PLoS Computational Biology,
Год журнала:
2024,
Номер
20(11), С. e1012531 - e1012531
Опубликована: Ноя. 4, 2024
Synapses
in
the
brain
are
highly
noisy,
which
leads
to
a
large
trial-by-trial
variability.
Given
how
costly
synapses
terms
of
energy
consumption
these
high
levels
noise
surprising.
Here
we
propose
that
use
represent
uncertainties
about
somatic
activity
postsynaptic
neuron.
To
show
this,
developed
mathematical
framework,
synapse
as
whole
interacts
with
soma
neuron
similar
way
an
agent
is
situated
and
behaves
uncertain,
dynamic
environment.
This
framework
suggests
implicit
internal
model
membrane
dynamics
being
updated
by
synaptic
learning
rule,
resembles
experimentally
well-established
LTP/LTD
mechanisms.
In
addition,
this
approach
entails
utilizes
its
inherently
noisy
release
also
encode
uncertainty
state
potential.
Although
each
strives
for
predicting
neuron,
emergent
many
neuronal
network
resolve
different
problems
such
pattern
classification
or
closed-loop
control
Hereby,
coordinate
themselves
utilize
on
level
behaviorally
ambiguous
situations.
Trends in Neurosciences,
Год журнала:
2023,
Номер
46(12), С. 1054 - 1066
Опубликована: Ноя. 2, 2023
Curiosity
refers
to
the
intrinsic
desire
of
humans
and
animals
explore
unknown,
even
when
there
is
no
apparent
reason
do
so.
Thus
far,
single,
widely
accepted
definition
or
framework
for
curiosity
has
emerged,
but
growing
consensus
that
curious
behavior
not
goal-directed
related
seeking
reacting
information.
In
this
review,
we
take
a
phenomenological
approach
group
behavioral
neurophysiological
studies
which
meet
these
criteria
into
three
categories
according
type
information
observed.
We
then
review
recent
computational
models
from
field
machine
learning
discuss
how
they
enable
integrating
different
types
one
theoretical
framework.
Combinations
along
with
modeling
will
be
instrumental
in
demystifying
notion
curiosity.
Theory & Psychology,
Год журнала:
2024,
Номер
unknown
Опубликована: Окт. 29, 2024
The
“standard
definition”
of
creativity
holds
that
a
creative
idea
is
one
novel
and
useful.
This
judgement
customarily
based
on
an
external
frame
reference
as
it
passed
by
people
who
are
receiving
the
(the
recipient).
internal
person
has
generated
creator)
usually
ignored.
I
make
two
cases
in
this
paper.
First,
employing
frames
assessing
products
been
erroneously
applied
to
understand
mind.
Second,
any
definition
needs
be
can
reasonably
whether
following
experience
or
product.
With
these
aims
mind,
propose
amendment
creativity:
both
satisfying.
Task-evoked
pupil
dilation
has
been
linked
to
many
cognitive
variables,
perhaps
most
notably
unexpected
events.
Zénon
(2019)
proposed
a
unifying
framework
stating
that
related
cognition
should
be
considered
from
an
information-theory
perspective.
In
the
current
study,
we
investigated
whether
pupil’s
response
decision
outcome
in
context
of
associative
learning
reflects
prediction
error
defined
formally
as
information
gain,
while
also
exploring
time
course
this
signal.
To
do
so,
adapted
simple
model
trial-by-trial
stimulus
probabilities
based
on
theory
previous
literature.
We
analyzed
two
data
sets
which
participants
performed
perceptual
decision-making
tasks
required
was
recorded.
Our
findings
consistently
showed
significant
proportion
variability
post-feedback
during
can
explained
by
formal
quantification
gain
shortly
after
feedback
presentation
both
task
contexts.
later
window,
relationship
between
information-theoretic
variables
and
differed
per
task.
For
first
time,
present
evidence
dilates
or
constricts
along
with
seems
dependent,
specifically
increasing
decreasing
average
uncertainty
(entropy)
across
trials.
This
study
offers
empirical
showcasing
how
offer
valuable
insights
into
process
updating
learning,
highlighting
promising
utility
readily
accessible
physiological
indicator
for
investigating
internal
belief
states.
Task-evoked
pupil
dilation
has
been
linked
to
many
cognitive
variables,
perhaps
most
notably
unexpected
events.
