Social media (SM) platforms have demonstrated their ability to facilitate knowledge sharing on the global scale. They are increasingly often employed in educational and humanitarian domains where, despite their general benefits, they expose challenges peculiar to these domains. Specifically, the research context of this talk is directed by my participation in the Go-Lab European project and my collaboration with Médecins Sans Frontières (MSF) where SM platforms were used extensively. In this talk, we address four challenges regarding analytics, privacy, discovery, and delivery, aiming to answer corresponding four research questions. How to provide user-oriented analytics in knowledge sharing systems to support awareness and reflection? What privacy management interfaces and mechanisms are suitable for knowledge analytics and learning analytics? How to enable discovery of knowledge relevant to user interests? How to facilitate knowledge delivery into settings where Internet connectivity is limited or absent? Henceforward, we provide an overview of our results. Analytics. To enable awareness and reflection for an SM platform users, we propose the embedded contextual analytics model where the analytics is embedded into the interaction context and presents information relevant to that particular context. Also, we propose two general architectures materializing this model respectfully based on real-time analytical applications and a scalable analytic back-end. Using these architectures, we provided analytics services to the SM platform users. We conducted an evaluation with the users demonstrating that embedded contextual analytics was useful to support their awareness and reflection. Discovery. Due to a large number of content items stored in SM platforms, it can be challenging for the users to find relevant knowledge. Addressing this challenge, we propose an interactive recommender system based on user interests enabling discovery of relevant content and people. We have implemented the proposed recommender in an SM platform and conducted two evaluations with platform users. The evaluations demonstrated the ability of the approach to identify relevant user interests and to recommend relevant content. Delivery. At the moment of writing in 2016, near half of the world's population still does not have reliable Internet access. Often, the places where humanitarian action is needed have limited Internet connection. We propose a novel knowledge delivery model that relies on a peer-to-peer middleware and uses low-cost computers for local knowledge replication. We have developed a system implementing the model and evaluated it during eight deployments in MSF missions. The evaluation demonstrated its knowledge delivery abilities and its usefulness for the field staff.

1. ENHANCING SOCIAL MEDIA PLATFORMS FOR
EDUCATIONAL AND HUMANITARIAN
KNOWLEDGE SHARING:
ANALYTICS, PRIVACY, DISCOVERY, AND
DELIVERY ASPECTS
PhD thesis: https://goo.gl/AK0DHb
ANDRII VOZNIUK
REACT-EPFL
PhD Private DefenseLausanne, December 2016
1

2. KNOWLEDGE SHARING
• Social media platforms are used for global
knowledge sharing
• 25% of time online is spent on Facebook
• New platforms are developed and adapted to needs
of specific domains
2

3. AS A COLLABORATIVE
KNOWLEDGE SHARING
ENVIRONMENT
graasp.net
GRAASP
3
A SOCIAL MEDIA PLATFORM
AS A TESTBED FOR
PROPOSALS
VALIDATION

4. Space s
4
SPACE
ABSTRACTS
INTERACTION CONTEXT
Items I(s)
Members M(s)
SETTINGS
PRESENTATION P(s)
Activities A(s)

5. GRAASP
5
M MEMBERS
A ACTIVITIES
S SPACES
I ITEMS

6. KNOWLEDGE SHARING
6
EDUCATIONAL
Orientation
Conceptuali-
zation
Discussion
Investigation
Conclusion
Inquiry Learning Spaces (ILSs)
to support Inquiry Learning
HUMANITARIAN
FR CH BE NL ES
MSF HQ: Geneva (CH)
19 national offices
70 missions
international office
operational offices
national &
regional offices
organisationaloperational
325 projects
field
operations
Space hierarchies to represent
large decentralized organization

7. KNOWLEDGE SHARING
TEACHER -TEACHERS
TEACHER - STUDENTS
EDUCATIONAL
7
HUMANITARIAN
BETWEEN MISSION MEMBERS
HQ - FIELD

8. METHODOLOGY
HUMANITARIANEDUCATIONAL
8
DESIGN AND IMPLEMENTATION
User-Centered
Participatory
Iterative
Interdisciplinary
Exploratory
VALIDATION AND EVALUATION
With users
Authentic settings
Surveys
Case studies
Tracking

