BACKGROUND
Eating
disorders
are
a
global
concern
and
access
to
early
treatment
is
critical
improve
prognosis.
Single
session
interventions
have
been
proposed
as
an
opportunity
provide
short
term
support
people
on
waitlists
for
eating
disorder
treatment,
however,
there
not
enough
clinicians
this
intervention.
Conversational
artificial
intelligence
agents
or
“chatbots”
reflect
unique
fill
gap
in
service
provision.
OBJECTIVE
To
co-design
novel
chatbot
capable
of
delivering
single
intervention
the
waitlist
across
diagnostic
spectrum
ascertain
its
preliminary
acceptability
feasibility.
METHODS
The
study
followed
design
process
Double
Diamond
model
including
four
phases:
discover,
define,
develop,
deliver.
involved
with
17
participants
total;
10
adults
lived
experience
7
clinicians,
by
conducting
interviews
workshops
participants.
Feedback
from
each
phase
informed
ideas
development
next
study.
A
final
prototype
was
presented
deliver
phase.
RESULTS
Qualitative
thematic
analysis
identified
main
themes
that
were
present
rounds
interviews/workshops:
conversational
tone,
safety
risk
management,
user
journey
structure,
content.
CONCLUSIONS
Overall,
feedback
positive
throughout
both
clinicians.
Incorporating
phases
allowed
refinement
chatbot.
Further
research
required
evaluate
chatbot’s
efficacy
settings.
CLINICALTRIAL
N/A
npj Mental Health Research,
Год журнала:
2024,
Номер
3(1)
Опубликована: Апрель 2, 2024
Abstract
Large
language
models
(LLMs)
such
as
Open
AI’s
GPT-4
(which
power
ChatGPT)
and
Google’s
Gemini,
built
on
artificial
intelligence,
hold
immense
potential
to
support,
augment,
or
even
eventually
automate
psychotherapy.
Enthusiasm
about
applications
is
mounting
in
the
field
well
industry.
These
developments
promise
address
insufficient
mental
healthcare
system
capacity
scale
individual
access
personalized
treatments.
However,
clinical
psychology
an
uncommonly
high
stakes
application
domain
for
AI
systems,
responsible
evidence-based
therapy
requires
nuanced
expertise.
This
paper
provides
a
roadmap
ambitious
yet
of
LLMs
First,
technical
overview
presented.
Second,
stages
integration
into
psychotherapy
are
discussed
while
highlighting
parallels
development
autonomous
vehicle
technology.
Third,
care,
training,
research
discussed,
areas
risk
given
complex
nature
Fourth,
recommendations
evaluation
provided,
which
include
centering
science,
involving
robust
interdisciplinary
collaboration,
attending
issues
like
assessment,
detection,
transparency,
bias.
Lastly,
vision
outlined
how
might
enable
new
generation
studies
interventions
at
scale,
these
may
challenge
assumptions
Journal of Eating Disorders,
Год журнала:
2025,
Номер
13(1)
Опубликована: Март 11, 2025
Abstract
Background
Early
treatment
is
critical
to
improve
eating
disorder
prognosis.
Single
session
interventions
have
been
proposed
as
a
strategy
provide
short
term
support
people
on
waitlists
for
treatment,
however,
it
not
always
possible
access
this
early
intervention.
Conversational
artificial
intelligence
agents
or
“chatbots”
reflect
unique
opportunity
attempt
fill
gap
in
service
provision.
The
aim
of
research
was
co-design
novel
chatbot
capable
delivering
single
intervention
adults
the
waitlist
across
diagnostic
spectrum
and
ascertain
its
preliminary
acceptability
feasibility.
Methods
A
Double
Diamond
approach
employed
which
included
four
phases:
discover,
define,
develop,
deliver.
There
were
17
participants
total
Australia;
ten
with
lived
experience
an
seven
registered
psychologists
working
field
disorders,
who
participated
online
interviews
workshops.