Zénon
(2019)
proposed
a
unifying
framework
stating
that
related
cognition
should
be
considered
from
an
information-theory
perspective.
In
the
current
study,
we
investigated
whether
pupil’s
response
decision
outcome
in
context
of
associative
learning
reflects
prediction
error
defined
formally
as
information
gain,
while
also
exploring
time
course
this
signal.
To
do
so,
adapted
simple
model
trial-by-trial
stimulus
probabilities
based
on
theory
previous
literature.
We
analyzed
two
data
sets
which
participants
performed
perceptual
decision-making
tasks
required
was
recorded.
Our
findings
consistently
showed
significant
proportion
variability
post-feedback
during
can
explained
by
formal
quantification
gain
shortly
after
feedback
presentation
both
task
contexts.
later
window,
relationship
between
information-theoretic
variables
and
differed
per
task.
For
first
time,
present
evidence
dilates
or
constricts
along
with
seems
dependent,
specifically
increasing
decreasing
average
uncertainty
(entropy)
across
trials.
This
study
offers
empirical
showcasing
how
offer
valuable
insights
into
process
updating
learning,
highlighting
promising
utility
readily
accessible
physiological
indicator
for
investigating
internal
belief
states.
Behavioral and Brain Sciences,
Год журнала:
2024,
Номер
47
Опубликована: Янв. 1, 2024
In
our
target
article,
we
proposed
that
curiosity
and
creativity
are
both
manifestations
of
the
same
novelty-seeking
process.
We
received
29
commentaries
from
diverse
disciplines
add
insights
to
initial
proposal.
These
ultimately
expanded
supplemented
model.
Here
draw
attention
five
central
practical
theoretical
issues
were
raised
by
commentators:
(1)
The
complex
construct
novelty
associated
concepts;
(2)
underlying
subsystems
possible
mechanisms;
(3)
different
pathways
subtypes
creativity;
(4)
"in
wild";
(5)
link(s)
between
curiosity.
'Why
are
we
curious?'
has
been
among
the
central
puzzles
of
neuroscience
and
psychology
in
past
decades.
Recent
'top-down'
theories
have
hypothesized
that
curiosity,
as
a
desire
for
some
intrinsically
generated
rewards
(e.g.,
novelty),
is
optimal
solution
survival
complex
environments
where
evolved.
To
formalize
test
this
hypothesis,
however,
it
necessary
to
understand
relationship
between
(i)
intrinsic
(as
drives
curiosity),
(ii)
optimality
conditions
objectives
(iii)
environment
structures.
Here,
demystify
through
systematic
simulation
study.
We
first
propose
an
algorithm
generating
capture
key
abstract
features
different
real-world
situations.
Then,
within
these
environments,
simulate
artificial
agents
seeking
six
representative
(novelty,
surprise,
information
gain,
empowerment,
MOP
SPIE)
evaluate
their
performance
regarding
three
potential
curiosity
(environment
exploration,
model
accuracy
uniform
state
visitation).
Our
results
show
comparative
each
reward
highly
dependent
on
structural
objective
under
consideration;
indicates
'optimality'
top-down
needs
precise
formulation
structure.
Nevertheless,
found
combination
novelty
gain
always
achieve
close-to-optimal
performance;
proposes
two
principal
axes
curiosity-driven
behavior.
These
results,
collectively,
pave
way
further
development
computational
models
design
theory-informed
experimental
paradigms.
Child Development,
Год журнала:
2024,
Номер
unknown
Опубликована: Сен. 2, 2024
Abstract
Humans
are
driven
by
an
intrinsic
motivation
to
learn,
but
the
developmental
origins
of
curiosity‐driven
exploration
remain
unclear.
We
investigated
computational
principles
guiding
4‐year‐old
children's
during
a
touchscreen
game
(
N
=
102,
F
49,
M
53,
primarily
white
and
middle‐class,
data
collected
in
Netherlands
from
2021–2023).
Children
guessed
location
characters
that
were
hiding
following
predictable
(yet
noisy)
patterns.
could
freely
switch
characters,
which
allowed
us
quantify
when
they
decided
explore
something
different
what
chose
explore.
Bayesian
modeling
their
responses
revealed
children
selected
activities
more
novel
offered
greater
learning
progress
(LP).
Moreover,
interest
making
LP
correlated
with
better
performance.
These
findings
highlight
importance
novelty
exploration.