<s>
In	O
information	O
science	O
,	O
profiling	B-Algorithm
refers	O
to	O
the	O
process	O
of	O
construction	O
and	O
application	O
of	O
user	O
profiles	O
generated	O
by	O
computerized	O
data	B-General_Concept
analysis	I-General_Concept
.	O
</s>
<s>
Other	O
than	O
a	O
discussion	O
of	O
profiling	B-Algorithm
technologies	O
or	O
population	B-Algorithm
profiling	I-Algorithm
,	O
the	O
notion	O
of	O
profiling	B-Algorithm
in	O
this	O
sense	O
is	O
not	O
just	O
about	O
the	O
construction	O
of	O
profiles	O
,	O
but	O
also	O
concerns	O
the	O
application	O
of	O
group	O
profiles	O
to	O
individuals	O
,	O
e.g.	O
,	O
in	O
the	O
cases	O
of	O
credit	O
scoring	O
,	O
price	O
discrimination	O
,	O
or	O
identification	O
of	O
security	O
risks	O
.	O
</s>
<s>
Profiling	B-Algorithm
is	O
being	O
used	O
in	O
fraud	B-Application
prevention	I-Application
,	O
ambient	B-Operating_System
intelligence	I-Operating_System
,	O
and	O
consumer	O
analytics	O
.	O
</s>
<s>
Statistical	O
methods	O
of	O
profiling	B-Algorithm
include	O
Knowledge	B-Application
Discovery	I-Application
in	I-Application
Databases	I-Application
(	O
KDD	O
)	O
.	O
</s>
<s>
The	O
technical	O
process	O
of	O
profiling	B-Algorithm
can	O
be	O
separated	O
in	O
several	O
steps	O
:	O
</s>
<s>
Preliminary	O
grounding	O
:	O
The	O
profiling	B-Algorithm
process	O
starts	O
with	O
a	O
specification	O
of	O
the	O
applicable	O
problem	O
domain	O
and	O
the	O
identification	O
of	O
the	O
goals	O
of	O
analysis	O
.	O
</s>
<s>
Data	B-General_Concept
preparation	I-General_Concept
:	O
The	O
data	O
are	O
preprocessed	O
for	O
removing	O
noise	O
and	O
reducing	O
complexity	O
by	O
eliminating	O
attributes	O
.	O
</s>
<s>
Data	B-Application
mining	I-Application
:	O
The	O
data	O
are	O
analysed	O
with	O
the	O
algorithm	O
or	O
heuristics	O
developed	O
to	O
suit	O
the	O
data	O
,	O
model	O
and	O
goals	O
.	O
</s>
<s>
However	O
,	O
profiling	B-Algorithm
also	O
refers	O
to	O
the	O
application	O
of	O
profiles	O
,	O
meaning	O
the	O
usage	O
of	O
profiles	O
for	O
the	O
identification	O
or	O
categorization	O
of	O
groups	O
or	O
individual	O
persons	O
.	O
</s>
<s>
The	O
interpretation	O
of	O
profiles	O
can	O
lead	O
to	O
the	O
reiterant	O
–	O
possibly	O
real-time	O
–	O
fine-tuning	O
of	O
specific	O
previous	O
steps	O
in	O
the	O
profiling	B-Algorithm
process	O
.	O
</s>
<s>
The	O
process	O
of	O
profiling	B-Algorithm
is	O
both	O
dynamic	O
and	O
adaptive	O
.	O
</s>
<s>
A	O
good	O
illustration	O
of	O
the	O
dynamic	O
and	O
adaptive	O
nature	O
of	O
profiling	B-Algorithm
is	O
the	O
Cross-Industry	B-Algorithm
Standard	I-Algorithm
Process	I-Algorithm
for	I-Algorithm
Data	I-Algorithm
Mining	I-Algorithm
(	O
CRISP-DM	B-Algorithm
)	O
.	O
</s>
<s>
In	O
order	O
to	O
clarify	O
the	O
nature	O
of	O
profiling	B-Algorithm
technologies	O
,	O
some	O
crucial	O
distinctions	O
have	O
to	O
be	O
made	O
between	O
different	O
types	O
of	O
profiling	B-Algorithm
practices	I-Algorithm
,	O
apart	O
from	O
the	O
distinction	O
between	O
the	O
construction	O
and	O
the	O
application	O
of	O
profiles	O
