<s>
A	O
filter	B-Algorithm
bubble	I-Algorithm
or	O
ideological	O
frame	O
is	O
a	O
state	O
of	O
intellectual	O
isolation	O
that	O
can	O
result	O
from	O
personalized	B-Application
searches	I-Application
,	O
where	O
a	O
website	O
algorithm	O
selectively	O
curates	O
what	O
information	O
a	O
user	O
would	O
like	O
to	O
see	O
based	O
on	O
information	O
about	O
the	O
user	O
,	O
such	O
as	O
location	O
,	O
past	O
click-behavior	O
,	O
and	O
search	O
history	O
.	O
</s>
<s>
Consequently	O
,	O
users	O
become	O
separated	O
from	O
information	O
that	O
disagrees	O
with	O
their	O
viewpoints	O
,	O
effectively	O
isolating	O
them	O
in	O
their	O
own	O
cultural	O
or	O
ideological	B-Algorithm
bubbles	I-Algorithm
,	O
resulting	O
in	O
a	O
limited	O
and	O
customized	O
view	O
of	O
the	O
world	O
.	O
</s>
<s>
Prime	O
examples	O
include	O
Google	B-Application
Personalized	I-Application
Search	I-Application
results	O
and	O
Facebook	B-Application
's	O
personalized	B-Application
news-stream	I-Application
.	O
</s>
<s>
The	O
term	O
filter	B-Algorithm
bubble	I-Algorithm
was	O
coined	O
by	O
internet	O
activist	O
Eli	O
Pariser	O
circa	O
2010	O
.	O
</s>
<s>
In	O
Pariser	O
's	O
influential	O
book	O
under	O
the	O
same	O
name	O
,	O
The	B-Algorithm
Filter	I-Algorithm
Bubble	I-Algorithm
(	O
2011	O
)	O
,	O
it	O
was	O
predicted	O
that	O
individualized	O
personalization	O
by	O
algorithmic	O
filtering	O
would	O
lead	O
to	O
intellectual	O
isolation	O
and	O
social	O
fragmentation	O
.	O
</s>
<s>
The	O
bubble	B-Algorithm
effect	O
may	O
have	O
negative	O
implications	O
for	O
civic	O
discourse	O
,	O
according	O
to	O
Pariser	O
,	O
but	O
contrasting	O
views	O
regard	O
the	O
effect	O
as	O
minimal	O
and	O
addressable	O
.	O
</s>
<s>
According	O
to	O
Pariser	O
,	O
users	O
get	O
less	O
exposure	O
to	O
conflicting	O
viewpoints	O
and	O
are	O
isolated	O
intellectually	O
in	O
their	O
informational	B-Algorithm
bubble	I-Algorithm
.	O
</s>
<s>
He	O
related	O
an	O
example	O
in	O
which	O
one	O
user	O
searched	O
Google	B-Application
for	O
"	O
BP	O
"	O
and	O
got	O
investment	O
news	O
about	O
British	O
Petroleum	O
,	O
while	O
another	O
searcher	O
got	O
information	O
about	O
the	O
Deepwater	O
Horizon	O
oil	O
spill	O
,	O
noting	O
that	O
the	O
two	O
search	O
results	O
pages	O
were	O
"	O
strikingly	O
different	O
"	O
despite	O
use	O
of	O
the	O
same	O
key	O
words	O
.	O
</s>
<s>
The	O
results	O
of	O
the	O
U.S.	O
presidential	O
election	O
in	O
2016	O
have	O
been	O
associated	O
with	O
the	O
influence	O
of	O
social	O
media	O
platforms	O
such	O
as	O
Twitter	B-Application
and	O
Facebook	B-Application
,	O
and	O
as	O
a	O
result	O
have	O
called	O
into	O
question	O
the	O
effects	O
of	O
the	O
"	O
filter	B-Algorithm
bubble	I-Algorithm
"	O
phenomenon	O
on	O
user	O
exposure	O
to	O
fake	O
news	O
and	O
echo	O
chambers	O
,	O
spurring	O
new	O
interest	O
in	O
the	O
term	O
,	O
with	O
many	O
concerned	O
that	O
the	O
phenomenon	O
may	O
harm	O
democracy	O
and	O
well-being	O
by	O
making	O
the	O
effects	O
of	O
misinformation	O
worse	O
.	O
</s>
<s>
Pariser	O
defined	O
his	O
concept	O
of	O
a	O
filter	B-Algorithm
bubble	I-Algorithm
in	O
more	O
formal	O
terms	O
as	O
"	O
that	O
personal	O
ecosystem	O
of	O
information	O
that	O
's	O
been	O
catered	O
by	O
these	O
algorithms.	O
"	O
</s>
<s>
Accessing	O
the	O
data	O
of	O
link	O
clicks	O
displayed	O
through	O
site	O
traffic	O
measurements	O
determines	O
that	O
filter	B-Algorithm
bubbles	I-Algorithm
can	O
be	O
collective	O
or	O
individual	O
.	O
</s>
<s>
As	O
of	O
2011	O
,	O
one	O
engineer	O
had	O
told	O
Pariser	O
that	O
Google	B-Application
looked	O
at	O
57	O
different	O
pieces	O
of	O
data	O
to	O
personally	O
tailor	O
a	O
user	O
's	O
search	O
results	O
,	O
including	O
non-cookie	O
data	O
such	O
as	O
the	O
type	O
of	O
computer	O
being	O
used	O
and	O
the	O
user	O
's	O
physical	O
location	O
.	O
</s>
<s>
Pariser	O
's	O
idea	O
of	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
was	O
popularized	O
after	O
the	O
TED	O
talk	O
in	O
May	O
2011	O
,	O
in	O
which	O
he	O
gave	O
examples	O
of	O
how	O
filter	B-Algorithm
bubbles	I-Algorithm
work	O
and	O
where	O
they	O
can	O
be	O
seen	O
.	O
</s>
<s>
In	O
a	O
test	O
seeking	O
to	O
demonstrate	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
effect	O
,	O
Pariser	O
asked	O
several	O
friends	O
to	O
search	O
for	O
the	O
word	O
"	O
Egypt	O
"	O
on	O
Google	B-Application
and	O
send	O
him	O
the	O
results	O
.	O
</s>
<s>
In	O
The	B-Algorithm
Filter	I-Algorithm
Bubble	I-Algorithm
,	O
Pariser	O
warns	O
that	O
a	O
potential	O
downside	O
to	O
filtered	O
searching	O
is	O
that	O
it	O
"	O
closes	O
us	O
off	O
to	O
new	O
ideas	O
,	O
subjects	O
,	O
and	O
important	O
information	O
,	O
"	O
and	O
"	O
creates	O
the	O
impression	O
that	O
our	O
narrow	O
self-interest	O
is	O
all	O
that	O
exists.	O
"	O
</s>
<s>
In	O
his	O
view	O
,	O