9. 9
ANALYTICS
•Providing user with analytics for awareness and reflection
analytics
PRIVACY
•Allowing easy and intuitive user privacy management
DISCOVERY
•Discovering knowledge relevant to user interests
DELIVERY
•Allowing users to access knowledge in underconnected
settings
TARGETED ASPECTS
OF KNOWLEDGE SHARING
Knowledge Sharing
System
Knowledge
Organisation
Knowledge
Acquisition
Knowledge
Access
Discovery
Delivery
Analytics Privacy

10. ANALYTICS ASPECT
EMBEDDED CONTEXTUAL ANALYTICS FOR
AWARENESS AND REFLECTION
10
Knowledge Sharing
System
Knowledge
Organisation
Knowledge
Acquisition
Knowledge
Access
Analytics Privacy
Discovery
Delivery

11. CHALLENGE
• Awareness is required to orchestrate and manage
• Reflection needed to understand and improve
11
RQ1. How to provide user-oriented analytics in
knowledge sharing systems to support awareness
and reflection?

12. RELATED WORK
• Not delivered for awareness: not integrated into
interaction context and not real-time
• Not designed for reflection: all information presented
altogether, not scoped per context
12
A. Vozniuk, S. Govaerts and D. Gillet. Towards portable learning analytics dashboards. ICALT,
Beijing, 2013.
B. A. Schwendimann, M. J. Rodriguez Triana, A. Vozniuk, L. P. Prieto and M. Shirvani Boroujeni
et al. Perceiving learning at a glance: A systematic literature review of learning dashboard
research. IEEE TLT, 2016.

13. External Stakeholders
PROPOSAL
EMBEDDED CONTEXTUAL
ANALYTICS MODEL
13
A. Vozniuk et al. Contextual Learning Analytics Apps to Create Awareness in Blended Inquiry
Learning. ITHET, Lisbon, 2015.
Space s
Items I(s)
Members M(s)
Activities A(s)
We call analytics embedded into the space s if and only if it is part of the space
presentation:
Embedded
We call analytics contextual in the space s if and only if
StandaloneAnalytics

14. ANALYTICS ASPECT
EMBEDDED CONTEXTUAL ANALYTICS
WITH WEB APPS
14

15. PROPOSAL
EMBEDDED CONTEXTUAL ANALYTICS
WITH WEB APPS
interacts
with the apps
User
Standardized APIs
Knowledge Sharing
Platform
retrieves activities
in the context
Front-end
Back-end
Analytics
Infrastructure Data
Storage
Context
Analytical App
stores activities
analyzes
and visualizes
activities
• Enables real-time awareness
• Allows customization and personalization
15
A. Vozniuk et al. Contextual Learning Analytics Apps to Create Awareness in Blended Inquiry
Learning. ITHET, Lisbon, 2015.

16. APPLICATION
EMBEDDED CONTEXTUAL ANALYTICS
WITH WEB APPS
16
In blended learning sessions
it is challenging to follow the
progress
We use the apps to improve
teacher’s awareness in
blended learning

17. APPLICATION
APPS TO SUPPORT AWARENESS
17
Conducted a survey with 32 teachers. Identified awareness needs.
Active Students App Time Spent App
Submitted Reports App

18. APPLICATION
PERSONALIZATION WITH DASHBOARDS
18
Multiple apps can be combined to construct a personalized dashboard
A. Vozniuk et. al. Towards portable learning analytics dashboards. ICALT, Beijing, 2013.