.	O
</s>
<s>
The	O
main	O
distinctions	O
are	O
those	O
between	O
bottom-up	O
and	O
top-down	O
profiling	B-Algorithm
(	O
or	O
supervised	O
and	O
unsupervised	B-General_Concept
learning	I-General_Concept
)	O
,	O
and	O
between	O
individual	O
and	O
group	O
profiles	O
.	O
</s>
<s>
This	O
is	O
called	O
top-down	O
profiling	B-Algorithm
or	O
supervised	B-General_Concept
learning	I-General_Concept
.	O
</s>
<s>
The	O
result	O
of	O
this	O
type	O
of	O
profiling	B-Algorithm
is	O
the	O
verification	O
or	O
refutation	O
of	O
the	O
hypothesis	O
.	O
</s>
<s>
One	O
could	O
also	O
speak	O
of	O
deductive	O
profiling	B-Algorithm
.	O
</s>
<s>
On	O
the	O
other	O
hand	O
,	O
profiles	O
can	O
be	O
generated	O
by	O
exploring	O
a	O
data	O
base	O
,	O
using	O
the	O
data	B-Application
mining	I-Application
process	O
to	O
detect	O
patterns	O
in	O
the	O
data	O
base	O
that	O
were	O
not	O
previously	O
hypothesized	O
.	O
</s>
<s>
This	O
is	O
called	O
unsupervised	B-General_Concept
learning	I-General_Concept
.	O
</s>
<s>
First	O
,	O
unsupervised	B-General_Concept
learning	I-General_Concept
algorithms	O
seem	O
to	O
allow	O
the	O
construction	O
of	O
a	O
new	O
type	O
of	O
knowledge	O
,	O
not	O
based	O
on	O
hypothesis	O
developed	O
by	O
a	O
researcher	O
and	O
not	O
based	O
on	O
causal	O
or	O
motivational	O
relations	O
but	O
exclusively	O
based	O
on	O
stochastical	O
correlations	O
.	O
</s>
<s>
Second	O
,	O
unsupervised	B-General_Concept
learning	I-General_Concept
algorithms	O
thus	O
seem	O
to	O
allow	O
for	O
an	O
inductive	O
type	O
of	O
knowledge	O
construction	O
that	O
does	O
not	O
require	O
theoretical	O
justification	O
or	O
causal	O
explanation	O
.	O
</s>
<s>
In	O
the	O
process	O
of	O
collecting	O
and	O
aggregating	O
data	O
into	O
a	O
database	O
(	O
the	O
first	O
three	O
steps	O
of	O
the	O
process	O
of	O
profile	O
construction	O
)	O
,	O
translations	O
are	O
made	O
from	O
real-life	O
events	O
to	O
machine-readable	B-General_Concept
data	I-General_Concept
.	O
</s>
<s>
In	O
the	O
case	O
of	O
machine	O
profiling	B-Algorithm
,	O
potential	O
bias	O
is	O
not	O
informed	O
by	O
common	O
sense	O
prejudice	O
or	O
what	O
psychologists	O
call	O
stereotyping	O
,	O
but	O
by	O
the	O
computer	O
techniques	O
employed	O
in	O
the	O
initial	O
steps	O
of	O
the	O
process	O
.	O
</s>
<s>
When	O
a	O
profile	O
is	O
constructed	O
with	O
the	O
data	O
of	O
a	O
single	O
person	O
,	O
this	O
is	O
called	O
individual	O
profiling	B-Algorithm
.	O
</s>
<s>
This	O
kind	O
of	O
profiling	B-Algorithm
is	O
used	O
to	O
discover	O
the	O
particular	O
characteristics	O
of	O
a	O
certain	O
individual	O
,	O
to	O
enable	O
unique	O
identification	O
or	O
the	O
provision	O
of	O
personalized	O
services	O
.	O
</s>
<s>
However	O
,	O
personalized	O
servicing	O
is	O
most	O
often	O
also	O
based	O
on	O
group	O
profiling	B-Algorithm
,	O
which	O
allows	O
categorisation	O
of	O
a	O
person	O
as	O
a	O
certain	O
type	O
of	O
person	O
,	O
based	O
on	O
the	O
fact	O
that	O
her	O
profile	O
matches	O
with	O
a	O
profile	O
that	O
has	O
been	O
constructed	O