filter	B-Algorithm
bubbles	I-Algorithm
are	O
potentially	O
harmful	O
to	O
both	O
individuals	O
and	O
society	O
.	O
</s>
<s>
He	O
criticized	O
Google	B-Application
and	O
Facebook	B-Application
for	O
offering	O
users	O
"	O
too	O
much	O
candy	O
and	O
not	O
enough	O
carrots.	O
"	O
</s>
<s>
According	O
to	O
Pariser	O
,	O
the	O
detrimental	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
include	O
harm	O
to	O
the	O
general	O
society	O
in	O
the	O
sense	O
that	O
they	O
have	O
the	O
possibility	O
of	O
"	O
undermining	O
civic	O
discourse	O
"	O
and	O
making	O
people	O
more	O
vulnerable	O
to	O
"	O
propaganda	O
and	O
manipulation.	O
"	O
</s>
<s>
Many	O
people	O
are	O
unaware	O
that	O
filter	B-Algorithm
bubbles	I-Algorithm
even	O
exist	O
.	O
</s>
<s>
This	O
can	O
be	O
seen	O
in	O
an	O
article	O
in	O
The	O
Guardian	O
,	O
which	O
mentioned	O
the	O
fact	O
that	O
"	O
more	O
than	O
60%	O
of	O
Facebook	B-Application
users	I-Application
are	O
entirely	O
unaware	O
of	O
any	O
curation	O
on	O
Facebook	B-Application
at	O
all	O
,	O
believing	O
instead	O
that	O
every	O
single	O
story	O
from	O
their	O
friends	O
and	O
followed	O
pages	O
appeared	O
in	O
their	O
news	O
feed.	O
"	O
</s>
<s>
A	O
brief	O
explanation	O
for	O
how	O
Facebook	B-Application
decides	O
what	O
goes	O
on	O
a	O
user	O
's	O
news	B-Application
feed	I-Application
is	O
through	O
an	O
algorithm	O
that	O
takes	O
into	O
account	O
"	O
how	O
you	O
have	O
interacted	O
with	O
similar	O
posts	O
in	O
the	O
past.	O
"	O
</s>
<s>
A	O
filter	B-Algorithm
bubble	I-Algorithm
has	O
been	O
described	O
as	O
exacerbating	O
a	O
phenomenon	O
that	O
called	O
splinternet	B-Application
or	O
cyberbalkanization	B-Application
,	O
which	O
happens	O
when	O
the	O
internet	O
becomes	O
divided	O
into	O
sub-groups	O
of	O
like-minded	O
people	O
who	O
become	O
insulated	O
within	O
their	O
own	O
online	O
community	O
and	O
fail	O
to	O
get	O
exposure	O
to	O
different	O
views	O
.	O
</s>
<s>
This	O
concern	O
dates	O
back	O
to	O
the	O
early	O
days	O
of	O
the	O
publicly	O
accessible	O
internet	O
,	O
with	O
the	O
term	O
"	O
cyberbalkanization	B-Application
"	O
being	O
coined	O
in	O
1996	O
.	O
</s>
<s>
The	O
concept	O
of	O
a	O
filter	B-Algorithm
bubble	I-Algorithm
has	O
been	O
extended	O
into	O
other	O
areas	O
,	O
to	O
describe	O
societies	O
that	O
self-segregate	O
according	O
political	O
views	O
but	O
also	O
economic	O
,	O
social	O
,	O
and	O
cultural	O
situations	O
.	O
</s>
<s>
Barack	O
Obama	O
's	O
farewell	O
address	O
identified	O
a	O
similar	O
concept	O
to	O
filter	B-Algorithm
bubbles	I-Algorithm
as	O
a	O
"	O
threat	O
to	O
 [ Americans' ] 	O
democracy	O
,	O
"	O
i.e.	O
,	O
the	O
"	O
retreat	O
into	O
our	O
own	O
bubbles	B-Algorithm
,	O
especially	O
...	O
our	O
social	O
media	O
feeds	O
,	O
surrounded	O
by	O
people	O
who	O
look	O
like	O
us	O
and	O
share	O
the	O
same	O
political	O
outlook	O
and	O
never	O
challenge	O
our	O
assumptions	O
...	O
And	O
increasingly	O
,	O
we	O
become	O
so	O
secure	O
in	O
our	O
bubbles	B-Algorithm
that	O
we	O
start	O
accepting	O
only	O
information	O
,	O
whether	O
it	O
's	O
true	O
or	O
not	O
,	O
that	O
fits	O
our	O
opinions	O
,	O
instead	O
of	O
basing	O
our	O
opinions	O
on	O
the	O
evidence	O
that	O
is	O
out	O
there.	O
"	O
</s>
<s>
Both	O
"	O
echo	O
chambers	O
"	O
and	O
"	O
filter	B-Algorithm
bubbles	I-Algorithm
"	O
describe	O
situations	O
where	O
individuals	O
are	O
exposed	O
to	O
a	O
narrow	O
range	O
of	O
opinions	O
and	O
perspectives	O
that	O
reinforce	O
their	O
existing	O
beliefs	O
and	O
biases	O
,	O
but	O
there	O
are	O
some	O
subtle	O
differences	O
between	O
the	O
two	O
,	O
especially	O
in	O
practices	O
surrounding	O
social	O
media	O
.	O
</s>
<s>
On	O
the	O
other	O
hand	O
,	O
filter	B-Algorithm
bubbles	I-Algorithm
are	O
implicit	O
mechanisms	O
of	O
pre-selected	O
personalization	O
,	O
where	O
a	O
user	O
's	O
media	O
consumption	O
is	O
created	O
by	O
personalized	O
algorithms	O
;	O
the	O
content	O
a	O
user	O
sees	O
is	O
filtered	O
through	O
an	O
AI-driven	O
algorithm	O
that	O
reinforces	O
their	O
existing	O
beliefs	O
and	O
preferences	O
,	O
potentially	O
excluding	O
contrary	O
or	O
diverse	O
perspectives	O
.	O
</s>
<s>
Some	O
researchers	O
argue	O
,	O
however	O
,	O
that	O
because	O
users	O
still	O
play	O
an	O
active	O
role	O
in	O
selectively	O
curating	O
their	O
own	O
newsfeeds	O
and	O
information	O
sources	O
through	O
their	O
interactions	O
with	O
search	O
engines	O
and	O
social	O
media	O
networks	O
,	O
that	O
they	O
directly	O
assist	O
in	O
the	O
filtering	O
process	O
by	O
AI-driven	O
algorithms	O
,	O
thus	O
effectively	O
engaging	O
in	O
self-segregating	O
filter	B-Algorithm
bubbles	I-Algorithm
.	O
</s>
<s>
The	O
results	O
varied	O
only	O
in	O
minor	O
respects	O
from	O
person	O
to	O
person	O
,	O
and	O
any	O
differences	O
did	O
not	O
appear	O
to	O
be	O
