USEMP - value of personal data CAISE 14 presentation
2. • A few words about VELTI
• USEMP project & value of personal data
• USEMP use cases related to value
• USEMP framework for the value of personal
data
• USEMP LIO platform architecture
3. What VELTI does
VELTI is a global provider of mobile marketing &
advertising solutions that enable brands,
advertising agencies, mobile operators, and media to
engage with consumers via their mobile devices.
Consumers’ personal data privacy protection &
consumer consent is key to VELTI solutions
VELTI conducts marketing campaigns in over 67
countries across the globe
Most of the TOP-20 largest mobile operators
worldwide have run campaigns with VELTI
4. Marketing & value of personal data
Marketing is effective when products are offered to consumers
that they are interested in them thus more prone to purchase
them or some other related marketing objective
Mobile & online marketing offers access to recorded personal
data that can be used to infer consumers’ interest graph and
the ability to offer them related information instead of
broadcasting it (targeting)
Value of personal data for brands/advertisers/marketers
depends on:
• how much it can increase the probability of a consumer
purchasing their product or meeting the campaign objective
• what is the perceived value of the marketing objective
6. • 3-year EU-funded R&D project to explore
how to help & consumers understand the
use of their personal data in Online Social
Networks
– http://www.usemp-project.eu
• Consortium
– Multimedia and semantic experts
• CEA LIST (France); CERTH (Greece);
– Social & Legal research experts/Living Labs
• Iminds (Belgium); Radboud University (The
Netherlands); Luleå University of Technology (Sweden);
– Industrial partners
• VELTI (Greece); HWC (United Kingdom);
7. • A large majority of Europeans engage with Online Social
Networks (OSNs)
– 74% of users consider that they do not have sufficient control
– 70% are concerned with the way such data are handled by OSNs
• Upcoming EU General Data Protection Regulation –
harmonisation of EU’s legal framework and improvement of
users’ control over their shared data
• Asymmetry between data processing and control means
available to OSNs and those afforded by citizens
• Personal data sharing is a complex and pervasive process that
is still not well understood
• Work in different relevant fields is most often performed in
isolation
9. • Objective: raising awareness about data shared
online and improving user’s control of them
• (a) Real-time OSN presence management
– Development of semi-automatic privacy preservation
tools
– Joint analysis of volunteered, observed and inferred
data
• (b) Long-term OSN presence management
– Visualisation tool which summarizes the privacy
status
– Controls for quick personal data visibility change
10. • Analyse the existing and proposed legal framework of privacy
and data protection with regard to Online Social Networks
(OSNs)
• Advance the understanding of personal data handling through
in-depth qualitative and quantitative user research
• Create multimedia information mining tools adapted to
personal information management
• Build semi-automatic awareness tools to assist the users in
their interaction with personal data
• Contribute to the current debates related to the way personal
data should be monetised
• Propose an innovative living labs approach, adapted for
personal data handling in OSNs
11. • Objective: assist the user in understanding the
economic value of data shared online
• (a) Awareness of the Economic Value of
Personal Information
– Modelling of the personal data monetisation process
performed by OSNs
– Contribution to the transparency of OSN business
models
• (b) Personal Content Licensing
– Simulation of a framework for licensing personal
information
– Avoidance of commodification through an adapted
rewards mechanism
12. • Objective: assist the user in understanding the
economic value of data shared online
• (a) Awareness of the Economic Value of
Personal Information
– Modelling of the personal data monetisation process
performed by OSNs
– Contribution to the transparency of OSN business
models
• (b) Personal Content Licensing
– Simulation of a framework for licensing personal
information
– Avoidance of commodification through an adapted
rewards mechanism
13. # Name Description Monetary Value
A Demographics Personal details, eg. as
Gender, Age, etc.
High: advertisers wish to
target users of certain
demographic criteria
B Psychological Traits Defined by psychologists
(extraversion, openness,
etc.)
