Transcript Slide 1

Managing the Privacy of Incidental Information During Collaboration
Kirstie Hawkey and Kori Inkpen
Concept
Ad hoc
Collaboration
Incidental
Information
Colleagues often gather
around a computer
Many traces of past
activities are visible with
casual inspection.
Individual
Behaviours
Privacy
Dalhousie University
{hawkey, inkpen}@cs.dal.ca
Challenges
Complexity Reduction
Privacy Patterns
Patterns on a per-window basis suggest a
semi-automated approach to privacy
management may be feasible:
Rapid Bursts of
Browsing
Normative privacy
for personal
displays does not
apply; the display is
an object in the
collaboration.
Privacy concerns
and personal
information
management styles
are very individual. Personal Information
Large Volume of
Pages
Management
Sequential
Representation
Managing large amounts of
personal information is
complex.
Data Analyses
 High volume of logged data during field studies.
 Numerical averages insufficient due to individual
differences.
 Data mining and visualization techniques may uncover
patterns.
System Evaluation
Streaks
Transitions
Field Study: Privacy Gradients
 Large volume of information.
 Multiple contexts of creation and viewing of incidental
information.
 Must balance amount of control with maintenance time
and effort.
 Privacy patterns suggest a semi-automated approach
is feasible.
 Field setting required for evaluation of management
solutions.
 To encourage natural behaviour, users need to
maintain privacy.
 Large volume make fine-grained self-reports tedious.
Next Steps
Temporal
Representation
Temporal
Patterns
Potential Solutions
Window
Re-visitation
 Browser windows of differing privacy levels?
 Filter which incidental information is displayed.
 Classify new information as it is generated.
Research studies
Methodology
Goodness of Fit
 Field study for one week.
 20 laptop users (multiple
contexts).
 Client-side logging.
 Electronic diary.
 Privacy gradients.
 Paper classification tasks.
 75% of gradients fit most
of the time; 20% fit all of
the time
 Hard to classify:
 Sites with multiple
purposes
 Sites with variable
content
Clusters
C1
Privacy Gradient
Overall
C2
C3
C4
Final Cluster Centers
Public
42% 22%
36%
62%
18%
Semi-Public
25% 58%
21%
16%
28%
15% 9%
36%
11%
9%
18% 11%
7%
11%
46%
5
10
2
Private
Don’t Save
Number of Participants
3
Overall
Privacy
Gradient
Usage
Patterns
 Survey of privacy concerns with respect to incidental
information (underway).
 Field study examining privacy in context of location
and pages (underway).
 Longitudinal field study of management solution
Acknowledgements
Thanks to the members of the EDGE Lab
for their support. This research is
funded in part by NSERC