Post on 27-Jan-2015
description
Rethinking Location Sharing: Exploring the Implications of Social-Driven vs. Purpose-Driven Location Sharing
Karen P. TangJialiu Lin, Jason Hong, Dan Siewiorek, Norman Sadeh
Human-Computer Interaction InstituteSchool of Computer ScienceCarnegie Mellon University
Location-Based Services Are Here
2
Types of Location-Based Services
tracking personal trends (no sharing)
doing local searches (sharing with a service provider)
3
[google latitude] [yelp]
Location Sharing Applications (LSAs)
tracking personal trends (no sharing)
doing local searches (sharing with a service provider)
share locations with other people (a social network)
4
activecampus[griswold, ’03]
lemming[hong, ’04]
Past Research Examples of LSAs
5
2003 2004 2005 20082007 2009
esm study[consolvo, ’05]
reno[smith, ’05]
whereabouts[brown, ’07]
watchme[marmasse, ’04]
contextcontacts[raento, ’05]
connecto[barkhuus, ’08]
locaccino[sadeh, ’09]
1992
active badge[want, ’92]
activecampus[griswold, ’03]
lemming[hong, ’04]
Past Research Examples of LSAs
6
2003 2004 2005 20082007 2009
esm study[consolvo, ’05]
reno[smith, ’05]
whereabouts[brown, ’07]
watchme[marmasse, ’04]
contextcontacts[raento, ’05]
connecto[barkhuus, ’08]
locaccino[sadeh, ’09]
1992
active badge[want, ’92]
The most common use of the system was by the receptionist who routinely used it when forwarding telephone calls from the main switchboard.
Groups of people who regularly wanted to hold meetings could find each other easily with very little notice.
“
activecampus[griswold, ’03]
lemming[hong, ’04]
Past Research Examples of LSAs
7
2003 2004 2005 20082007 2009
esm study[consolvo, ’05]
reno[smith, ’05]
whereabouts[brown, ’07]
watchme[marmasse, ’04]
contextcontacts[raento, ’05]
connecto[barkhuus, ’08]
locaccino[sadeh, ’09]
1992
active badge[want, ’92]
Given mobile users’ fragmented attention, the time it takes to make a phone call must remain extremely short…These [context] cues [which include location] should facilitate decisions about whether to call, and if so, which communication channel to use.
“
activecampus[griswold, ’03]
lemming[hong, ’04]
Past Research Examples of LSAs
8
2003 2004 2005 20082007 2009
esm study[consolvo, ’05]
reno[smith, ’05]
whereabouts[brown, ’07]
watchme[marmasse, ’04]
contextcontacts[raento, ’05]
connecto[barkhuus, ’08]
locaccino[sadeh, ’09]
1992
active badge[want, ’92]
Phoebe wonders what she and her husband, Ross, will do for the evening, so she sends a location query to Ross. While he is waiting at the bus stop near his office, Ross sends a location update to Phoebe. Phoebe receives the message at home, eagerly anticipating Ross’ arrival home. When Ross gets off the bus, a location update is sent to Phoebe and she knows that he’s only 10 minutes away. She sets out dinner just in time for her husband’s arrival.
“
Common Themes for Past LSAs
driven by functional purposes:• coordination• collaboration• interruptibility• event planning
one-to-one sharing or small group sharing
9
Industry Trends for Information Sharing
integrated with online social networks (OSNs)• diverse networks, lots of weak links [wellman, ‘01]
• very large networks [donah, ‘04]
sharing is often not because one needs to share, but because one wants to share
driven by a social reason for sharing
10
11
Commercial Examples of LSAsmostly aimed at social-driven sharing
2005 2006 2009 20102007 2008
Commercial Examples of LSAsmostly aimed at social-driven sharing
12
2005 2006 2009 20102007 2008
“I'm just down the street!” Never miss another chance to connect when you happen to be at the same place at the same time. [facebook places]
Find out who’s around, what to do, and where to go. Introducing…the new Loopt so you can always stay connected… [loopt]
Share your location and stay connected with your friends. [plazes]“““
Reframing Location SharingPurpose-Driven Social-Driven
motivations
coordination, collaboration, interruptibility, planning
want (vs. need) to share, social awareness
features
one-to-oneclose-knit relationships
one-to-manydiverse relationship types
13
Understanding the DifferencesQ1: what are people sharing?will social-driven sharing lead to different sharing decisions?
