<s>
Human-AI	O
collaboration	O
is	O
the	O
study	O
of	O
how	O
humans	O
and	O
artificial	B-Application
intelligence	I-Application
(	O
AI	B-Application
)	O
agents	O
work	O
together	O
to	O
accomplish	O
a	O
shared	O
goal	O
.	O
</s>
<s>
AI	B-Application
systems	O
can	O
aid	O
humans	O
in	O
everything	O
from	O
decision	O
making	O
tasks	O
to	O
art	O
creation	O
.	O
</s>
<s>
Examples	O
of	O
collaboration	O
include	O
medical	B-Application
decision	O
making	O
aids.	O
,	O
hate	O
speech	O
detection	O
,	O
and	O
music	O
generation	O
.	O
</s>
<s>
As	O
AI	B-Application
systems	O
are	O
able	O
to	O
tackle	O
more	O
complex	O
tasks	O
,	O
studies	O
are	O
exploring	O
how	O
different	O
models	O
and	O
explanation	O
techniques	O
can	O
improve	O
human-AI	O
collaboration	O
.	O
</s>
<s>
When	O
a	O
human	O
uses	O
an	O
AI	B-Application
's	O
output	O
,	O
they	O
often	O
want	O
to	O
understand	O
why	O
a	O
model	O
gave	O
a	O
certain	O
output	O
.	O
</s>
<s>
Various	O
Explainable	O
artificial	B-Application
intelligence	I-Application
methods	O
aim	O
to	O
describe	O
model	O
outputs	O
with	O
post-hoc	O
explanations	O
or	O
visualizations	O
,	O
these	O
methods	O
can	O
often	O
provide	O
misleading	O
and	O
false	O
explanations	O
.	O
</s>
<s>
Studies	O
have	O
also	O
found	O
that	O
explanations	O
may	O
not	O
improve	O
the	O
performance	O
of	O
a	O
human-AI	O
team	O
,	O
but	O
simply	O
increase	O
a	O
human	O
's	O
reliance	O
on	O
the	O
model	O
's	O
output	O
.	O
</s>
<s>
A	O
human	O
's	O
trust	O
in	O
an	O
AI	B-Application
agent	O
is	O
an	O
important	O
factor	O
in	O
human-AI	O
collaboration	O
,	O
dictating	O
whether	O
the	O
human	O
should	O
follow	O
or	O
override	O
the	O
AI	B-Application
's	O
input	O
.	O
</s>