19. EVALUATION
METHODOLOGY
Interview with teachers before the
class to assess apps relevance
Attending the class using the apps,
two case studies
Interview with teachers after the class
to get feedback
Interview
Before
Case Studies
Interview
After
19
Case Study 1
90 minutes
1 teacher T1
11 students
Case Study 2
90 minutes
1 teacher T2
17 students

20. EVALUATION
OUTCOMES
• Teachers agreed on the usefulness of the Apps for
awareness
• Teachers stated that tools helped them to monitor
the progress of the students
• Teachers embedded analytics into their context by
projecting on the wall
• Students as well expressed awareness needs,
mainly to have progress reference
20

21. LIMITATIONS
FOR REFLECTION
• REFLECTION IS DONE OVER LONGER PERIODS
• Scalability is required
• Web apps do not scale
• REFLECTION REQUIRES EXPLORATION
• Interactivity is necessary for reflection
• Inter-apps communication is challenging
21

22. ANALYTICS ASPECT
EMBEDDED CONTEXTUAL ANALYTICS
WITH ANALYTICS ENGINE AND
INTERACTIVE DASHBOARDS
22

23. PROPOSAL
EMBEDDED CONTEXTUAL ANALYTICS
WITH ANALYTICS ENGINE & INTERACTIVE
DASHBOARDS
User
Platform APIs
Knowledge Sharing
Platform
Front-end
Back-end
Analytics
Infrastructure
Context
Interactive dashboard
Analytics
Engine
stores and analyzes
activities
Permissions check
and proxy
retrieves results
of analysis
visualizes
analysis
results
interacts
with the dashboard
• Enables
interactive
reflection
• Allows scalable
analysis
23
A. Vozniuk et al. Workshop. Building Interactive Learning Dashboards with
Logstash, Elasticsearch and Kibana (ELK). LASI, Bilbao, 2016.

24. IMPLEMENTATION IN GRAASP
24
Which users were the most active last week?
When did the users download the documents yesterday?
Which users did comment the most last month?
How active were the users in the space last month?
Which items were viewed the most last week?
Analytics
Which actions happened the most last year?
How the views are distributed between documents?
Briefing
paper
C
had
Safe
W
ate
r.pdfm
easle
s
in
te
rn
albriefingPrise
en
charg
e
de
m
enin
gite
FR
.p
df
D
ocum
ent5.p
df
D
ocum
ent6.d
oc
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Briefing
paper
C
had
Safe
W
ate
r.pdf
Supporting knowledge managers at MSF

25. ANALYTICS AND
PRIVACY
• Detailed analytics triggered privacy concerns with
knowledge workers
• Some users expressed concerns about the analytics
resolution and interactiveness
• Highlighted conflict between analytics and privacy
25

26. PRIVACY ASPECT
AN AGENT-BASED CONTEXTUAL PRIVACY
MANAGEMENT
26
Knowledge Sharing
System
Knowledge
Organisation
Knowledge
Acquisition
Knowledge
Access
Analytics Privacy
Discovery
Delivery

27. CHALLENGE
• Users expressed concerned about being tracked
• Better the analytics, stronger the concerns
• Particular contexts have legal regulations (medical,
primary school)
• Users want to be in control
27
RQ2. What privacy management interfaces and
mechanisms are suitable for knowledge analytics and
learning analytics?

28. RELATED WORK
• PRIVACY SETTINGS
• Hidden, complicated, changing
• Black patterns
• GROUP ANONYMIZATION
• Limits the possibilities of analytics
• Not suitable for educational settings
• AWARENESS CUES
• Make user know
28

29. PROPOSAL
AGENT-BASED PRIVACY
29
A. Vozniuk et al. AngeLA: Putting the Teacher in Control of Student Privacy in the
Online Classroom. ITHET, York, 2014.
Space s
Items I(s)
Members M(s)
Activities A(s)
Analytics Infrastructure
Data Storage
Data Analysis
Agent

30. IMPLEMENTATION
IN GRAASP
30

31. • Survey-based evaluation with workshop participants, f2f interviews
• N=16 respondents (7 male, 9 female)
• Defined and adjusted tracking policy using the Tracking Agent
• Graasp usage data
EVALUATION
31
SETTING
OUTCOMES
• SUS score 74, GOOD usability rating
• Usage: 1234 ILSs, 91% have Agent, 9% do note have agent
• 168 ILSs (1%) have Agent removed manually by the user
• Agent-based management adopted by Go-Lab and SiWay EU projects

32. DISCOVERY ASPECT
KNOWLEDGE DISCOVERY WITH INTERACTIVE
INTERESTS-BASED RECOMMENDER
32
Knowledge Sharing
System
Knowledge
Organisation
Knowledge
Acquisition
Knowledge
Access
Analytics Privacy
Discovery
Delivery

33. CHALLENGE
• Information overload in social media platforms
• Hard to find relevant content
• Not aware of its existence
• Do not know particular keywords
• New content appearing
• Both in educational and humanitarian domains
• NGOs have high turnover, previous knowledge can be lost
33
RQ3. How to enable discovery of knowledge relevant
to user interests?

