Transcript Co-presence Communities

Co-presence
Communities
Using pervasive computing to
support weak social networks
Jamie Lawrence
Terry Payne & David De Roure
DMC 2006
Introduction
http://eprints.ecs.soton.ac.uk/12684/
Focus on relating this work to the DMC
workshop (and WETICE in general)
• Weak Social Networks
• Co-presence
• Co-presence Communities
– Application
– Discovery Algorithm
– Worked Example
Weak Social Networks
• Weak & Informal Networks
– Familiar Stranger
– Communities of Practice
• Shrinking circle of “best friends”
• Weak relationships are important
– Familiar Strangers provide social support in times of
crisis
– CoP are vital sources of information and expertise in
an enterprise
• Ironically, weak relationships are often
– based on physical interaction
– least served by technological solutions.
Co-presence
• “sense that they are close enough to be
perceived in whatever they are doing,
including their experiencing of others, and
close enough to be perceived in this sensing
of being perceived” – Goffman
• “corporeal copresence” – Zhao’s taxonomy
– Natural state of co-presence: all parties are
physically proximate and present at the same
site.
Co-presence Detection
• Must correspond
to human sensory
limits
• Ego-centric
– Bluetooth
– IrDA badges
• Omniscient
– GPS tracking
Co-presence Communities
A group of people that you are usually
around during a particular time period
• People
• Time
• Context must be defined by the user
– regular meeting, sports club, lunch, coffee
break, Friday evening pints, …
Applications
• Ambient Information Dissemination
Environment (AIDE)
– Use Co-presence Communities to control the
flow of information
– For example, distributing a URL to the
“afternoon coffee crew”
• Context-aware computing
– Co-presence Communities can add context to
other information sources, e.g. diaries
• Building a social networking service from
real-world interaction data
Mining Algorithm Attributes
•
•
•
•
Incremental
Probabilistic
High-dimensional data
Error smoothing (missing values)
• Transform from…
– <start, end, device, device>
– <time, device, device>
• To…
– <~start, ~end, ~{devices}>
Mining Algorithm Overview
• Discretisation
– Produces groups of co-present devices at
each time interval
• Feature Extraction
– Finds periods of continuously similar copresence
• Clustering
– Cluster the co-presence periods across all
historical data
– The clusters provide the Co-presence
Community definitions
Discretisation
• Transform the co-presence events into
discrete time slots
• Useful if the data comes from multiple
sources
Feature Extraction
• Detects changes in the co-presence
membership
• Use a Laplacian of Gaussian (LoG) edge
detection routine averaged across devices
• Period boundaries occur at the zerocrossings
1
0
-1
Clustering
•
•
•
•
Clusters periods of co-presence together
Uses a implementation of COBWEB
Modified to accept Nominal Set attributes
The resulting clusters define the copresence communities
• Can be weighted to find temporal or
membership-stable communities
Worked Example
Interval 1 Interval 2 Interval 3 Interval 4
Day 1
B
B,C
Day 2
C
Day 3
D
Day 4
E
B,C
D
Day 5
F
B,C
D
B,C
D
D
B,C
D
Day 1: Periods
Day 1: Clusters
Day 2: Periods
Day 2: Clusters
Day 3: Periods
Day 3: Clusters
Day 4: Periods
Day 4: Clusters
Day 5: Periods
Day 5: Clusters
Conclusions
• Introduced the idea of Co-presence
Communities
• Discussed how they might capture weak
social networks
• Presented a method of discovering these
communities
• Demonstrated a simple example
Questions?
[email protected]
http://eprints.ecs.soton.ac.uk/12684/