ideology-related	O
,	O
leading	O
Weisberg	O
to	O
conclude	O
that	O
a	O
filter	B-Algorithm
bubble	I-Algorithm
was	O
not	O
in	O
effect	O
,	O
and	O
to	O
write	O
that	O
the	O
idea	O
that	O
most	O
internet	O
users	O
were	O
"	O
feeding	O
at	O
the	O
trough	O
of	O
a	O
Daily	B-Application
Me	I-Application
"	O
was	O
overblown	O
.	O
</s>
<s>
Weisberg	O
asked	O
Google	B-Application
to	O
comment	O
,	O
and	O
a	O
spokesperson	O
stated	O
that	O
algorithms	O
were	O
in	O
place	O
to	O
deliberately	O
"	O
limit	O
personalization	O
and	O
promote	O
variety.	O
"	O
</s>
<s>
Interviewing	O
programmers	O
at	O
Google	B-Application
,	O
off	O
the	O
record	O
,	O
journalist	O
Per	O
Grankvist	O
found	O
that	O
user	O
data	O
used	O
to	O
play	O
a	O
bigger	O
role	O
in	O
determining	O
search	O
results	O
but	O
that	O
Google	B-Application
,	O
through	O
testing	O
,	O
found	O
that	O
the	O
search	O
query	O
is	O
by	O
far	O
the	O
best	O
determinator	O
of	O
what	O
results	O
to	O
display	O
.	O
</s>
<s>
There	O
are	O
reports	O
that	O
Google	B-Application
and	O
other	O
sites	O
maintain	O
vast	O
"	O
dossiers	O
"	O
of	O
information	O
on	O
their	O
users	O
,	O
which	O
might	O
enable	O
them	O
to	O
personalize	O
individual	O
internet	O
experiences	O
further	O
if	O
they	O
choose	O
to	O
do	O
so	O
.	O
</s>
<s>
For	O
instance	O
,	O
the	O
technology	O
exists	O
for	O
Google	B-Application
to	O
keep	O
track	O
of	O
users	O
 '	O
histories	O
even	O
if	O
they	O
do	O
n't	O
have	O
a	O
personal	O
Google	B-Application
account	O
or	O
are	O
not	O
logged	O
into	O
one	O
.	O
</s>
<s>
One	O
report	O
stated	O
that	O
Google	B-Application
had	O
collected	O
"	O
10	O
years	O
 '	O
worth	O
"	O
of	O
information	O
amassed	O
from	O
varying	O
sources	O
,	O
such	O
as	O
Gmail	B-Application
,	O
Google	B-Application
Maps	I-Application
,	O
and	O
other	O
services	O
besides	O
its	O
search	O
engine	O
,	O
although	O
a	O
contrary	O
report	O
was	O
that	O
trying	O
to	O
personalize	O
the	O
internet	O
for	O
each	O
user	O
,	O
was	O
technically	O
challenging	O
for	O
an	O
internet	O
firm	O
to	O
achieve	O
despite	O
the	O
huge	O
amounts	O
of	O
available	O
data	O
.	O
</s>
<s>
Analyst	O
Doug	O
Gross	O
of	O
CNN	B-Architecture
suggested	O
that	O
filtered	O
searching	O
seemed	O
to	O
be	O
more	O
helpful	O
for	O
consumers	O
than	O
for	O
citizens	O
,	O
and	O
would	O
help	O
a	O
consumer	O
looking	O
for	O
"	O
pizza	O
"	O
find	O
local	O
delivery	O
options	O
based	O
on	O
a	O
personalized	B-Application
search	I-Application
and	O
appropriately	O
filter	O
out	O
distant	O
pizza	O
stores	O
.	O
</s>
<s>
A	O
scientific	O
study	O
from	O
Wharton	O
that	O
analyzed	O
personalized	B-Application
recommendations	I-Application
also	O
found	O
that	O
these	O
filters	O
can	O
create	O
commonality	O
,	O
not	O
fragmentation	O
,	O
in	O
online	O
music	O
taste	O
.	O
</s>
<s>
Harvard	O
law	O
professor	O
Jonathan	O
Zittrain	O
disputed	O
the	O
extent	O
to	O
which	O
personalization	O
filters	O
distort	O
Google	B-Application
search	O
results	O
,	O
saying	O
that	O
"	O
the	O
effects	O
of	O
search	O
personalization	O
have	O
been	O
light.	O
"	O
</s>
<s>
Further	O
,	O
Google	B-Application
provides	O
the	O
ability	O
for	O
users	O
to	O
shut	O
off	O
personalization	O
features	O
if	O
they	O
choose	O
by	O
deleting	O
Google	B-Application
's	I-Application
record	O
of	O
their	O
search	O
history	O
and	O
setting	O
Google	B-Application
not	O
to	O
remember	O
their	O
search	O
keywords	O
and	O
visited	O
links	O
in	O
the	O
future	O
.	O
</s>
<s>
A	O
study	O
from	O
Internet	O
Policy	O
Review	O
addressed	O
the	O
lack	O
of	O
a	O
clear	O
and	O
testable	O
definition	O
for	O
filter	B-Algorithm
bubbles	I-Algorithm
across	O
disciplines	O
;	O
this	O
often	O
results	O
in	O
researchers	O
defining	O
and	O
studying	O
filter	B-Algorithm
bubbles	I-Algorithm
in	O
different	O
ways	O
.	O
</s>
<s>
Subsequently	O
,	O
the	O
study	O
explained	O
a	O
lack	O
of	O
empirical	O
data	O
for	O
the	O
existence	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
across	O
disciplines	O
and	O
suggested	O
that	O
the	O
effects	O
attributed	O
to	O
them	O
may	O
stem	O
more	O
from	O
preexisting	O
ideological	O
biases	O
than	O
from	O
algorithms	O
.	O
</s>
<s>
Similar	O
views	O
can	O
be	O
found	O
in	O
other	O
academic	O
projects	O
,	O
which	O
also	O
address	O
concerns	O
with	O
the	O
definitions	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
the	O
relationships	O
between	O
ideological	O
and	O
technological	O
factors	O
associated	O
with	O
them	O
.	O
</s>
<s>
A	O
critical	O
review	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
suggested	O
that	O
"	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
thesis	O
often	O
posits	O
a	O
special	O
kind	O
of	O
political	O
human	O
who	O
has	O
opinions	O
that	O
are	O
strong	O
,	O
but	O
at	O
the	O
same	O
time	O
highly	O
malleable	O
"	O
and	O