Thematic
content
analyses
undertaken
interview/workshop
transcriptions
findings
from
previous
phase
informing
ideas
development
next
phase.
final
prototype
presented
deliver
Results
identified
main
themes
that
present
phases
interviews/workshops:
conversational
tone,
safety
risk
management,
user
journey
structure,
content.
Conclusions
Overall,
feedback
positive
throughout
process
both
psychologists.
Incorporating
allowed
refinement
chatbot.
Further
required
evaluate
chatbot’s
efficacy
settings.
Journal of Eating Disorders,
Год журнала:
2022,
Номер
10(1)
Опубликована: Май 8, 2022
Abstract
Advances
in
machine
learning
and
digital
data
provide
vast
potential
for
mental
health
predictions.
However,
research
using
the
field
of
eating
disorders
is
just
beginning
to
emerge.
This
paper
provides
a
narrative
review
existing
explores
benefits,
limitations,
ethical
considerations
aid
detection,
prevention,
treatment
disorders.
Current
primarily
uses
predict
disorder
status
from
females’
responses
validated
surveys,
social
media
posts,
or
neuroimaging
often
with
relatively
high
levels
accuracy.
early
work
evidence
improve
current
screening
methods.
ability
these
algorithms
generalise
other
samples
be
used
on
mass
scale
only
explored.
One
key
benefit
over
traditional
statistical
methods
simultaneously
examine
large
numbers
(100s
1000s)
multimodal
predictors
their
complex
non-linear
interactions,
but
few
studies
have
explored
this
Machine
also
being
develop
chatbots
psychoeducation
coping
skills
training
around
body
image
disorders,
implications
intervention.
The
use
personalise
options,
ecological
momentary
interventions,
clinicians
discussed.
accurate,
rapid,
cost-effective
More
needed
diverse
participants
ensure
that
models
are
unbiased,
generalisable
all
people
There
important
limitations
utilising
practice.
Thus,
rather
than
magical
solution,
should
seen
as
an
tool
researchers,
eventually
clinicians,
identification,
International Journal of Eating Disorders,
Год журнала:
2022,
Номер
55(9), С. 1229 - 1244
Опубликована: Авг. 18, 2022
Abstract
Objective
A
significant
gap
exists
between
those
who
need
and
receive
care
for
eating
disorders
(EDs).
Novel
solutions
are
needed
to
encourage
service
use
address
treatment
barriers.
This
study
developed
evaluated
the
usability
of
a
chatbot
designed
pairing
with
online
ED
screening.
The
tool
aimed
promote
mental
health
utilization
by
improving
motivation
self‐efficacy
among
individuals
EDs.
Methods
prototype,
Alex,
was
using
decision
trees
theoretically‐informed
components:
psychoeducation,
motivational
interviewing,
personalized
recommendations,
repeated
administration.
Usability
testing
conducted
over
four
iterative
cycles,
user
feedback
informing
refinements
next
iteration.
Post‐testing,
participants
(N=
21)
completed
System
Scale
(SUS),
Usefulness,
Satisfaction,
Ease
Use
Questionnaire
(USE),
semi‐structured
interview.
Results
Interview
detailed
aspects
enjoyed
necessitating
improvement.
Feedback
converged
on
themes:
experience,
qualities,
content,
ease
use.
Following
refinements,
users
described
Alex
as
humanlike,
supportive,
encouraging.
Content
perceived
novel
personally
relevant.
USE
scores
across
domains
were
generally
above
average
(~5
out
7),
SUS
indicated
“good”
“excellent”
final
iteration
receiving
highest
score.
Discussion
Overall,
reflected
positively
interactions
including
initial
version.
Refinements
cycles
further
improved
experiences.
provides
preliminary
evidence
feasibility
acceptance
services
Public
Significance
Low
rates
have
been
observed
following
disorder
Tools
needed,
scalable,
digital
options,
that
can
be
easily
paired
screening,
improve
addressing
utilization.
JMIR Mental Health,
Год журнала:
2024,
Номер
11, С. e60589 - e60589
Опубликована: Окт. 11, 2024
Abstract
Background
Artificial
intelligence
(AI)
has
been
increasingly
recognized
as
a
potential
solution
to
address
mental
health
service
challenges
by
automating
tasks
and
providing
new
forms
of
support.
