<s>
A	O
Twitter	B-Protocol
bot	I-Protocol
is	O
a	O
type	O
of	O
software	O
bot	B-Protocol
that	O
controls	O
a	O
Twitter	B-Application
account	I-Application
via	O
the	O
Twitter	B-Application
API	B-General_Concept
.	O
</s>
<s>
The	O
social	O
bot	B-Protocol
software	O
may	O
autonomously	O
perform	O
actions	O
such	O
as	O
tweeting	O
,	O
re-tweeting	O
,	O
liking	O
,	O
following	O
,	O
unfollowing	O
,	O
or	O
direct	O
messaging	O
other	O
accounts	O
.	O
</s>
<s>
The	O
automation	O
of	O
Twitter	B-Application
accounts	I-Application
is	O
governed	O
by	O
a	O
set	O
of	O
automation	O
rules	O
that	O
outline	O
proper	O
and	O
improper	O
uses	O
of	O
automation	O
.	O
</s>
<s>
Improper	O
usage	O
includes	O
circumventing	O
API	B-General_Concept
rate	O
limits	O
,	O
violating	O
user	O
privacy	O
,	O
spamming	O
,	O
and	O
sockpuppeting	O
.	O
</s>
<s>
Twitter	B-Protocol
bots	I-Protocol
may	O
be	O
part	O
of	O
a	O
larger	O
botnet	B-Architecture
.	O
</s>
<s>
Twitter	B-Application
's	O
policies	O
do	O
allow	O
non-abusive	O
bots	B-Protocol
,	O
such	O
as	O
those	O
created	O
as	O
a	O
benign	O
hobby	O
or	O
for	O
artistic	O
purposes	O
,	O
or	O
posting	O
helpful	O
information	O
.	O
</s>
<s>
Many	O
non-malicious	O
bots	B-Protocol
are	O
popular	O
for	O
their	O
entertainment	O
value	O
.	O
</s>
<s>
However	O
,	O
as	O
technology	O
and	O
the	O
creativity	O
of	O
bot-makers	O
improves	O
,	O
so	O
does	O
the	O
potential	O
for	O
Twitter	B-Protocol
bots	I-Protocol
that	O
fill	O
social	O
needs	O
.	O
</s>
<s>
@tinycarebot	O
is	O
a	O
Twitter	B-Protocol
bot	I-Protocol
that	O
encourages	O
followers	O
to	O
practice	O
self	O
care	O
,	O
and	O
brands	O
are	O
increasingly	O
using	O
automated	O
Twitter	B-Protocol
bots	I-Protocol
to	O
engage	B-Application
with	I-Application
customers	I-Application
in	I-Application
interactive	I-Application
ways	I-Application
.	O
</s>
<s>
Concerns	O
about	O
political	O
Twitter	B-Protocol
bots	I-Protocol
include	O
the	O
promulgation	O
of	O
malicious	O
content	O
,	O
increased	O
polarization	O
,	O
and	O
the	O
spreading	O
of	O
fake	O
news	O
.	O
</s>
<s>
A	O
subset	O
of	O
Twitter	B-Protocol
bots	I-Protocol
programmed	O
to	O
complete	O
social	O
tasks	O
played	O
an	O
important	O
role	O
in	O
the	O
United	O
States	O
2016	O
Presidential	O
Election	O
.	O
</s>
<s>
Researchers	O
estimated	O
that	O
pro-Trump	O
bots	B-Protocol
generated	O
four	O
tweets	O
for	O
every	O
pro-Clinton	O
automated	O
account	O
and	O
out-tweeted	O
pro-Clinton	O
bots	B-Protocol
7:1	O
on	O
relevant	O
hashtags	O
during	O
the	O
final	O
debate	O
.	O
</s>
<s>
Deceiving	O
Twitter	B-Protocol
bots	I-Protocol
fooled	O
candidates	O
and	O
campaign	O
staffers	O
into	O
retweeting	O
misappropriated	O
quotes	O
and	O
accounts	O
affiliated	O
with	O
incendiary	O
ideals	O
.	O
</s>
<s>
Twitter	B-Protocol
bots	I-Protocol
have	O
also	O
been	O
documented	O
to	O
influence	O
online	O
politics	O
in	O
Venezuela	O
.	O
</s>
<s>
In	O
2019	O
,	O
20%	O
of	O