that	O
it	O
is	O
a	O
"	O
paradox	O
that	O
people	O
have	O
an	O
active	O
agency	O
when	O
they	O
select	O
content	O
but	O
are	O
passive	O
receivers	O
once	O
they	O
are	O
exposed	O
to	O
the	O
algorithmically	O
curated	O
content	O
recommended	O
to	O
them.	O
"	O
</s>
<s>
They	O
then	O
identified	O
whether	O
news	O
stories	O
were	O
read	O
after	O
accessing	O
the	O
publisher	O
's	O
site	O
directly	O
,	O
via	O
the	B-Application
Google	I-Application
News	O
aggregation	O
service	O
,	O
web	O
searches	O
,	O
or	O
social	O
media	O
.	O
</s>
<s>
A	O
study	O
by	O
Princeton	O
University	O
and	O
New	O
York	O
University	O
researchers	O
aimed	O
to	O
study	O
the	O
impact	O
of	O
filter	B-Algorithm
bubble	I-Algorithm
and	O
algorithmic	O
filtering	O
on	O
social	O
media	O
polarization	O
.	O
</s>
<s>
They	O
used	O
a	O
mathematical	O
model	O
called	O
the	O
"	O
stochastic	O
block	O
model	O
"	O
to	O
test	O
their	O
hypothesis	O
on	O
the	O
environments	O
of	O
Reddit	O
and	O
Twitter	B-Application
.	O
</s>
<s>
The	O
researchers	O
gauged	O
changes	O
in	O
polarization	O
in	O
regularized	O
social	O
media	O
networks	O
and	O
non-regularized	O
networks	O
,	O
specifically	O
measuring	O
the	O
percent	O
changes	O
in	O
polarization	O
and	O
disagreement	O
on	O
Reddit	O
and	O
Twitter	B-Application
.	O
</s>
<s>
While	O
algorithms	O
do	O
limit	O
political	O
diversity	O
,	O
some	O
of	O
the	B-Algorithm
filter	I-Algorithm
bubbles	I-Algorithm
are	O
the	O
result	O
of	O
user	O
choice	O
.	O
</s>
<s>
A	O
study	O
by	O
data	O
scientists	O
at	O
Facebook	B-Application
found	O
that	O
users	O
have	O
one	O
friend	O
with	O
contrasting	O
views	O
for	O
every	O
four	O
Facebook	B-Application
friends	O
that	O
share	O
an	O
ideology	O
.	O
</s>
<s>
No	O
matter	O
what	O
Facebook	B-Application
's	O
algorithm	O
for	O
its	O
News	B-Application
Feed	I-Application
is	O
,	O
people	O
are	O
more	O
likely	O
to	O
befriend/follow	O
people	O
who	O
share	O
similar	O
beliefs	O
.	O
</s>
<s>
Although	O
algorithms	O
and	O
filter	B-Algorithm
bubbles	I-Algorithm
weaken	O
content	O
diversity	O
,	O
this	O
study	O
reveals	O
that	O
political	O
polarization	O
trends	O
are	O
primarily	O
driven	O
by	O
pre-existing	O
views	O
and	O
failure	O
to	O
recognize	O
outside	O
sources	O
.	O
</s>
<s>
Basing	O
their	O
study	O
on	O
the	O
notion	O
that	O
the	O
number	O
of	O
news	O
sources	O
that	O
users	O
consume	O
impacts	O
their	O
likelihood	O
to	O
be	O
caught	O
in	O
a	O
filter	B-Algorithm
bubble	I-Algorithm
—	O
with	O
higher	O
media	O
diversity	O
lessening	O
the	O
chances	O
—	O
their	O
results	O
suggest	O
that	O
certain	O
demographics	O
(	O
higher	O
age	O
and	O
male	O
)	O
along	O
with	O
certain	O
personality	O
traits	O
(	O
high	O
openness	O
)	O
correlate	O
positively	O
with	O
a	O
number	O
of	O
news	O
sources	O
consumed	O
by	O
individuals	O
.	O
</s>
<s>
Beyond	O
offering	O
different	O
individual	O
user	O
factors	O
that	O
may	O
influence	O
the	O
role	O
of	O
user	O
choice	O
,	O
this	O
study	O
also	O
raises	O
questions	O
and	O
associations	O
between	O
the	O
likelihood	O
of	O
users	O
being	O
caught	O
in	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
user	O
voting	O
behavior	O
.	O
</s>
<s>
The	B-Application
Facebook	I-Application
study	O
found	O
that	O
it	O
was	O
"	O
inconclusive	O
"	O
whether	O
or	O
not	O
the	O
algorithm	O
played	O
as	O
big	O
a	O
role	O
in	O
filtering	O
News	B-Application
Feeds	I-Application
as	O
people	O
assumed	O
.	O
</s>
<s>
The	O
study	O
also	O
found	O
that	O
"	O
individual	O
choice	O
,	O
"	O
or	O
confirmation	O
bias	O
,	O
likewise	O
affected	O
what	O
gets	O
filtered	O
out	O
of	O
News	B-Application
Feeds	I-Application
.	O
</s>
<s>
Some	O
social	O
scientists	O
criticized	O
this	O
conclusion	O
because	O
the	O
point	O
of	O
protesting	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
is	O
that	O
the	O
algorithms	O
and	O
individual	O
choice	O
work	O
together	O
to	O
filter	O
out	O
News	B-Application
Feeds	I-Application
.	O
</s>
<s>
They	O
also	O
criticized	O
Facebook	B-Application
's	O
small	O
sample	O
size	O
,	O
which	O
is	O
about	O
"	O
9%	O
of	O
actual	O
Facebook	B-Application
users	I-Application
,	O
"	O
and	O
the	O
fact	O
that	O
the	O
study	O
results	O
are	O
"	O
not	O
reproducible	O
"	O
due	O
to	O
the	O
fact	O
that	O
the	O
study	O
was	O
conducted	O
by	O
"	O
Facebook	B-Application
scientists	O
"	O
who	O
had	O
access	O
to	O
data	O
that	O
Facebook	B-Application
does	O
not	O
make	O
available	O
to	O
outside	O
researchers	O
.	O
</s>
<s>
Though	O
the	O
study	O
found	O
that	O
only	O
about	O
15	O
–	O
20%	O
of	O
the	O
average	O
user	O
's	O
Facebook	B-Application
friends	O
subscribe	O
to	O
the	O
opposite	O
side	O
of	O
the	O
political	O
spectrum	O
,	O
Julia	O
Kaman	O
from	O
Vox	O
theorized	O
that	O
this	O
could	O
have	O
potentially	O
positive	O
implications	O
for	O
viewpoint	O
diversity	O