34. PROPOSAL
RECOMMENDATION MODEL
Provide
Recommendations
Record
User-
Content
Interactions
Extract
Concepts
from the
Content Build
User
Interests
Profile
Interpretable
Interactive
34
A. Vozniuk et al. Combining Content Analytics and Activity Tracking to Identify
User Interests and Enable Knowledge Discovery. PALE, UMAP, Halifax, 2016.

35. Pdf Report
Powerpoint
Presentation
Image with
Text
Youtube
Video
Σw*UA
*DC
accessed
rated
commented
downloaded
Education
Educational psychology
Knowledge
Learning
Knowledge Management
Human-Computer Interaction
Interdisciplinarity
Academia
Systems thinking
Scientific method
Educational technology
Virtual learning environment
User
Identified Concepts (DC)
Identified User Concepts
(UC)
Tracked Activities (UA)
Education
Educational psychology
Knowledge
Learning
Knowledge Management
Systems thinking
Scientific method
Educational technology
Virtual learning environment
Learning
Knowledge Management
Human-Computer Interaction
Interdisciplinarity
Education
Educational psychology
Academia
35
PROPOSAL
INTERESTS PROFILE

36. RANKING
RECOMMENDATIONS
36
STEP 2.
Use vector cosine similarity for scoring and ranking
STEP 1.
Compute TF-IDF for each term in the vectors
STEP 0.
Represent each content item concepts using the document
vector model

37. IMPLEMENTATION
USER INTERFACE
37

38. EVALUATION
PRELIMINARY
• Six pre-service teachers
• Workshop on inquiry-based learning
• Newly registered users, no interaction data
• Interacted for 2 hours
• Survey with three parts: disposition, usability (SUS), quality
38
SETTING
OUTCOMES
• SUS score 63, OK usability
• Positive regarding interests-based recommendation idea, useful to
get recommendations
• Positive about seeing their identified interests (interpretability)
• Want to be able to edit the interests (interactivity)
• Interests and recommended items relevant in general
• Misidentified interests for two users due to inaccurate concepts

39. EVALUATION
WITH GRAASP USERS
39
SETTING
OUTCOMES
• At least five interests identified in profile
• Sent emails with the same survey of three parts
• 40 Graasp users (22 female, 16 male, 2 - not provided)
ADDRESSING CONCEPT IDENTIFICATION ISSUES
• Consider only concepts with relevance more than 0.8
• Consider a concept if it occurred at least in 3 items
• SUS score 70, OK usability approaching GOOD
• Positive regarding interests-based recommendation idea, useful to
get recommendations
• Positive about seeing their identified interests (interpretability).
• Want to be able to edit the interests (interactivity)
• Interests and recommended items were relevant (3.7 vs 3.17 and
3.52 vs 3.67)
• Would benefit from dataset-driven evaluation, challenging to find

40. DELIVERY ASPECT
MOBILE KNOWLEDGE DELIVERY INTO
UNDERCONNECTED ENVIRONMENTS
40
Knowledge Sharing
System
Knowledge
Organisation
Knowledge
Acquisition
Knowledge
Access
Analytics Privacy
Discovery
Delivery

41. CHALLENGE
LIMITED INTERNET ACCESS
• Half of world’s population without Internet
• Hight cost, low performance
• Unreliable or damaged infrastructure, disasters
41
RQ4. How to facilitate knowledge delivery into
settings where Internet connection is limited or
absent?
CONTENT INCREASES IN SIZE
• Video is the dominant knowledge medium
• Growing multimedia usage in education (MOOCs) and
in humanitarian (MSF org training)
• In emergency access time is critical

