Ethics and Privacy in the Application of Learning Analytics (#EP4LA)
1. Response
to
talks
at
Big
Data
and
Privacy
in
Human
Subject
Research
(1st
day)
Hendrik
Drachsler,
@hdrachsler
Welten
Ins4tute
Research
Centre,
Open
University
of
the
Netherlands
Presenta4on
given
at:
NSF
expert
mee4ng
on
‘Big
Data
and
Privacy
in
Human
Subjects
Research’
(#BDEDU)
11
November
2014
2. 3
WhoAmI
• Hendrik
Drachsler,
Open
University
of
the
Netherlands
• Research
topics:
Personaliza4on,
Recommender
Systems,
Learning
Analy4cs,
Mobile
devices
• Applica4on
domains:
Science
2.0
Health
2.0
5. They
brought
together
some
Super
Hero’s
5
Who
of
you
considers
him-‐
herself
to
be
a
Super
Hero?
• You
are
passionate
about
what
you
are
doing.
• You
shape
the
future
of
society.
• You
touch
ethical
ques4ons
with
your
super
power.
• You
want
to
follow
societal
norms
and
advance
those.
With
Big
Power
comes
great
responsibili5es.
6. Big
Power
-‐>
Big
Data
=
Repurposing
data
6
Jawbone
data
repurposed
to
measure
earthquake
strength
7. 7
Big
Data
is
the
new
truth
(the
ulHmate
truth?)
8. 8
Big
Data
is
the
new
truth
(the
ulHmate
truth?)
Inaccurate
Google
Flue
trend
measures
compared
to
CDC
9. Big
Data
has
the
potenHal
to
change
EducaHon
9
• First
4me
monitoring
learning
while
it
happens
• Personalize
Educa4on
• Iden4fy
students
at
Risk
• Learning
Measures
on
demand
• More
…
10. Some
QuesHons,
Super
Hero’s
10
Some
Demographics:
Who
of
you
are
data
scien4sts,
legal
or
educa4onal
experts?
Who
of
you
read
TOC
of
your
online
services?
Who
of
you
cares
about
his/her
privacy?
Who
sees
Privacy
and
Legal
regula4ons
as
a
burden
we
need
to
overcome?
11. Learning
AnalyHcs
Research
Issues
11
Learning
Analy4cs
research
always
raises
the
P-‐Word
in
EU
(University
of
Amsterdam,
2014)
This
stops
innova4on
and
advancing
research
(dataTEL
2010)
12. 12
• Privacy
changes
overHme
• Privacy
is
bind
to
context
• Privacy
is
bind
to
culture
Slide
supported
byTore
Hoel,
@Tore
13. What
if
I
would
know
…
• How
many
days
you
have
NOT
been
at
school
without
any
excuse.
• All
read
and
wrihen
pages,
and
what
your
annota4ons
have
been.
• The
people
you
hangout
with
in
your
youth.
• If
you
cheated
in
a
test
and
how
many
ahempts
you
needed
for
your
math
class.
• What
if
I
use
all
those
informaHon
and
predict
your
chances
to
be
good
or
bad
in
a
certain
job
aSer
school?
• How
representaHve
and
reliable
is
this
data
I’m
capturing
to
predict
those
chances?
• And
what
if
all
this
informaHon
will
be
last
forever!
13
14. Approaches
to
prevent
another
inBloom
…
• Transparency
(Purpose
of
analysis,
Raw
data
access,
opt-‐out)
• Data
Security
• Contextual
Integrity
(Smart
Informed
Consents)
• Anonymisa4on
&
Data
degrada4on
14
15. Ethics
&
Privacy
Issue
in
the
ApplicaHon
of
Learning
AnalyHcs
(#EP4LA)
Hendrik
Drachsler,
@hdrachsler
Welten
InsHtute
Research
Centre,
Open
University
of
the
Netherlands
Presenta4on
given
at:
NSF
expert
mee4ng
on
‘Big
Data
and
Privacy
in
Human
Subjects
Research’
(#BDEDU)
11
November
2014
16. Building
bridges
between
research,
policy
and
prac4ce
to
realise
the
poten4al
of
learning
analy4cs
in
EU
FP7
LACE
–
Hendrik
Drachsler,
@Hdrachsler,
28
October
2014
16
17. Who
we
are
FP7
LACE
–
Hendrik
Drachsler,
@Hdrachsler,
28
October
2014
17
LACE
Network
LACE
ConsorHum
18. Data
Geology
18
PAST,
single,
centered
IT
solu4ons
with
single
purpose
(loosely
couple
data)
PRESENT,
mul4ple
ubiquitous
IT
systems
mul4ple
func4onali4es
(highly
connected
but
unstructured
data)
FP7
LACE
–
Hendrik
Drachsler,
@Hdrachsler,
28
October
2014
FUTURE,
learner
ac4vity
tracking
of
ubiquitous
systems
(structured
learner
data)
19. Data
Geology
FP7
LACE
–
Hendrik
Drachsler,
@Hdrachsler,
28
October
2014
19
• Are
our
instruments
measuring
what
we
expect
them
to
measure?
