Does
the
notion
of
intrinsic
reward
explain
context-dependent
valuation
in
reinforcement
learning?
A
comment
on
Molinaro
&
Collins,
2023
-
Intrinsic
rewards
context-sensitive
learning;
https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002201
Trends in Cognitive Sciences,
Год журнала:
2024,
Номер
28(5), С. 441 - 453
Опубликована: Фев. 26, 2024
What
drives
our
curiosity
remains
an
elusive
and
hotly
debated
issue,
with
multiple
hypotheses
proposed
but
a
cohesive
account
yet
to
be
established.
This
review
discusses
traditional
emergent
theories
that
frame
as
desire
know
drive
learn,
respectively.
We
adopt
model-based
approach
maps
the
temporal
dynamics
of
various
factors
underlying
curiosity-based
exploration,
such
uncertainty,
information
gain,
learning
progress.
In
so
doing,
we
identify
limitations
past
posit
integrated
harnesses
their
strengths
in
describing
tool
for
optimal
environmental
exploration.
unified
account,
serves
'common
currency'
which
must
balanced
other
safety
hunger
achieve
efficient
action.
Nature Human Behaviour,
Год журнала:
2024,
Номер
unknown
Опубликована: Апрель 17, 2024
When
striking
a
balance
between
commitment
to
goal
and
flexibility
in
the
face
of
better
options,
people
often
demonstrate
strong
perseveration.
Here,
using
functional
MRI
(n
=
30)
lesion
patient
26)
studies,
we
argue
that
ventromedial
prefrontal
cortex
(vmPFC)
drives
linked
changes
goal-directed
selective
attention.
Participants
performed
an
incremental
pursuit
task
involving
sequential
decisions
persisting
with
versus
abandoning
progress
for
alternative
options.
Individuals
stronger
perseveration
showed
higher
attention
interleaved
task.
Increasing
also
affected
abandonment
decisions:
while
pursuing
goal,
lost
their
sensitivity
valuable
goals
remaining
more
sensitive
current
goal.
In
healthy
population,
individual
differences
both
biases
goal-oriented
were
predicted
by
baseline
goal-related
activity
vmPFC.
Among
patients,
vmPFC
damage
reduced
commitment,
leading
performance
benefit.
Motor
learning
is
often
viewed
as
a
unitary
process
that
operates
outside
of
conscious
awareness.
This
perspective
has
led
to
the
development
sophisticated
models
designed
elucidate
mechanisms
implicit
sensorimotor
learning.
In
this
review,
we
argue
for
broader
perspective,
emphasizing
contribution
explicit
strategies
tasks.
Furthermore,
propose
theoretical
framework
motor
consists
three
fundamental
processes:
reasoning,
understanding
action–outcome
relationships;
refinement,
optimizing
and
cognitive
parameters
achieve
goals;
retrieval,
inferring
context
recalling
control
policy.
We
anticipate
‘3R’
how
complex
movements
are
learned
will
open
exciting
avenues
future
research
at
intersection
between
cognition
action.
[Now
published
in
eLife:
https://elifesciences.org/articles/91839]
Motor
learning
is
often
viewed
as
a
unitary
process
that
operates
outside
of
conscious
awareness.
This
perspective
has
led
to
the
development
sophisticated
models
designed
elucidate
mechanisms
implicit
sensorimotor
learning.
In
this
review
we
argue
for
broader
perspective,
emphasizing
contribution
explicit
strategies
tasks.
Furthermore,
propose
theoretical
framework
motor
consists
three
fundamental
processes:
Reasoning,
understanding
action-outcome
relationships;
Refinement,
optimizing
and
cognitive
parameters
achieve
goals;
Retrieval,
inferring
context
recalling
control
policy.
We
anticipate
“3R”
how
complex
movements
are
learned
will
open
exciting
avenues
future
research
at
intersection
between
cognition
action.
Brain,
Год журнала:
2024,
Номер
147(6), С. 2230 - 2244
Опубликована: Апрель 8, 2024
Despite
a
theory
that
an
imbalance
in
goal-directed
versus
habitual
systems
serve
as
building
blocks
of
compulsions,
research
has
yet
to
delineate
how
this
occurs
during
arbitration
between
the
two
obsessive-compulsive
disorder.
Inspired
by
brain
model
which
inferior
frontal
cortex
selectively
gates
putamen
guide
or
actions,
study
aimed
examine
whether
disruptions
process
via
fronto-striatal
circuit
would
underlie
imbalanced
decision-making
and
compulsions
patients.
Thirty
patients
with
disorder
[mean
(standard
deviation)
age
=
26.93
(6.23)
years,
12
females
(40%)]
30
healthy
controls
24.97
(4.72)
17
(57%)]
underwent
functional
MRI
scans
while
performing
two-step
Markov
decision
task,
was
designed
dissociate
behaviour
from
behaviour.
We
employed
neurocomputational
account
for
uncertainty-based
process,
prefrontal
arbitrator
(i.e.
gyrus)
allocates
behavioural
control
more
reliable
strategy
gating
putamen.
analysed
group
differences
neural
estimates
uncertainty
each
strategy.
also
compared
psychophysiological
interaction
effects
system
preference
(goal-directed
habitual)
on
coupling
groups.
examined
correlation
compulsivity
score
activity
connectivity
involved
process.
The
computational
captured
subjects'
preferences
strategies.
Compared
controls,
had
stronger
(t
-2.88,
P
0.006),
attributed
uncertain
2.72,
0.009).
Before
allocation
exhibited
hypoactivity
gyrus
when
region
tracked
inverse
reliability)
(P
0.001,
family-wise
error
rate
corrected).
When
reorienting
behaviours
reach
specific
goals,
weaker
right
ipsilateral
ventrolateral
prefronto-putamen
than
This
hypoconnectivity
correlated
severe
(r
-0.57,
0.002).
Our
findings
suggest
attenuated
top-down
underlies
Enhancing
may
be
potential
neurotherapeutic
approach
adaptive
decision-making.