Objective
This
study
is
the
first
in
series
which
aims
estimate
current
rates
AI
technology
use
well
perceived
benefits,
harms,
risks
experienced
community
members
(CMs)
professionals
(MHPs).
Methods
involved
2
web-based
surveys
conducted
Australia.
The
collected
data
on
demographics,
comfort,
attitudes
toward
AI,
specific
cases,
experiences
benefits
harms
from
use.
Descriptive
statistics
were
calculated,
thematic
analysis
open-ended
responses
conducted.
Results
final
sample
consisted
107
CMs
86
MHPs.
General
varied,
with
reporting
neutral
MHPs
more
positive
attitudes.
Regarding
usage,
28%
(30/108)
used
primarily
for
quick
support
(18/30,
60%)
personal
therapist
(14/30,
47%).
Among
MHPs,
43%
(37/86)
AI;
mostly
research
(24/37,
65%)
report
writing
(20/37,
54%).
While
majority
found
be
generally
beneficial
(23/30,
77%
34/37,
92%
MHPs),
concerns
47%
(14/30)
51%
(19/37)
There
was
an
equal
mix
negative
sentiment
future
care
open
feedback.
Conclusions
Commercial
tools
are
being
Respondents
believe
will
offer
advantages
terms
accessibility,
cost
reduction,
personalization,
work
efficiency.
However,
they
equally
concerned
about
reducing
human
connection,
ethics,
privacy
regulation,
medical
errors,
misuse,
security.
Despite
immense
potential,
integration
into
systems
must
approached
caution,
addressing
legal
ethical
while
developing
safeguards
mitigate
harms.
Future
planned
track
acceptability
associated
issues
over
time.
Large
language
models
(LLMs)
such
as
Open
AI’s
GPT-3
and
-4
(which
power
ChatGPT)
Google’s
PaLM,
built
on
artificial
intelligence,
hold
immense
potential
to
support,
augment,
or
even
eventually
fully
automate
psychotherapy.
Enthusiasm
about
applications
is
mounting
in
the
field
well
industry.
These
developments
promise
address
insufficient
mental
healthcare
system
capacity
scale
individual
access
personalized
treatments.
However,
clinical
psychology
an
uncommonly
high
stakes
application
domain
for
AI
systems,
responsible
evidence-based
therapy
requires
nuanced
expertise.
This
paper
provides
a
roadmap
ambitious
yet
of
LLMs
First,
technical
overview
presented.
Second,
stages
integration
into
psychotherapy
are
discussed
while
highlighting
parallels
development
autonomous
vehicle
technology.
Third,
care,
training,
research
discussed,
areas
risk
given
complex
nature
Fourth,
recommendations
evaluation
provided,
which
include
centering
science,
involving
robust
interdisciplinary
collaboration,
attending
issues
like
assessment,
detection,
transparency,
bias.
Lastly,
vision
outlined
how
might
enable
new
generation
studies
interventions
at
scale,
these
may
challenge
assumptions
Journal of Medical Internet Research,
Год журнала:
2023,
Номер
25, С. e50696 - e50696
Опубликована: Авг. 14, 2023
The
use
of
artificial
intelligence
(AI)
to
assist
with
the
prevention,
identification,
and
management
eating
disorders
body
image
concerns
is
exciting,
but
it
not
without
risk.
Technology
advancing
rapidly,
ensuring
that
responsible
standards
are
in
place
mitigate
risk
protect
users
vital
success
safety
technologies
users.
Interactive Journal of Medical Research,
Год журнала:
2024,
Номер
13, С. e53672 - e53672
Опубликована: Авг. 12, 2024
Background
Mental
disorders
have
ranked
among
the
top
10
prevalent
causes
of
burden
on
a
global
scale.
Generative
artificial
intelligence
(GAI)
has
emerged
as
promising
and
innovative
technological
advancement
that
significant
potential
in
field
mental
health
care.