• Can
we
isolate
the
noise
in
the
data?
• Are
the
measures
accurate?
Picture
from:
hhp://wsnblog.com/2012/05/28/how-‐sensors-‐can-‐lead-‐us-‐to-‐beher-‐self-‐
knowledge/human-‐body-‐sensors/
21. Privacy
as
Showstopper
–
The
inBloom
case
• $100
million
investment
• Aim:
Personalized
learning
in
public
schools,
through
data
&
technology
standards
• 9
US
states
par4cipated
• In
2013
the
database
held
informa4on
on
millions
of
children
21
23. Privacy
• What is privacy?
– Right to be let alone (Warren and Brandeis)
– Informational self-determination (Westin)
– Degree of access (Gavison)
– … Right to be forgotten …
• Three dimensions (Roessler)
– Informational privacy
– Decisional privacy
– Local privacy
• What it is not
– Anonymity, secrecy, data protection
24. What
are
the
dangers
of
learning
analyHcs?
– Missing
legal
obligaHons:
• Data
protec4on
• IRB
• Educa4on
laws
– InflicHng
harm:
• Unfair
discrimina4on
• Unjus4fied
discrimina4on
(through
errors)
• Subjec4ve
privacy
harm
(panop4c
effect)
• Unintended
pressure
to
perform
/
wrong
incen4ves?
• De-‐iden4fica4on
– ViolaHng
human
dignity
– Unintended
changes
of
educaHon
norms?
24
25. ModernizaHon
of
EU
UniversiHes
report
RecommendaHon
14
Member
States
should
ensure
that
legal
frameworks
allow
higher
educa4on
ins4tu4ons
to
collect
and
analyse
learning
data.
The
full
and
informed
consent
of
students
must
be
a
requirement
and
the
data
should
only
be
used
for
educa4onal
purposes.
RecommendaHon
15
Online
plaoorms
should
inform
users
about
their
privacy
and
data
protec4on
policy
in
a
clear
and
understandable
way.
Individuals
should
always
have
the
choice
to
anonymise
their
data.
hgp://ec.europa.eu/educaHon/
library/reports/modernisaHon-‐
universiHes_en.pdf
26. #EP4LA
on
the
European
Agenda
• Round
table
meeHng
‘Ethiek
en
Learning
AnalyHcs’
(Jan
2014)
hhps://www.surfspace.nl/media/bijlagen/ar4kel-‐1499-‐
b315e61001041bf52a6b1c5d80053cea.pdf
• Learning
AnalyHcs
Summer
InsHtute
(July
2014)
hhp://lasiutrecht.wordpress.com/
• Call
for
a
‘Code
of
Ethics
for
LA’
in
NL
(August
2014)
hhps://www.surfspace.nl/ar4kel/1311-‐towards-‐a-‐uniform-‐code-‐of-‐ethics-‐and-‐
prac4ces-‐for-‐learning-‐analy4cs/
• Call
for
a
‘Code
of
Ethics
for
LA’
in
the
UK
(September
2014)
hhp://analy4cs.jiscinvolve.org/wp/2014/09/18/code-‐of-‐prac4ce-‐essen4al-‐
for-‐learning-‐analy4cs/
26
30. Example
Issues
from
Stakeholders
• Who
is
in
charge
(who
is
the
owners)
of
the
data
created
by
persons?
• What
is
the
impact
of
privacy
concerns
for
the
management?
How
to
deal
with
these
concerns?
• Should
students
be
allowed
to
opt-‐out
of
having
their
personal
digital
footprints
harvested
and
analysed?
• How
to
prevent
reuse
of
collected
data
for
non-‐educa4onal
needs.
(e.g.
finance,
insurance,
research),
or
is
it
no
problem?
Full
list:
hgp://bit.ly/raw_ep4la
31. We
are
pracHcal
people
–
our
approach
31
• Invite
5
legal
experts,
15
members
of
the
SURF
SIG
LA
• Task
groups
to
answer
issues
of
the
stakeholders
• Open
Working
doc
for
all
#EP4LA
events
32. 32
Four
examples
how
we
addressed
the
issues
submiged
by
the
stakeholders
33. 1.