42. SITUATION MSF
• Low bandwidth
• High latency
• Often unreliable, outages are common
• High cost in remote areas
42

43. RELATED WORK
Global Cloud
or CDN
ISP
Cache or Proxy
Community
Village
Cache or Proxy
Local
Cache or Proxy
User
Device
Cache
ISP Network Village Network Local Network User Device
Possibility of mechanical delivery (first group solutions)
Possibility of caching and asynchronous delivery
(second group solutions)
Possibility of caching, async delivery,
minimizing the number of request,
prioritizing types of content
(third group solutions)
43
MECHANICAL DELIVERY
• resilient, on-the-go
• not shared downloads
USER DEVICE OPTIMIZATIONS
• requires custom software
INTERMEDIATE SERVERS
• shared downloads
• not portable

44. PROPOSED MODEL
Local WiFi
Connection
Internet Connection
LAN / WiFi / 3G
Online System
Local System A
Browser
navigating
Online System
Browser
navigating
Local System 1
Personal System 1
p2p
p2p
Local System B
p2p
p2p
Personal System 2
p2p
http
http
p2p
p2p
smartphone laptop smartphone tablet
• Peer-to-peer middleware
• Local system with low-cost-computers
• Personal system with mobile devices
44
A. Vozniuk et al. Graspeo: a Social Media Platform for Knowledge Management in NGOs. ICTD, Singapore,
2015.
A. Vozniuk et al. GraaspBox: Enabling Mobile Knowledge Delivery into Underconnected Environments. To be
submitted, 2017.

45. 45
IMPLEMENTATION
GRAASPBOX

46. 46
IMPLEMENTATION
GRAASPBOX
Local System
via Browser
Personal
System
Online System

47. DEPLOYMENT
IN THE FIELD
EIGHT DEPLOYMENTS
47
MAPUTO
MOZAMBIQUE
FIELD
N’DJAMENA
CHAD
BASE
ON THE ROAD

48. EVALUATION
48
• Eight field trips by MSF partner
• Demonstrating GraaspBox and letting to use
• Survey of 3 parts: Internet situation, delivery needs, SUS
• Used by 45 people, 14 respondents (11 male, 3 female)
SETTING
OUTCOMES
• Lab tests: download 90 Mbps, WiFi 20Mbps, 32 connections, 45
hours on battery
• Internet access is slow and unreliable
• Strong need to have faster access to MSF content
• Strong agreement with GraaspBox making content access faster
• Need to have content on personal device, GraaspBox, and online
• SUS score 78, GOOD close to EXCELLENT usability
• General positive feedback by MSF field staff

49. SUMMARY
AND FUTURE RESEARCH
49
Knowledge Sharing
System
Knowledge
Organisation
Knowledge
Acquisition
Knowledge
Access
Discovery
Delivery
Analytics Privacy

50. SUMMARY
ANALYTICS
• Embedded contextual analytics model to support
awareness and reflection
• Two general architectures materializing the model
• Integrated into Graasp, evaluated with users and adopted
PRIVACY
• Agent-based privacy management model
• Integrated into Graasp and evaluated with users
• Adopted by two EU projects
50

51. SUMMARY
51
DISCOVERY
• Recommender system for content and people based on
user interests
• Interactive and interpretable
• Integrated into Graasp and conducted two evaluations
with users
DELIVERY
• Knowledge delivery model with P2P middleware and low-
cost computers
• Integrated into Graasp with GraaspBox
• evaluated during eight deployments in MSF missions

52. FUTURE RESEARCH
DIRECTIONS
PRIVACY & ANALYTICS
• How to deal with incomplete analytics picture and make
the user aware of it?
PRIVACY & DISCOVERY
• What if user interacts in contexts where tracking is
disabled?
PRIVACY & DELIVERY
• How to deliver offline the privacy policy that is set online,
and should it be the same?
DELIVERY & DISCOVERY
• Predictively delivering content matching user interests.
52

53. QUESTIONS?
FEEDBACK?
andrii@vozniuk.com
53

54. ANDRII VOZNIUK
https://about.me/vozniuk
RESEARCHER
CS PhD Candidate
SOFTWARE ENGINEER
Web, Mobile, Data
ENTREPRENEUR
Knowledge Sharing Systems