.	O
</s>
<s>
Facebook	B-Application
may	O
foster	O
a	O
unique	O
environment	O
where	O
a	O
user	O
sees	O
and	O
possibly	O
interacts	O
with	O
content	O
posted	O
or	O
re-posted	O
by	O
these	O
"	O
second-tier	O
"	O
friends	O
.	O
</s>
<s>
Similarly	O
,	O
a	O
study	O
of	O
Twitter	B-Application
's	O
filter	B-Algorithm
bubbles	I-Algorithm
by	O
New	O
York	O
University	O
concluded	O
that	O
"	O
Individuals	O
now	O
have	O
access	O
to	O
a	O
wider	O
span	O
of	O
viewpoints	O
about	O
news	O
events	O
,	O
and	O
most	O
of	O
this	O
information	O
is	O
not	O
coming	O
through	O
the	O
traditional	O
channels	O
,	O
but	O
either	O
directly	O
from	O
political	O
actors	O
or	O
through	O
their	O
friends	O
and	O
relatives	O
.	O
</s>
<s>
According	O
to	O
these	O
studies	O
,	O
social	O
media	O
may	O
be	O
diversifying	O
information	O
and	O
opinions	O
users	O
come	O
into	O
contact	O
with	O
,	O
though	O
there	O
is	O
much	O
speculation	O
around	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
their	O
ability	O
to	O
create	O
deeper	O
political	O
polarization	O
.	O
</s>
<s>
Social	O
bots	O
have	O
been	O
utilized	O
by	O
different	O
researchers	O
to	O
test	O
polarization	O
and	O
related	O
effects	O
that	O
are	O
attributed	O
to	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
echo	O
chambers	O
.	O
</s>
<s>
A	O
2018	O
study	O
used	O
social	O
bots	O
on	O
Twitter	B-Application
to	O
test	O
deliberate	O
user	O
exposure	O
to	O
partisan	O
viewpoints	O
.	O
</s>
<s>
The	O
study	O
claimed	O
it	O
demonstrated	O
partisan	O
differences	O
between	O
exposure	O
to	O
differing	O
views	O
,	O
although	O
it	O
warned	O
that	O
the	O
findings	O
should	O
be	O
limited	O
to	O
party-registered	O
American	O
Twitter	B-Application
users	O
.	O
</s>
<s>
A	O
different	O
study	O
from	O
The	O
People	O
's	O
Republic	O
of	O
China	O
utilized	O
social	O
bots	O
on	O
Weibo	O
—	O
the	O
largest	O
social	O
media	O
platform	O
in	O
China	O
—	O
to	O
examine	O
the	O
structure	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
regarding	O
to	O
their	O
effects	O
on	O
polarization	O
.	O
</s>
<s>
By	O
utilizing	O
social	O
bots	O
instead	O
of	O
human	O
volunteers	O
and	O
focusing	O
more	O
on	O
information	O
polarization	O
rather	O
than	O
opinion-based	O
,	O
the	O
researchers	O
concluded	O
that	O
there	O
are	O
two	O
essential	O
elements	O
of	O
a	O
filter	B-Algorithm
bubble	I-Algorithm
:	O
a	O
large	O
concentration	O
of	O
users	O
around	O
a	O
single	O
topic	O
and	O
a	O
uni-directional	O
,	O
star-like	O
structure	O
that	O
impacts	O
key	O
information	O
flows	O
.	O
</s>
<s>
In	O
June	O
2018	O
,	O
the	O
platform	O
DuckDuckGo	B-Application
conducted	O
a	O
research	O
study	O
on	O
the	B-Application
Google	I-Application
Web	O
Browser	O
Platform	O
.	O
</s>
<s>
Google	B-Application
included	O
certain	O
links	O
for	O
some	O
that	O
it	O
did	O
not	O
include	O
for	O
other	O
participants	O
,	O
and	O
the	O
News	O
and	O
Videos	O
infoboxes	O
showed	O
significant	O
variation	O
.	O
</s>
<s>
Google	B-Application
publicly	O
disputed	O
these	O
results	O
saying	O
that	O
Search	O
Engine	O
Results	O
Page	O
(	O
SERP	O
)	O
personalization	O
is	O
mostly	O
a	O
myth	O
.	O
</s>
<s>
Google	B-Application
Search	O
Liaison	O
,	O
Danny	O
Sullivan	O
,	O
stated	O
that	O
“	O
Over	O
the	O
years	O
,	O
a	O
myth	O
has	O
developed	O
that	O
Google	B-Application
Search	O
personalizes	O
so	O
much	O
that	O
for	O
the	O
same	O
query	O
,	O
different	O
people	O
might	O
get	O
significantly	O
different	O
results	O
from	O
each	O
other	O
.	O
</s>
<s>
When	O
filter	B-Algorithm
bubbles	I-Algorithm
are	O
in	O
place	O
,	O
they	O
can	O
create	O
specific	O
moments	O
that	O
scientists	O
call	O
'	O
Whoa	O
 '	O
moments	O
.	O
</s>
<s>
"	O
Sat	O
down	O
and	O
opened	O
up	O
Facebook	B-Application
this	O
morning	O
while	O
having	O
my	O
coffee	O
,	O
and	O
there	O
they	O
were	O
two	O
ads	O
for	O
Nespresso	O
.	O
</s>
<s>
Several	O
designers	O
have	O
developed	O
tools	O
to	O
counteract	O
the	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
(	O
see	O
)	O
.	O
</s>
<s>
Swiss	O
radio	O
station	O
SRF	O
voted	O
the	O
word	O
filterblase	O
(	O
the	O
German	O
translation	O
of	O
filter	B-Algorithm
bubble	I-Algorithm
)	O
word	O
of	O
the	O
year	O
2016	O
.	O
</s>
<s>
In	O
The	B-Algorithm
Filter	I-Algorithm
Bubble	I-Algorithm
:	O
What	O
the	O
Internet	O
Is	O
Hiding	O
from	O
You	O
,	O
internet	O
activist	O
Eli	O
Pariser	O
highlights	O
how	O
the	O
increasing	O
occurrence	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
further	O
emphasizes	O
the	O
value	O
of	O
one	O
's	O
bridging	O
social	O
capital	O
as	O
defined	O
by	O
Robert	O
Putman	O
.	O
</s>
<s>
Pariser	O
argues	O
that	O
filter	B-Algorithm
bubbles	I-Algorithm
reinforce	O
a	O
sense	O
of	O
social	O
homogeneity	O
,	O
which	O
weakens	O
ties	O
between	O
people	O
with	O
potentially	O
diverging	O