the	O
global	O
Twitter	B-Application
trends	O
were	O
found	O
to	O
be	O
created	O
automatically	O
using	O
bots	B-Protocol
originating	O
from	O
Turkey	O
.	O
</s>
<s>
It	O
is	O
reported	O
that	O
108,000	O
bot	B-Protocol
accounts	I-Protocol
were	O
bulk	O
tweeting	O
to	O
push	O
19,000	O
keywords	O
to	O
top	O
trends	O
in	O
Turkey	O
,	O
to	O
promote	O
slogans	O
such	O
as	O
political	O
campaigns	O
related	O
to	O
the	O
2019	O
Turkish	O
local	O
elections	O
.	O
</s>
<s>
In	O
November	O
2022	O
,	O
Chinese	O
bots	B-Protocol
coordinately	O
flood	O
Twitter	B-Application
with	O
garbage	O
information	O
(	O
e.g.	O
</s>
<s>
These	O
bots	B-Protocol
,	O
disguised	O
as	O
attractive	O
girls	O
,	O
hashtag	O
the	O
major	O
cities	O
in	O
China	O
.	O
</s>
<s>
The	O
majority	O
of	O
Twitter	B-Application
accounts	I-Application
following	O
public	O
figures	O
and	O
brands	O
are	O
often	O
fake	O
or	O
inactive	O
,	O
making	O
the	O
number	O
of	O
Twitter	B-Application
followers	O
a	O
celebrity	O
has	O
a	O
difficult	O
metric	O
for	O
gauging	O
popularity	O
.	O
</s>
<s>
While	O
this	O
cannot	O
always	O
be	O
helped	O
,	O
some	O
public	O
figures	O
who	O
have	O
gained	O
or	O
lost	O
huge	O
quantities	O
of	O
followers	O
in	O
short	O
periods	O
of	O
time	O
have	O
been	O
accused	O
of	O
discreetly	O
paying	O
for	O
Twitter	B-Application
followers	O
.	O
</s>
<s>
For	O
example	O
,	O
the	O
Twitter	B-Application
accounts	I-Application
of	O
Sean	O
Combs	O
,	O
Rep	O
Jared	O
Polis	O
(	O
D-Colo	O
)	O
,	O
PepsiCo	O
,	O
Mercedes-Benz	O
,	O
and	O
50	O
Cent	O
have	O
come	O
under	O
scrutiny	O
for	O
possibly	O
engaging	O
in	O
the	O
buying	O
and	O
selling	O
of	O
Twitter	B-Application
followers	O
,	O
which	O
is	O
estimated	O
to	O
be	O
between	O
a	O
$40	O
million	O
and	O
$360	O
million	O
business	O
annually	O
.	O
</s>
<s>
Account	O
sellers	O
may	O
charge	O
a	O
premium	O
for	O
more	O
realistic	O
accounts	O
that	O
have	O
Twitter	B-Application
profile	O
pictures	O
and	O
bios	O
and	O
retweet	O
the	O
accounts	O
they	O
follow	O
.	O
</s>
<s>
In	O
addition	O
to	O
an	O
ego	O
boost	O
,	O
public	O
figures	O
may	O
gain	O
more	O
lucrative	O
endorsement	O
contracts	O
from	O
inflated	O
Twitter	B-Application
metrics	O
.	O
</s>
<s>
It	O
is	O
sometimes	O
desirable	O
to	O
identify	O
when	O
a	O
Twitter	B-Application
account	I-Application
is	O
controlled	O
by	O
a	O
internet	B-Protocol
bot	I-Protocol
.	O
</s>
<s>
Following	O
a	O
test	O
period	O
,	O
Twitter	B-Application
rolled	O
out	O
labels	O
to	O
identify	O
bot	B-Protocol
accounts	I-Protocol
and	O
automated	O
tweets	O
in	O
February	O
2022	O
.	O
</s>
<s>
Detecting	O
non-human	O
Twitter	B-Application
users	O
has	O
been	O
of	O
interest	O
to	O
academics	O
.	O
</s>
<s>
propose	O
the	O
following	O
criteria	O
that	O
indicate	O
that	O
an	O
account	O
may	O
be	O
a	O
bot	B-Protocol
(	O
they	O
were	O
designing	O
an	O
automated	O
system	O
)	O
:	O
</s>
<s>
The	O
ratio	O
of	O
tweets	O
from	O
mobile	O
