websit
ena
me + tit
le
certif
icatio
n n=90
percent
June 7, 2011
first experiment: conclusions
54
! test subjects by and large pick first search result! … occupational title is strongest indicator! name (and title) weak indicator
! good from a privacy point of view
! in general, source of information was not deemed important
June 7, 2011
conclusions continued
55
! test subjects believe they base their choice on information quality! in practice they pick the first search result
! indicators reported as important are! professional writing! perceived text relevance! lack of spelling mistakes
June 7, 201156
experiment 2:using ratings
June 7, 201157
fake medical wiki
task: answer questions about 5-HT
using information in wikiread a lemma
[observe rating, explore individual reviews]
-> answer questions
complete survey
June 7, 201158
June 7, 201159
June 7, 2011
second experiment: conclusions
60
! test subjects do notice the stars! they report correct number
! test subjects did not inspect the reviews! may be design related
! text quality is appreciated in accordance with match of rating and actual text quality! neg rating for low quality text ! lower perceived quality! pos rating for high quality text ! higher perceived
quality
June 7, 2011
conclusions
61
! technology is part of the problem & solution:! context matters, architecture too
! demonstrators! simple, user friendly tool embedded in workflow: Clique! general tool using PKI: Scramble!
! determining information quality ill understood! lessons incorporated in architecture for reputation
system
June 7, 201162
A research project funded by the European Commission’s7th Framework Programme
PrimeLife SUMMITJune 7, 2011
thank you
June 7, 2011
search result selection preferences
65
! first result pickers! role of indicator vs preference for spot 1 is unknown
! other result pickers ! preference for indicator, but unknown # of subjects
missing due to preferred option being in spot 1! shows same pattern as first result pickers
June 7, 201166
rating text quality
reviews
experimental stars# reviewsreviews
low negexperimental stars# reviewsreviews
low
pos
experimental stars# reviewsreviews
high neg
experimental stars# reviewsreviews
high
pos
control none low
June 7, 201167
conditioncondition
Frequency Percent
Valid experimental 1a low pos 20 16,1Valid
experimental 1b high neg 16 12,9
Valid
experimental 1c low neg 20 16,1
Valid
experimental 1d high pos 25 20,2
Valid
control 1a high 38 30,6
Valid
control 1b low 5 40,0
Valid
Total 124 100
June 7, 201168
condition * How many stars did the text about 5-HT receive?condition * How many stars did the text about 5-HT receive?condition * How many stars did the text about 5-HT receive?condition * How many stars did the text about 5-HT receive?condition * How many stars did the text about 5-HT receive?condition * How many stars did the text about 5-HT receive?condition * How many stars did the text about 5-HT receive?How many stars did the text about 5-HT receive?
How many stars did the text about 5-HT receive?
TotalHow many stars did the text about 5-HT receive?
How many stars did the text about 5-HT receive? 1 3 4 5
condition experimental 1a low pos 0 1 1 12 14conditionexperimental 1b high neg 11 2 1 0 14
condition
experimental 1c low neg 13 2 0 0 15
condition
experimental 1d high pos 0 1 3 14 18
condition
control 1a high 2 2 0 1 5Total 26 8 5 27 66
June 7, 2011
“friends”do people care?
June 7, 201170
! do people use audience segregation on Facebook and Hyves?
! if so, who, what and why?! control for age, gender, personality (introvert/
extravert)
June 7, 201171
! 1163 (Dutch) participants (906 complete), avg age 25, 65% female
! awareness of technical access control measures,but …
! hardly any use in practice (18%)! reasons: hassle (22%), how? (19%), never bothered
(57%)! AC users: females, privacy concerned, older
June 7, 201172
<Section title>