Q2: how are making their sharing decisions?what privacy strategies are used in social-driven sharing?
Q3: are people making good choices?do people’s preferences result in privacy-preserving choices?
14
User Study: Participants2-week user study 9 participants, 3 female18-46 years old (μ=27.1, σ=8.3)
⅔ undergrad & grad students, ⅓ staff
15
User Study: Part 1 (in the field)participants given custom Nokia N95s
• treated as primary phone
collected continuous GPS tracesextracted significant places
• dwell time ≥ 5 mins
16
User Study: Part 2 (in the lab)1. shown a map of each place2. generate as many labels as possible
17
[sample labeling exercise given to everyone as training]
Heinz Field
Football field
Steelers vs. Bengals
downtown Pittsburgh
Steelers’ home
100 Art Rooney Ave
Near golden triangle
User Study: Part 2 (in the lab)purpose-driven scenario:
social-driven scenario:
18
User Study: Part 2 (in the lab)purpose-driven scenario:
social-driven scenario:
19
Analysis: Taxonomycoded each label:
20
Heinz Field
Football field
Steelers vs Bengals
downtown Pittsburgh
Steelers’ home
100 Art Rooney Ave
Near golden triangle
Analysis: Taxonomycoded each label:
21
Heinz Field
Football field
Steelers vs Bengals
downtown Pittsburgh
Steelers’ home
100 Art Rooney Ave
Near golden triangletype of description example
geographic100 Art Rooney AveNear Golden TriangleDowntownPittsburgh
semanticHeinz FieldSteelers vs. BengalsSteelers’ homeFootball field
hybrid Heinz Field @ downtown
Q1: What Do Users Share? [semantic]
social sharing preferences:• more semantic labels*
• fewer hybrid labels**
social sharing had different semantic labels**
• prefer activity & personal labels (“home”, “work”)
• purpose-driven sharing preferred type of place & business names (“coffee shop”, “Starbucks”)
22
*p<0.01**p<0.005
Q2: How Do Users Decide? [blurring]
insider knowledge“If I just say Giant Eagle [a regional grocery store chain], my friends will know which one I’m at.”
sharing activity vs. location“I’d rather say what I am doing than that I’m at a certain place.”
protecting friends’ locations“I’m uncomfortable sharing where I am at, since it’s someone else's place.”
23
Q2: How Do Users Share? [blurring intent]
purpose-driven: used to convey unavailability
social-driven: used to explicitly hide location
24
Q2: How Do Users Share? [blurring intent]purpose-driven: used to convey unavailability
social-driven: used to explicitly hide location…but also considered:
• social capital & image management• what would appear more interesting to others?
25
Q3: Do Users Make Good Choices?
examine 3 techniques for reverse engineering• google maps• google search + google maps• routines + google search + google maps
“bad” choice = physically locatable (stalker threat)
26
Result: Leaky Privacy Decisionspurpose-driven: easily locatablesocial-driven: susceptible to being located
27
resource(s) purpose-driven social-driven
map 50.0% 10.2%
map + web 62.3% 19.4%
map + web + routines 90.8% 51.0%
Summary & Conclusionsreframing: purpose- vs. social-driven sharingsignificant differences for social sharing:
• what: different types of disclosures [semantic]
• how: different intentions for blurring [to hide]
• how: considered social issues [impressions]
• actual privacy: still susceptible to attacks
28
Summary & Conclusionsreframing: purpose- vs. social-driven sharingsignificant differences for social sharing:
• what: different types of disclosures [semantic]
• how: different intentions for blurring [to hide]
• how: considered social issues [impressions]
• actual privacy: still susceptible to attacks
context for sharing is an important factor
29
Limitations & Future Workhypothetical disclosure scenariossmall, homogenous participant pool
• predominantly college students• already familiar social network users
comparing two extremes of location sharing• many other types of possible location sharing• one-to-one vs. one-to-many purpose-driven• one-to-many vs. one-to-one social-driven
30
Questions?Karen P. TangHuman-Computer Interaction InstituteSchool of Computer ScienceCarnegie Mellon University
kptang@cs.cmu.edu
This research has been supported in part by the National Science Foundation under grants CNS-0627513, IIS-0534406, and ITR-032535, by the CyLab at Carnegie Mellon University under grants DAAD19-02-1-0389 from the Army Research Office, by Nokia, by Portugal ICTI, and by a Microsoft Computational Thinking grant.
31