on	O
the	O
basis	O
of	O
massive	O
amounts	O
of	O
data	O
about	O
massive	O
numbers	O
of	O
other	O
people	O
.	O
</s>
<s>
A	O
group	O
profile	O
can	O
refer	O
to	O
the	O
result	O
of	O
data	B-Application
mining	I-Application
in	O
data	O
sets	O
that	O
refer	O
to	O
an	O
existing	O
community	O
that	O
considers	O
itself	O
as	O
such	O
,	O
like	O
a	O
religious	O
group	O
,	O
a	O
tennis	O
club	O
,	O
a	O
university	O
,	O
a	O
political	O
party	O
etc	O
.	O
</s>
<s>
If	O
an	O
individual	O
profile	O
is	O
applied	O
to	O
the	O
individual	O
that	O
it	O
was	O
mined	O
from	O
,	O
then	O
that	O
is	O
direct	O
individual	O
profiling	B-Algorithm
.	O
</s>
<s>
If	O
a	O
group	O
profile	O
is	O
applied	O
to	O
an	O
individual	O
whose	O
data	O
match	O
the	O
profile	O
,	O
then	O
that	O
is	O
indirect	O
individual	O
profiling	B-Algorithm
,	O
because	O
the	O
profile	O
was	O
generated	O
using	O
data	O
of	O
other	O
people	O
.	O
</s>
<s>
Similarly	O
,	O
if	O
a	O
group	O
profile	O
is	O
applied	O
to	O
the	O
group	O
that	O
it	O
was	O
mined	O
from	O
,	O
then	O
that	O
is	O
direct	O
group	O
profiling	B-Algorithm
.	O
</s>
<s>
However	O
,	O
in	O
as	O
far	O
as	O
the	O
application	O
of	O
a	O
group	O
profile	O
to	O
a	O
group	O
implies	O
the	O
application	O
of	O
the	O
group	O
profile	O
to	O
individual	O
members	O
of	O
the	O
group	O
,	O
it	O
makes	O
sense	O
to	O
speak	O
of	O
indirect	O
group	O
profiling	B-Algorithm
,	O
especially	O
if	O
the	O
group	O
profile	O
is	O
non-distributive	O
.	O
</s>
<s>
This	O
has	O
far-reaching	O
implications	O
for	O
the	O
accuracy	O
of	O
indirect	O
individual	O
profiling	B-Algorithm
based	O
on	O
data	O
matching	O
with	O
non-distributive	O
group	O
profiles	O
.	O
</s>
<s>
In	O
the	O
financial	O
sector	O
,	O
institutions	O
use	O
profiling	B-Algorithm
technologies	O
for	O
fraud	B-Application
prevention	I-Application
and	O
credit	O
scoring	O
.	O
</s>
<s>
On	O
the	O
basis	O
of	O
the	O
extensive	O
group	O
,	O
profiling	B-Algorithm
customers	O
are	O
assigned	O
a	O
certain	O
scoring	O
value	O
that	O
indicates	O
their	O
creditworthiness	O
.	O
</s>
<s>
Financial	O
institutions	O
like	O
banks	O
and	O
insurance	O
companies	O
also	O
use	O
group	O
profiling	B-Algorithm
to	O
detect	O
fraud	O
or	O
money-laundering	O
.	O
</s>
<s>
Profiling	B-Algorithm
can	O
also	O
be	O
used	O
to	O
support	O
people	O
at	O
work	O
,	O
and	O
also	O
for	O
learning	O
,	O
by	O
intervening	O
in	O
the	O
design	O
of	O
adaptive	O
hypermedia	O
systems	O
personalizing	O
the	O
interaction	O
.	O
</s>
<s>
Consumer	O
profiling	B-Algorithm
is	O
a	O
form	O
of	O
customer	O
analytics	O
,	O
where	O
customer	O
data	O
is	O
used	O
to	O
make	O
decisions	O
on	O
product	O
promotion	O
,	O
the	O
pricing	O
of	O
products	O
,	O
as	O
well	O
as	O
personalized	O
advertising	O
.	O
</s>
<s>
To	O
establish	O
predictive	B-General_Concept
models	I-General_Concept
on	O
the	O
basis	O
of	O
existing	O
databases	O
,	O
the	O
Knowledge	B-Application
Discovery	I-Application
in	I-Application
Databases	I-Application
(	O
KDD	O
)	O
statistical	O
method	O