versus	O
desktop	O
,	O
as	O
compared	O
to	O
an	O
average	O
human	O
Twitter	B-Application
user	O
.	O
</s>
<s>
Emilio	O
Ferrara	O
at	O
the	O
University	O
of	O
Southern	O
California	O
used	O
artificial	O
intelligence	O
to	O
identify	O
Twitter	B-Protocol
bots	I-Protocol
.	O
</s>
<s>
He	O
found	O
that	O
humans	O
reply	O
to	O
other	O
tweets	O
four	O
or	O
five	O
times	O
more	O
than	O
bots	B-Protocol
and	O
that	O
bots	B-Protocol
continue	O
to	O
post	O
longer	O
tweets	O
over	O
time	O
.	O
</s>
<s>
Bots	B-Protocol
also	O
post	O
at	O
more	O
regular	O
time	O
gaps	O
,	O
for	O
example	O
,	O
tweeting	O
at	O
30-minute	O
or	O
60-minute	O
intervals	O
.	O
</s>
<s>
Indiana	O
University	O
has	O
developed	O
a	O
free	O
service	O
called	O
Botometer	O
(	O
formerly	O
BotOrNot	O
)	O
,	O
which	O
scores	O
Twitter	B-Application
handles	I-Application
based	O
on	O
their	O
likelihood	O
of	O
being	O
a	O
Twitterbot	B-Protocol
.	O
</s>
<s>
Recent	O
research	O
from	O
EPFL	O
argued	O
that	O
classifying	O
a	O
Twitter	B-Application
account	I-Application
as	O
bot	B-Protocol
or	O
not	O
may	O
not	O
be	O
always	O
possible	O
because	O
hackers	O
take	O
over	O
human	O
accounts	O
and	O
use	O
them	O
as	O
bots	B-Protocol
temporarily	O
or	O
permanently	O
and	O
in	O
parallel	O
to	O
the	O
owner	O
of	O
the	O
account	O
in	O
some	O
cases	O
.	O
</s>
<s>
There	O
are	O
many	O
different	O
types	O
of	O
Twitter	B-Protocol
bots	I-Protocol
and	O
their	O
purposes	O
vary	O
from	O
one	O
to	O
another	O
.	O
</s>
<s>
to	O
anyone	O
on	O
Twitter	B-Application
that	O
used	O
the	O
internet	O
meme	O
phrase	O
"	O
over	O
9000.	O
"	O
</s>
<s>
@DearAssistant	O
sends	O
auto-reply	O
tweets	O
responding	O
to	O
complex	O
queries	O
in	O
simple	O
English	O
by	O
utilizing	O
Wolfram	B-Application
Alpha	I-Application
.	O
</s>
<s>
@DeepDrumpf	O
is	O
a	O
recurrent	B-Algorithm
neural	I-Algorithm
network	I-Algorithm
,	O
created	O
at	O
MIT	O
,	O
that	O
releases	O
tweets	O
imitating	O
Donald	O
Trump	O
's	O
speech	O
patterns	O
.	O
</s>
<s>
Try	O
saying	O
'	O
undocumented	O
immigrant	O
 '	O
or	O
'	O
unauthorized	O
immigrant	O
 '	O
instead	O
"	O
to	O
Twitter	B-Application
users	O
who	O
have	O
sent	O
a	O
tweet	O
containing	O
the	O
phrase	O
"	O
illegal	O
immigrant	O
"	O
.	O
</s>
<s>
@fuckeveryword	O
was	O
tweeting	O
every	O
word	O
in	O
the	O
English	O
language	O
preceded	O
by	O
"	O
fuck	O
"	O
,	O
but	O
Twitter	B-Application
suspended	O
it	O
midway	O
through	O
operation	O
because	O
the	O
account	O
tweeted	O
"	O
fuck	O
niggers	O
"	O
.	O
</s>
<s>
@Horse	O
ebooks	O
was	O
a	O
bot	B-Protocol
that	O
gained	O
a	O
following	O
among	O
people	O
who	O
found	O
its	O
tweets	O
poetic	O
.	O
</s>
<s>
It	O
has	O
inspired	O
various	O
_ebooks-suffixed	O
Twitter	B-Protocol
bots	I-Protocol
which	O
use	O
Markov	O
text	B-General_Concept
generators	I-General_Concept
(	O
or	O
similar	B-General_Concept
techniques	I-General_Concept
)	O
to	O
create	O
new	O
tweets	O
by	O
mashing	O