Boundaries
of
Learning
AnalyHcs
data
Where
is
the
boundary
on
data
use
for
learning
analy3cs
(courses,
grades,
LMS,
GoogleDrive,
library
system,
residence
halls,
dining
halls,
…)?
– Contextual
Integrity:
context
and
norms
of
learning
environment
– It
depends
on
• Awareness
of
students
about
processes
• Possible
consequences
for
students
• Safeguards
that
are
in
place
33
34. 2.
Outsourcing
What
are
the
concerns
when
outsourcing
the
collec3on
and
analysis
of
data?
Who
owns
the
data?
– Concerns:
• Undue
third
country
data
transfers
• Less
control
about
processing
• Less
transparency
for
the
data
subject
– Ownership:
• No
complete
ownership
for
any
party
• Relevant:
data
protec4on
and
intellectual
property
rights
• See
discussion
concerning
`data
portability’
in
DP
regula4on
(NDA
agreement
required)
34
35. 3.
Undesirable
data
collecHon
Are
there
any
circumstances
when
collecHng
data
about
students
is
unacceptable/undesirable?
– Yes,
there
are:
• Data
which
is
not
of
any
purpose
• Data
outside
of
the
learning
context
• Data
of
which
the
student
is
not
aware
• Data
which
poses
a
risk
to
the
student
• Data
which
is
not
well
protected
35
36. 4.
Data
access
by
students
What
data
should
students
be
able
to
view,
i.e.
what
and
how
much
informaHon
should
be
provided
to
the
student?
– Data
Protec4on
Direc4ve
(ar4cle
12):
• Everything
concerning
them
(at
least
upon
request)
– Human
subjects
research:
• Everything
concerning
study
(at
least
arer
experiment)
• Avoidance
of
decep4on
– But
• Possible
conflict
of
full
data
access
with
goals
of
LA?
• How
to
provide
meaningful
access
while
excluding
other
students
data?
36
37. We
idenHfied
9
main
themes
that
are
relevant
for
LA
in
Europe
37
38. 9
Themes
around
privacy
(1/3)
1.
LegiHmate
grounds
-‐
Why
are
you
allowed
to
have
the
data?
2.
Purpose
of
the
data
-‐
Repurposing
is
an
issue
vs.
MIT
Social
Machine
lab
3.
Inventory
of
data
-‐
What
data
do
you
have?
-‐
What
can
you
do
with
that
data
already?
39. 9
Themes
around
privacy
(2/3)
4.
Data
quality
-‐
How
good
is
the
data?
(eg.
Bb
log
file
is
weak
predictor)
-‐
When
do
you
I
delete
data
and
what
data?
5.
Transparency
-‐
Informing
students
(Purpose,
Approach)
-‐
Checklist
what
to
communicate
for
researchers
6.
The
rights
of
the
data
subject
to
access
their
data
from
the
data
client
-‐
For
teachers
who
are
employees
other
rights
apply
40. 9
Themes
around
privacy
(3/3)
7.
Outsource
processing
to
external
parHes
-‐
Prevent
external
par4es
to
not
do
addi4onal
analysis
(NDA
agreement)
8.
Transport
data,
legal
locaHon
-‐
e.g.
Safe
Harbour
agreement
9.
Data
Security
-‐>
Shuangbao
Wang
41. Value Sensitive Design
(Batya Friedman)
Goal:
address
human
values
in
a
technical
design
Source: presentation by Jeroen van den Hoven
42. www.laceproject.eu
@laceproject
“Ethics
&
Privacy
Issues
in
the
Applica4on
of
Learning
Analy4cs”
by
Hendrik
Drachsler,
Open
University
of
the
Netherlands
was
presented
at
NSF
Mee4ng
–
Big
Data
in
Educa4on,
Washington,
USA,
on
09-‐11.10.2014.
Hendrik.drachsler@ou.nl,
@hdrachsler
This
work
was
undertaken
as
part
of
the
LACE
Project,
supported
by
the
European
Commission
Seventh
Framework
Programme,
grant
619424.
These
slides
are
provided
under
the
Crea4ve
Commons
Ahribu4on
Licence:
hhp://
crea4vecommons.org/licenses/by/4.0/.
Some
images
used
may
have
different
licence
terms.
42
43. 43
43
You
are
free
to:
copy,
share,
adapt,
or
re-‐mix;
photograph,
film,
or
broadcast;
blog,
live-‐blog,
or
post
video
of
this
presenta4on
provided
that:
You
ahribute
the
work
to
its
author
and
respect
the
rights
and
licences
associated
with
its
components.