is	O
used	O
.	O
</s>
<s>
In	O
a	O
built	O
environment	O
with	O
ambient	B-Operating_System
intelligence	I-Operating_System
everyday	O
objects	O
have	O
built-in	O
sensors	O
and	O
embedded	B-Architecture
systems	I-Architecture
that	O
allow	O
objects	O
to	O
recognise	O
and	O
respond	O
to	O
the	O
presence	O
and	O
needs	O
of	O
individuals	O
.	O
</s>
<s>
Ambient	B-Operating_System
intelligence	I-Operating_System
relies	O
on	O
automated	O
profiling	B-Algorithm
and	O
human	O
–	O
computer	O
interaction	O
designs	O
.	O
</s>
<s>
Early	O
examples	O
of	O
consumer	O
electronics	O
with	O
ambient	B-Operating_System
intelligence	I-Operating_System
include	O
mobile	B-Application
apps	I-Application
,	O
augmented	B-General_Concept
reality	I-General_Concept
and	O
location-based	B-Application
service	I-Application
.	O
</s>
<s>
Profiling	B-Algorithm
technologies	O
have	O
raised	O
a	O
host	O
of	O
ethical	O
,	O
legal	O
and	O
other	O
issues	O
including	O
privacy	O
,	O
equality	O
,	O
due	B-Application
process	I-Application
,	O
security	O
and	O
liability	O
.	O
</s>
<s>
Numerous	O
authors	O
have	O
warned	O
against	O
the	O
affordances	O
of	O
a	O
new	O
technological	O
infrastructure	O
that	O
could	O
emerge	O
on	O
the	O
basis	O
of	O
semi-autonomic	O
profiling	B-Algorithm
technologies	O
.	O
</s>
<s>
Profiling	B-Algorithm
technologies	O
make	O
possible	O
a	O
far-reaching	O
monitoring	O
of	O
an	O
individual	O
's	O
behaviour	O
and	O
preferences	O
.	O
</s>
<s>
Profiling	B-Algorithm
technologies	O
are	O
by	O
their	O
very	O
nature	O
discriminatory	O
tools	O
.	O
</s>
<s>
In	O
most	O
cases	O
they	O
will	O
not	O
be	O
aware	O
of	O
this	O
,	O
since	O
profiling	B-Algorithm
practices	I-Algorithm
are	O
mostly	O
invisible	O
and	O
the	O
profiles	O
themselves	O
are	O
often	O
protected	O
by	O
intellectual	O
property	O
or	O
trade	O
secret	O
.	O
</s>
<s>
One	O
of	O
the	O
problems	O
underlying	O
potential	O
violations	O
of	O
privacy	O
and	O
non-discrimination	O
is	O
that	O
the	O
process	O
of	O
profiling	B-Algorithm
is	O
more	O
often	O
than	O
not	O
invisible	O
for	O
those	O
that	O
are	O
being	O
profiled	O
.	O
</s>
<s>
This	O
disturbs	O
principles	O
of	O
due	B-Application
process	I-Application
:	O
if	O
a	O
person	O
has	O
no	O
access	O
to	O
information	O
on	O
the	O
basis	O
of	O
which	O
they	O
are	O
withheld	O
benefits	O
or	O
attributed	O
certain	O
risks	O
,	O
they	O
cannot	O
contest	O
the	O
way	O
they	O
are	O
being	O
treated	O
.	O
</s>
<s>
Is	O
the	O
software	O
programmer	O
,	O
the	O
profiling	B-Algorithm
service	O
provider	O
,	O
or	O
the	O
profiled	O
user	O
to	O
be	O
held	O
accountable	O
?	O
</s>
<s>
This	O
issue	O
of	O
liability	O
is	O
especially	O
complex	O
in	O
the	O
case	O
the	O
application	O
and	O
decisions	O
on	O
profiles	O
have	O
also	O
become	O
automated	O
like	O
in	O
Autonomic	B-Architecture
Computing	I-Architecture
or	O
ambient	B-Operating_System
intelligence	I-Operating_System
decisions	O
of	O
automated	B-General_Concept
decisions	I-General_Concept
based	O
on	O
profiling	B-Algorithm
.	O
</s>