Low
C Sexual Profile Relationship status,
preferences, habits
High: advertisers wish to
target consumers based on
their relationship
status/lifestyle related to their
sexual profile
D Political Attitudes Supported politicians, parties
and stance
High: advertisers wish to
target consumers based on
the political affiliations since
these are related to their
general profile
E Religious Beliefs &
Cultural Heritage
Religion (if any) and beliefs Moderate:advertisers wish to
target consumers based on
their religious and cultural
beliefs
F Health Factors &
Condition
Habits (e.g. smoking,
drinking), medical
conditions, health factors
(exercise)
High: advertisers wish to
target consumers based on
their habits
G Location Characteristic locations of
the individual and history of
previous locations
High: advertisers wish to
target consumers based on
their current location or their
home location
H Consumer Profile Preferred products and
brands
High: advertisers wish to
target consumers based on
their consumer profile
attributes like the devices the
use to access digital content
14. 3 approaches are explored further:
collecting any explicit data about value ($) from records:
f.e: how much advertisers/marketers actually pay for
campaigns using personal data
modeling users utility functions for their personal data &
conducting focus groups/user tests:
f.e: what are the consumers perceived value of their personal
data
collecting & computing indicators of value (scores) related to
each consumer audience in Online Social network
f.e: an Online Social network user with larger audience has
larger value to other users (for audiences with similar
attributes)
15. • OSN users roles:
– user generated content producer
– content consumer
– part of a network of value
• OSN users generating value activities
– sharing content (images/posts)
– interest graph/sharing preferences
(likes/comments/follows/retweets/sharing)
• this has a network effect since it allows to better estimate other
members’ social graph interest graphs
– sharing demographic/location/contextual personal data
• this has a network effect since it generates also value for his
network
17. As a producer of user generated content in a social
graph I want to know what is the expected audience of
my activities related to a category that maybe of interest
to advertisers/marketers
[the more estimated audience I have the more valuable
my contribution would be]
[Lio should be able to compute an OSN's user audience
for different type of categories that maybe of interest to
advertisers based on past behaviour, klout-type of
algorithms can be used to estimate the audience]
[Lio should have access to the social graph data of the
consumer/producer of user generated content]
18. Weighted average of:
• Twitter repeats and mentions
• Facebook comments, wall posts and
likes
• Google+ comments, reshares and +1s
• LinkedIn Comments and Likes
• Foursquare tips
19. As a producer of user generated content I want to know what is
the ranking of my content base [affinity, weight, time decay] what
is have my comment shown to a friend to be consumed
[each post/like from my network affect the ranking of my posts to
my audience, the higher the rank to more of my audience the
better]
[Lio should be able to compute the potential ranking of a type of
post]
[Lio should be able to compute the potential ranking of a type of
post]
[Lio should have access to the social graph data of the
consumer/producer of user generated content]
21. Edge rank basics
-------------------------------------------------
• each user creates new objects e that will be displayed
to the news feed page of each of his/her friends in the
social graph
• the original creator of the object is the owner of the
object
• other users interact with the object
(share/comments/like)
• each time a user other than the owner interacts with
an object, a new edge is created for the object
• edge (user that created it, type of interaction, time)
-------------------------------------------------
the rank of an object e is defined:
-------------------------------------------------
edgerank(object e) = sum (edge score, over all edges)
each edge score is computed as the product of:
• affinity score between the user that has created the
action and the owner of the object
• weight of the action
• time decay
22. As a consumer/audience for advertisement/marketing
campaigns I want to know how value-able my shared or non-
shared personal data are in a given time for advertisers in a
OSN
[certain personal data maybe more valuable to advertisers than
others depending on the demand/offer market type of
dynamics]
[Lio should be able to compute an indication of what is the CPM
value for a campaign running on based on my share or NON-
share personal data I have included in OSN]
[Lio should have access to the advertising data/API of an OSN
to estimate the value of CPM for different filters/classifications
and to OSN user social graph ]
23. As a consumer of OSN application services I want to know how
my shared data may be valued by different OSN services
provider based on known values/classification schemes for
application providers
[different application providers classify their users in different
classes, we need a reference for this]
[Lio should provide my current ranking in terms of value for
different industries]
[Lio implementation requires access to OSN social graph data
and additional external data from application providers, this
needs to be explored further]