interests	O
and	O
viewpoints	O
.	O
</s>
<s>
Fostering	O
one	O
's	O
bridging	O
capital	O
,	O
such	O
as	O
by	O
connecting	O
with	O
more	O
people	O
in	O
an	O
informal	O
setting	O
,	O
may	O
be	O
an	O
effective	O
way	O
to	O
reduce	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
phenomenon	O
.	O
</s>
<s>
Users	O
can	O
take	O
many	O
actions	O
to	O
burst	O
through	O
their	O
filter	B-Algorithm
bubbles	I-Algorithm
,	O
for	O
example	O
by	O
making	O
a	O
conscious	O
effort	O
to	O
evaluate	O
what	O
information	O
they	O
are	O
exposing	O
themselves	O
to	O
,	O
and	O
by	O
thinking	O
critically	O
about	O
whether	O
they	O
are	O
engaging	O
with	O
a	O
broad	O
range	O
of	O
content	O
.	O
</s>
<s>
Technology	O
can	O
also	O
play	O
a	O
valuable	O
role	O
in	O
combating	O
filter	B-Algorithm
bubbles	I-Algorithm
.	O
</s>
<s>
Some	O
additional	O
plug-ins	B-Application
,	O
such	O
as	O
Media	O
Bias	O
Fact	O
Check	O
,	O
aimed	O
to	O
help	O
people	O
step	O
out	O
of	O
their	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
make	O
them	O
aware	O
of	O
their	O
personal	O
perspectives	O
;	O
thus	O
,	O
these	O
media	O
show	O
content	O
that	O
contradicts	O
with	O
their	O
beliefs	O
and	O
opinions	O
.	O
</s>
<s>
For	O
instance	O
,	O
Escape	O
Your	O
Bubble	B-Algorithm
asks	O
users	O
to	O
indicate	O
a	O
specific	O
political	O
party	O
they	O
want	O
to	O
be	O
more	O
informed	O
about	O
.	O
</s>
<s>
The	O
plug-in	B-Application
will	O
then	O
suggest	O
articles	O
from	O
well-established	O
sources	O
to	O
read	O
relating	O
to	O
that	O
political	O
party	O
,	O
encouraging	O
users	O
to	O
become	O
more	O
educated	O
about	O
the	O
other	O
party	O
.	O
</s>
<s>
In	O
addition	O
to	O
plug-ins	B-Application
,	O
there	O
are	O
apps	O
created	O
with	O
the	O
mission	O
of	O
encouraging	O
users	O
to	O
open	O
their	O
echo	O
chambers	O
.	O
</s>
<s>
UnFound.news	O
offers	O
an	O
AI	B-Application
(	O
Artificial	B-Application
Intelligence	I-Application
)	O
curated	O
news	O
app	O
to	O
readers	O
presenting	O
them	O
news	O
from	O
diverse	O
and	O
distinct	O
perspectives	O
,	O
helping	O
them	O
form	O
rationale	O
and	O
informed	O
opinion	O
rather	O
than	O
succumbing	O
to	O
their	O
own	O
biases	O
.	O
</s>
<s>
Although	O
apps	O
and	O
plug-ins	B-Application
are	O
tools	O
humans	O
can	O
use	O
,	O
Eli	O
Pariser	O
stated	O
"	O
certainly	O
,	O
there	O
is	O
some	O
individual	O
responsibility	O
here	O
to	O
really	O
seek	O
out	O
new	O
sources	O
and	O
people	O
who	O
are	O
n't	O
like	O
you.	O
"	O
</s>
<s>
Since	O
web-based	O
advertising	O
can	O
further	O
the	O
effect	O
of	O
the	B-Algorithm
filter	I-Algorithm
bubbles	I-Algorithm
by	O
exposing	O
users	O
to	O
more	O
of	O
the	O
same	O
content	O
,	O
users	O
can	O
block	O
much	O
advertising	O
by	O
deleting	O
their	O
search	O
history	O
,	O
turning	O
off	O
targeted	O
ads	O
,	O
and	O
downloading	O
browser	O
extensions	O
.	O
</s>
<s>
Extensions	O
such	O
as	O
Escape	O
your	O
Bubble	B-Algorithm
for	O
Google	B-Application
Chrome	O
aim	O
to	O
help	O
curate	O
content	O
and	O
prevent	O
users	O
from	O
only	O
being	O
exposed	O
to	O
biased	O
information	O
,	O
while	O
Mozilla	B-Application
Firefox	I-Application
extensions	O
such	O
as	O
Lightbeam	O
and	O
Self-Destructing	O
Cookies	B-Application
enable	O
users	O
to	O
visualize	O
how	O
their	O
data	O
is	O
being	O
tracked	O
,	O
and	O
lets	O
them	O
remove	O
some	O
of	O
the	O
tracking	O
cookies	B-Application
.	O
</s>
<s>
Some	O
use	O
anonymous	O
or	O
non-personalized	O
search	O
engines	O
such	O
as	O
YaCy	B-Application
,	O
DuckDuckGo	B-Application
,	O
Qwant	O
,	O
Startpage.com,	O
Disconnect	B-Operating_System
,	O
and	O
Searx	B-Application
in	O
order	O
to	O
prevent	O
companies	O
from	O
gathering	O
their	O
web-search	O
data	O
.	O
</s>
<s>
The	O
European	O
Union	O
is	O
taking	O
measures	O
to	O
lessen	O
the	O
effect	O
of	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
.	O
</s>
<s>
The	O
European	O
Parliament	O
is	O
sponsoring	O
inquiries	O
into	O
how	O
filter	B-Algorithm
bubbles	I-Algorithm
affect	O
people	O
's	O
ability	O
to	O
access	O
diverse	O
news	O
.	O
</s>
<s>
In	O
light	O
of	O
recent	O
concerns	O
about	O
information	O
filtering	O
on	O
social	O
media	O
,	O
Facebook	B-Application
acknowledged	O
the	O
presence	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
has	O
taken	O
strides	O
toward	O
removing	O
them	O
.	O
</s>
<s>
In	O
January	O
2017	O
,	O
Facebook	B-Application
removed	O
personalization	O
from	O
its	O
Trending	O
Topics	O
list	O
in	O
response	O
to	O
problems	O
with	O
some	O
users	O
not	O
seeing	O
highly	O
talked-about	O
events	O
there	O
.	O
</s>
<s>
Facebook	B-Application
's	O
strategy	O
is	O
to	O
reverse	O
the	O
Related	O
Articles	O
feature	O
that	O
it	O
had	O
implemented	O
in	O
2013	O
,	O
which	O
would	O
post	O
related	O
news	O
stories	O
after	O
the	O
user	O
read	O
a	O