Nevertheless,
there
is
scarcity
research
dedicated
to
examining
understanding
application
landscape
GAI
within
this
domain.
Objective
This
review
aims
inform
current
state
knowledge
identify
its
key
uses
domain
by
consolidating
relevant
literature.
Methods
Records
were
searched
8
reputable
sources
including
Web
Science,
PubMed,
IEEE
Xplore,
medRxiv,
bioRxiv,
Google
Scholar,
CNKI
Wanfang
databases
between
2013
2023.
Our
focus
was
original,
empirical
with
either
English
or
Chinese
publications
use
technologies
benefit
health.
For
an
exhaustive
search,
we
also
checked
studies
cited
Two
reviewers
responsible
for
data
selection
process,
all
extracted
synthesized
summarized
brief
in-depth
analyses
depending
approaches
used
(traditional
retrieval
rule-based
techniques
vs
advanced
techniques).
Results
In
144
articles,
44
(30.6%)
met
inclusion
criteria
detailed
analysis.
Six
emerged:
disorder
detection,
counseling
support,
therapeutic
application,
clinical
training,
decision-making
goal-driven
optimization.
Advanced
systems
been
mainly
focused
applications
(n=19,
43%)
support
(n=13,
30%),
training
being
least
common.
Most
(n=28,
64%)
broadly
health,
while
specific
conditions
such
anxiety
(n=1,
2%),
bipolar
(n=2,
5%),
eating
posttraumatic
stress
schizophrenia
2%)
received
limited
attention.
Despite
use,
efficacy
ChatGPT
detection
remains
insufficient.
addition,
100
articles
traditional
found,
indicating
diverse
areas
where
could
enhance
Conclusions
study
provides
comprehensive
overview
care,
which
serves
valuable
guide
future
research,
practical
applications,
policy
development
While
demonstrates
promise
augmenting
care
services,
inherent
limitations
emphasize
role
supplementary
tool
rather
than
replacement
trained
providers.
A
conscientious
ethical
integration
necessary,
ensuring
balanced
approach
maximizes
benefits
mitigating
challenges
practices.
JMIR Mental Health,
Год журнала:
2025,
Номер
12, С. e69294 - e69294
Опубликована: Янв. 17, 2025
Abstract
Background
Mental
health
disorders
significantly
impact
global
populations,
prompting
the
rise
of
digital
mental
interventions,
such
as
artificial
intelligence
(AI)-powered
chatbots,
to
address
gaps
in
access
care.
This
review
explores
potential
for
a
“digital
therapeutic
alliance
(DTA),”
emphasizing
empathy,
engagement,
and
alignment
with
traditional
principles
enhance
user
outcomes.
Objective
The
primary
objective
this
was
identify
key
concepts
underlying
DTA
AI-driven
psychotherapeutic
interventions
health.
secondary
propose
an
initial
definition
based
on
these
identified
concepts.
Methods
PRISMA
(Preferred
Reporting
Items
Systematic
Reviews
Meta-Analyses)
scoping
reviews
Tavares
de
Souza’s
integrative
methodology
were
followed,
encompassing
systematic
literature
searches
Medline,
Web
Science,
PsycNet,
Google
Scholar.
Data
from
eligible
studies
extracted
analyzed
using
Horvath
et
al’s
conceptual
framework
alliance,
focusing
goal
alignment,
task
agreement,
bond,
quality
assessed
Newcastle-Ottawa
Scale
Cochrane
Risk
Bias
Tool.
Results
A
total
28
pool
1294
articles
after
excluding
duplicates
ineligible
studies.
These
informed
development
DTA,
elements
facilitators
barriers
affecting
primarily
focused
AI-powered
psychotherapy,
other
tools.
Conclusions
findings
provide
foundational
concept
report
its
replicate
mechanisms
trust,
collaboration
While
shows
promise
enhancing
accessibility
engagement
care,
further
research
innovation
are
needed
challenges
personalization,
ethical
concerns,
long-term
impact.