up	O
the	O
tweets	O
of	O
their	O
owner	O
.	O
</s>
<s>
@MetaphorMagnet	O
is	O
an	O
AI	O
bot	B-Protocol
that	O
generates	O
metaphorical	O
insights	O
using	O
its	O
knowledge-base	O
of	O
stereotypical	O
properties	O
and	O
norms	O
.	O
</s>
<s>
A	O
companion	O
bot	B-Protocol
@MetaphorMirror	O
pairs	O
these	O
metaphors	O
to	O
news	O
tweets	O
.	O
</s>
<s>
Another	O
companion	O
bot	B-Protocol
,	O
@BestOfBotWorlds	O
,	O
uses	O
metaphor	O
to	O
generate	O
faux-religious	O
insights	O
.	O
</s>
<s>
@Pentametron	O
finds	O
tweets	O
incidentally	O
written	O
in	O
iambic	O
pentameter	O
using	O
the	O
CMU	B-General_Concept
Pronouncing	I-General_Concept
Dictionary	I-General_Concept
,	O
pairs	O
them	O
into	O
couplets	O
using	O
a	O
rhyming	O
dictionary	O
,	O
and	O
retweets	O
them	O
as	O
couplets	O
into	O
followers	O
 '	O
feeds	O
.	O
</s>
<s>
@RedScareBot	O
tweets	O
in	O
the	O
persona	O
of	O
Joseph	O
McCarthy	O
in	O
response	O
to	O
Twitter	B-Application
posts	O
mentioning	O
"	O
socialist	O
"	O
,	O
"	O
communist	O
"	O
,	O
or	O
"	O
communism	O
"	O
.	O
</s>
<s>
@DisinfoNews	O
Disinformation	O
News	O
Aggregator	O
automatically	O
retweets	O
tweets	O
that	O
shares	O
news	O
articles	O
or	O
scientific	O
work	O
related	O
to	O
disinformation	O
,	O
bots	B-Protocol
or	O
trolls	O
from	O
experts	O
relevant	O
to	O
those	O
topics	O
.	O
</s>
<s>
In	O
2009	O
,	O
based	O
on	O
a	O
study	O
by	O
Sysomos	O
,	O
Twitter	B-Protocol
bots	I-Protocol
were	O
estimated	O
to	O
create	O
approximately	O
24%	O
of	O
tweets	O
on	O
Twitter	B-Application
.	O
</s>
<s>
According	O
to	O
the	O
company	O
,	O
there	O
were	O
20million	O
,	O
fewer	O
than	O
5%	O
,	O
of	O
accounts	O
on	O
Twitter	B-Application
that	O
were	O
fraudulent	O
in	O
2013	O
.	O
</s>
<s>
In	O
2013	O
,	O
two	O
Italian	O
researchers	O
calculated	O
10	O
percent	O
of	O
total	O
accounts	O
on	O
Twitter	B-Application
were	O
"	O
bots	B-Protocol
"	O
although	O
other	O
estimates	O
have	O
placed	O
the	O
figure	O
even	O
higher	O
.	O
</s>
<s>
One	O
significant	O
academic	O
study	O
in	O
2017	O
estimated	O
that	O
up	O
to	O
15%	O
of	O
Twitter	B-Application
users	O
were	O
automated	O
bot	B-Protocol
accounts	I-Protocol
.	O
</s>
<s>
The	O
prevalence	O
of	O
Twitter	B-Protocol
bots	I-Protocol
coupled	O
with	O
the	O
ability	O
of	O
some	O
bots	B-Protocol
to	O
give	O
seemingly	O
human	O
responses	O
has	O
enabled	O
these	O
non-human	O
accounts	O
to	O
garner	O
widespread	O
influence	O
.	O
</s>
<s>
The	O
social	O
implications	O
these	O
Twitter	B-Protocol
bots	I-Protocol
potentially	O
have	O
on	O
human	O
perception	O
are	O
sizeable	O
according	O
to	O
a	O
study	O
published	O
by	O
the	O
ScienceDirect	O
Journal	O
.	O
</s>
<s>
The	O
study	O
concluded	O
that	O
Twitter	B-Protocol
bots	I-Protocol
were	O
viewed	O
as	O
credible	O
and	O
competent	O
in	O
communication	O
and	O
interaction	O
making	O
them	O
suitable	O
for	O
transmitting	O
information	O
in	O
the	O
social	O
media	O
sphere	O
.	O
</s>