shared	O
article	O
.	O
</s>
<s>
Facebook	B-Application
is	O
also	O
attempting	O
to	O
go	O
through	O
a	O
vetting	O
process	O
whereby	O
only	O
articles	O
from	O
reputable	O
sources	O
will	O
be	O
shown	O
.	O
</s>
<s>
Along	O
with	O
the	O
founder	O
of	O
Craigslist	O
and	O
a	O
few	O
others	O
,	O
Facebook	B-Application
has	O
invested	O
$14	O
million	O
into	O
efforts	O
"	O
to	O
increase	O
trust	O
in	O
journalism	O
around	O
the	O
world	O
,	O
and	O
to	O
better	O
inform	O
the	O
public	O
conversation	O
"	O
.	O
</s>
<s>
Similarly	O
,	O
Google	B-Application
,	O
as	O
of	O
January	O
30	O
,	O
2018	O
,	O
has	O
also	O
acknowledged	O
the	O
existence	O
of	O
a	O
filter	B-Algorithm
bubble	I-Algorithm
difficulties	O
within	O
its	O
platform	O
.	O
</s>
<s>
Because	O
current	O
Google	B-Application
searches	O
pull	O
algorithmically	O
ranked	O
results	O
based	O
upon	O
"	O
authoritativeness	O
"	O
and	O
"	O
relevancy	O
"	O
which	O
show	O
and	O
hide	O
certain	O
search	O
results	O
,	O
Google	B-Application
is	O
seeking	O
to	O
combat	O
this	O
.	O
</s>
<s>
By	O
training	O
its	O
search	O
engine	O
to	O
recognize	O
the	O
intent	O
of	O
a	O
search	O
inquiry	O
rather	O
than	O
the	O
literal	O
syntax	O
of	O
the	O
question	O
,	O
Google	B-Application
is	O
attempting	O
to	O
limit	O
the	O
size	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
.	O
</s>
<s>
In	O
April	O
2017	O
news	O
surfaced	O
that	O
Facebook	B-Application
,	O
Mozilla	B-Operating_System
,	O
and	O
Craigslist	O
contributed	O
to	O
the	O
majority	O
of	O
a	O
$14M	O
donation	O
to	O
CUNY	O
's	O
"	O
News	O
Integrity	O
Initiative	O
,	O
"	O
poised	O
at	O
eliminating	O
fake	O
news	O
and	O
creating	O
more	O
honest	O
news	O
media	O
.	O
</s>
<s>
Later	O
,	O
in	O
August	O
,	O
Mozilla	B-Operating_System
,	O
makers	O
of	O
the	O
Firefox	B-Application
web	I-Application
browser	I-Application
,	O
announced	O
the	O
formation	O
of	O
the	O
Mozilla	B-Operating_System
Information	O
Trust	O
Initiative	O
(	O
MITI	O
)	O
.	O
</s>
<s>
The	O
+MITI	O
would	O
serve	O
as	O
a	O
collective	O
effort	O
to	O
develop	O
products	O
,	O
research	O
,	O
and	O
community-based	O
solutions	O
to	O
combat	O
the	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
the	O
proliferation	O
of	O
fake	O
news	O
.	O
</s>
<s>
Mozilla	B-Operating_System
's	O
Open	O
Innovation	O
team	O
leads	O
the	O
initiative	O
,	O
striving	O
to	O
combat	O
misinformation	O
,	O
with	O
a	O
specific	O
focus	O
on	O
the	O
product	O
with	O
regards	O
to	O
literacy	O
,	O
research	O
and	O
creative	O
interventions	O
.	O
</s>
<s>
As	O
the	O
popularity	O
of	O
cloud	B-Architecture
services	I-Architecture
increases	O
,	O
personalized	O
algorithms	O
used	O
to	O
construct	O
filter	B-Algorithm
bubbles	I-Algorithm
are	O
expected	O
to	O
become	O
more	O
widespread	O
.	O
</s>
<s>
Scholars	O
have	O
begun	O
considering	O
the	O
effect	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
on	O
the	O
users	O
of	O
social	O
media	O
from	O
an	O
ethical	O
standpoint	O
,	O
particularly	O
concerning	O
the	O
areas	O
of	O
personal	O
freedom	O
,	O
security	O
,	O
and	O
information	O
bias	O
.	O
</s>
<s>
Filter	B-Algorithm
bubbles	I-Algorithm
in	O
popular	O
social	O
media	O
and	O
personalized	B-Application
search	I-Application
sites	O
can	O
determine	O
the	O
particular	O
content	O
seen	O
by	O
users	O
,	O
often	O
without	O
their	O
direct	O
consent	O
or	O
cognizance	O
,	O
due	O
to	O
the	O
algorithms	O
used	O
to	O
curate	O
that	O
content	O
.	O
</s>
<s>
Critics	O
of	O
the	O
use	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
speculate	O
that	O
individuals	O
may	O
lose	O
autonomy	O
over	O
their	O
own	O
social	O
media	O
experience	O
and	O
have	O
their	O
identities	O
socially	O
constructed	O
as	O
a	O
result	O
of	O
the	O
pervasiveness	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
.	O
</s>
<s>
Technologists	O
,	O
social	O
media	O
engineers	O
,	O
and	O
computer	O
specialists	O
have	O
also	O
examined	O
the	O
prevalence	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
.	O
</s>
<s>
Mark	O
Zuckerberg	O
,	O
founder	O
of	O
Facebook	B-Application
,	O
and	O
Eli	O
Pariser	O
,	O
author	O
of	O
The	B-Algorithm
Filter	I-Algorithm
Bubble	I-Algorithm
,	O
have	O
expressed	O
concerns	O
regarding	O
the	O
risks	O
of	O
privacy	O
and	O
information	O
polarization	O
.	O
</s>
<s>
The	O
information	O
of	O
the	O
users	O
of	O
personalized	B-Application
search	I-Application
engines	O
and	O
social	O
media	O
platforms	O
is	O
not	O
private	O
,	O
though	O
some	O
people	O
believe	O
it	O
should	O
be	O
.	O
</s>
<s>
Some	O
scholars	O
have	O
expressed	O
concerns	O
regarding	O
the	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
on	O
individual	O
and	O
social	O
well-being	O
,	O
i.e.	O
</s>
<s>
A	O
2019	O
multi-disciplinary	O
book	O
reported	O
research	O
and	O
perspectives	O
on	O
the	O
roles	O
filter	B-Algorithm
bubbles	I-Algorithm
play	O
in	O
regards	O
to	O
health	O
misinformation	O
.	O
</s>
<s>
alternative	O
medicine	O
and	O
pseudoscience	O
)	O
as	O
well	O
as	O
potential	O
remedies	O
to	O
the	O
negative	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
echo	O
chambers	O
on	O
different	O
topics	O
in	O
health	O
discourse	O
.	O
</s>
<s>
A	O
2016	O
study	O
on	O
the	O
potential	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
on	O
search	O
engine	O
results	O
related	O
to	O
suicide	O
found	O
that	O
algorithms	O
play	O
an	O
important	O
role	O
in	O
whether	O
or	O
not	O
helplines	O
and	O
similar	O
search	O
results	O
are	O
displayed	O
to	O
users	O
and	O
discussed	O
the	O
implications	O
their	O
research	O
may	O
have	O
for	O
health	O
policies	O
.	O
</s>
<s>
Another	O
2016	O
study	O
from	O
the	O
Croatian	O
Medical	O
journal	O
proposed	O
some	O
strategies	O
for	O
mitigating	O
the	O
potentially	O
harmful	O
effects	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
on	O
health	O
information	O
,	O
such	O
as	O
:	O
informing	O
the	O
public	O
more	O
about	O
filter	B-Algorithm
bubbles	I-Algorithm
and	O
their	O
associated	O
effects	O
,	O
users	O
choosing	O
to	O
try	O
alternative	O
[	O
to	O
Google ]	O
search	O
engines	O
,	O
and	O
more	O
explanation	O
of	O
the	O
processes	O
search	O
engines	O
use	O
to	O
determine	O
their	O
displayed	O
results	O
.	O
</s>
<s>
Since	O
the	O
content	O
seen	O
by	O
individual	O
social	O
media	O
users	O
is	O
influenced	O
by	O
algorithms	O
that	O
produce	O
filter	B-Algorithm
bubbles	I-Algorithm
,	O
users	O
of	O
social	O
media	O
platforms	O
are	O
more	O
susceptible	O
to	O
confirmation	O
bias	O
,	O
and	O
may	O
be	O
exposed	O
to	O
biased	O
,	O
misleading	O
information	O
.	O
</s>
<s>
In	O
light	O
of	O
the	O
2016	O
U.S.	O
presidential	O
election	O
scholars	O
have	O
likewise	O
expressed	O
concerns	O
about	O
the	O
effect	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
on	O
democracy	O
and	O
democratic	O
processes	O
,	O
as	O
well	O
as	O
the	O
rise	O
of	O
"	O
ideological	O
media	O
"	O
.	O
</s>
<s>
These	O
scholars	O
fear	O
that	O
users	O
will	O
be	O
unable	O
to	O
"[think]	O
beyond	O
 [ their ] 	O
narrow	O
self-interest	O
"	O
as	O
filter	B-Algorithm
bubbles	I-Algorithm
create	O
personalized	O
social	O
feeds	O
,	O
isolating	O
them	O
from	O
diverse	O
points	O
of	O
view	O
and	O
their	O
surrounding	O
communities	O
.	O
</s>
<s>
For	O
this	O
reason	O
,	O
an	O
increasingly	O
discussed	O
possibility	O
is	O
to	O
design	O
social	O
media	O
with	O
more	O
serendipity	O
,	O
that	O
is	O
,	O
to	O
proactively	O
recommend	O
content	O
that	O
lies	O
outside	O
one	O
's	O
filter	B-Algorithm
bubble	I-Algorithm
,	O
including	O
challenging	O
political	O
information	O
and	O
,	O
eventually	O
,	O
to	O
provide	O
empowering	O
filters	O
and	O
tools	O
to	O
users	O
.	O
</s>
<s>
A	O
related	O
concern	O
is	O
in	O
fact	O
how	O
filter	B-Algorithm
bubbles	I-Algorithm
contribute	O
to	O
the	O
proliferation	O
of	O
"	O
fake	O
news	O
"	O
and	O
how	O
this	O
may	O
influence	O
political	O
leaning	O
,	O
including	O
how	O
users	O
vote	O
.	O
</s>
<s>
Revelations	O
in	O
March	O
2018	O
of	O
Cambridge	O
Analytica	O
's	O
harvesting	O
and	O
use	O
of	O
user	O
data	O
for	O
at	O
least	O
87	O
million	O
Facebook	B-Application
profiles	O
during	O
the	O
2016	O
presidential	O
election	O
highlight	O
the	O
ethical	O
implications	O
of	O
filter	B-Algorithm
bubbles	I-Algorithm
.	O
</s>
<s>
Access	O
to	O
user	O
data	O
by	O
third	O
parties	O
such	O
as	O
Cambridge	O
Analytica	O
can	O
exasperate	O
and	O
amplify	O
existing	O
filter	B-Algorithm
bubbles	I-Algorithm
users	O
have	O
created	O
,	O
artificially	O
increasing	O
existing	O
biases	O
and	O
further	O
divide	O
societies	O
.	O
</s>
<s>
Filter	B-Algorithm
bubbles	I-Algorithm
have	O
stemmed	O
from	O
a	O
surge	O
in	O
media	O
personalization	O
,	O
which	O
can	O
trap	O
users	O
.	O
</s>
<s>
The	O
use	O
of	O
AI	B-Application
to	O
personalize	O
offerings	O
can	O
lead	O
to	O
users	O
viewing	O
only	O
content	O
that	O
reinforces	O
their	O
own	O
viewpoints	O
without	O
challenging	O
them	O
.	O
</s>
<s>
Social	O
media	O
websites	O
like	O
Facebook	B-Application
may	O
also	O
present	O
content	O
in	O
a	O
way	O
that	O
makes	O
it	O
difficult	O
for	O
users	O
to	O
determine	O
the	O
source	O
of	O
the	O
content	O
,	O
leading	O
them	O
to	O
decide	O
for	O
themselves	O
whether	O
the	O
source	O
is	O
reliable	O
or	O
fake	O
.	O
</s>
<s>
The	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
may	O
cause	O
the	O
person	O
to	O
see	O
any	O
opposing	O
viewpoints	O
as	O
incorrect	O
and	O
so	O
could	O
allow	O
the	O
media	O
to	O
force	O
views	O
onto	O
consumers	O
.	O
</s>
<s>
Researches	O
explain	O
that	O
the	B-Algorithm
filter	I-Algorithm
bubble	I-Algorithm
reinforces	O
what	O
one	O
is	O
already	O
thinking	O
.	O
</s>
