ORIGINAL ARTICLE
Cue ‐ induced effects on decisio n ‐ making distinguish subjects
with gambling disorder from healthy controls
Alexander Genauck
1,2
| Milan Andrejevic
1,3
| Katharina Brehm
1
| Caroline Matthis
5,2
|
Andreas Heinz
1
| André Weinreich
4
| Norbert Kathmann
4
| Nina Romanczuk ‐ Seiferth
1
1
Department of Psyc hiatry and
Psychotherapy, Charit é – Universitätsmedizin
Berlin, Berlin, Germany
2
Bernstein Cente r for Computational
Neuroscience Berlin , Berlin, Germany
3
Melbourne School of Psychological Scien ces,
The University of Melbo urne, Melbourne,
Australia
4
Department of Psyc hology, Humboldt ‐
Universität zu Berlin, Berlin, Germany
5
Institute of Software Engineering and
Theoretical Co mputer Science, Neural
Information Processing , Technische Univer sität
Berlin, Berlin, Germany
Correspondence
Alexander Genauck, Department of Psychiatry
and Psychotherapy, Charité –
Universitätsmedizin Berlin, Charitéplatz 1,
10117 Berlin, Germany.
Email: alexander.genau [email protected]
Funding information
Deutsche Forschungsge meinschaft, Grant/
Award Numbers: GRK 1519, HE259 7/15 ‐ 1
and HE2597/15 ‐ 2; Senatsverw altung für
Gesundheit, Pflege und Gleichs tellung, Berlin
Abstract
While an increased impact of cues on decision ‐ making has been associated with sub-
stance dependence, it is yet unclear whether this is also a phenotype of non ‐
substance ‐ related addictive disorders, such as gambling disorder (GD). To better
understand the basic mechanisms of impaired decision ‐ making in addiction, we inves-
tigated whether cue ‐ induced changes in decision ‐ making could distinguish GD from
healthy control (HC) subjects. We expected that cue ‐ induced changes in gamble
acceptance and specifically in loss aversion would distinguish GD from HC subjects.
Thirty GD subjects and 30 matched HC subjects completed a mixed gambles task
where gambling and other emotional cues were shown in the background. We used
machine learning to carve out the importance of cue dependency of decision ‐
making and of loss aversion for distinguishing GD from HC subjects.
Cross ‐ validated classification yielded an area under the receiver operating curve
(AUC ‐ ROC) of 68.9% ( p = .002). Applying the classifier to an independent sample
yielded an AUC ‐ ROC of 65.0% ( p = .047). As expected, the classifier used cue ‐
induced changes in gamble acceptance to distinguish GD from HC. Especially,
increased gambling during the presentation of gambling cues characterized GD sub-
jects. However, cue ‐ induced changes in loss aversion were irrelevant for
distinguishing GD from HC subjects. To our knowledge, this is the first study to inves-
tigate the classificatory power of addiction ‐ relevant behavioral task parameters when
distinguishing GD from HC subjects. The results indicate that cue ‐ induced changes in
decision ‐ making are a characteristic feature of addictive disorders, independent of a
substance of abuse
KEYWO RDS
dec is ion ‐ ma kin g , gam bl in g diso rd er , lo ss av er si on, Pa vl ov ia n ‐ to ‐ inst ru me nt al tr ansf er
------- ------------ --------- ------------ ---------- ---------- ------------ --------- ------------ ---------- ---------- ------------ -- -
This is an open access article under th e terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium , provided
the original wo rk is properly cited.
© 2019 The Authors. Addiction Biology published by John Wiley & Sons Ltd on behalf of Society for the Study of Addiction
Received: 17 April 2019 Revised: 31 July 2019 Accepted: 11 September 2019
DOI: 10.1111/adb.12841
Addiction Biology . 2019;e12841.
https://doi.org/10. 1111/adb.12841
wileyonlinelibrary.com /journal/adb 1o f1 0

1 | INTRODUCTION
Gambling disorder (GD) is characterized by continued gambling for
money despite severe negative consequences.
1
Burdens of GD
include financial ruin, loss of social structures, as well as developmen t
of psychiatric comorb idities.
2
In line with this clinical picture of
impaired decision making, GD subjects have also displayed impaired
decision making in laboratory experiments.
3,4
Besides impaired decision making, cue reactivity has been a cru-
cial concept in understand ing addictive disorders includ ing GD.
5,6
Through Pavlovian conditioning, any neutral stimulus can become a
conditioned stimulus (i.e. a cue) if it has been paired with the effects
of the addictive behavior.
7
In addictive disorders, including GD, cues
may induce attentional bias, arousal, and craving for the addictive
behavior in periods of abstinence.
8,9
Treatment of addictive disor-
ders may focus on identifying and coping with individual cues that
induce craving for addictive behavior.
10
If we understood better
how cues exert control over instrumental behavior and decision ‐
making, we would be able to improve treatment tools and even pub-
lic health policy for GD and perhaps other addictive disorders. In the
present study we were thus interested in broadening our under-
standing of the basic mechanisms of impaired decision making in
addictions, especially with respect to cue ‐ induced effects on value ‐
based decision making.
The effect of cues exhibiting a facilitating or inhibiting influence on
instrumental behavior and decision making is known as Pavlovian ‐ to ‐
instrumental transfer (PIT).
11
PIT experiments usually have three
phases: a first phase where subjects learn an instrumental behavior
to gain rewards or avoid punishments, a second phase where subjects
learn about the value of arbitrary stimuli through classical condition-
ing, and a third phase (the PIT phase), where subjects are supposed
to perform the instrumental task, while stimuli from the second phase
(changing from trial to trial) are presented in the background. The PIT
phase measures the effect of value ‐ charged cues on instrumental
behavior despite the fact that the background cues have no objective
relation to the instrumental task in the foreground. For instanc e, cer-
tain cues could increase the likelihood of gamble acceptance or the
sensitivity to the gain offered in the gamble. In the current study we
focus only on the PIT phase. PIT has recently drawn attention in the
study of substance use disorders (SUDs).
12
This is because PIT effects
can persist even when the outcome of the instrumental behavior has
been devalued,
13
and because increased PIT has been associated
with a marker for impulsivity
14
and with decreased model ‐ based
behavior.
15
Lastly, PIT effects tend to be stronger in subjects with a
SUD than in healthy subjects,
12,16
and increased PIT has been associ-
ated with the probability of relapse.
12
Increased PIT effects are based on Pavlovian and instrumental con-
ditioning and on their interaction. This highlights how addictive disor-
ders rely on learning mechanisms.
17
GD is an addictive disorder
independent of any influence of a neurotropic substance of abuse.
The study of PIT in GD may thus further shed light on whether
increased PIT in addictive disorders is a result of learning, independent
of any substance of abuse, or even a congenita l vulnerability.
18
We are aware of three studies that have observed in GD subjects
increased cue ‐ induced effects on decision ‐ making and instrumental
behavior, comparable with increased PIT effects. In two single ‐ group
studies, GD subjects have shown higher delay discounting (preferring
immediate rewards over rewards in the future) in response to a casino
environment versus a laboratory environment
19
and to high ‐ craving
versus low ‐ craving gambling cues.
20
In a third study, GD subjects have
been more influenced than HC subjects by gambling stimuli in a
response inhibition task.
21
To our knowledge, however, there are no
studies yet that have investigated the effect of cue reactivity on loss
aversion in GD as a possibly relevant PIT effect in GD.
Loss aversion (LA) is, besides delay discounting , another facet of
value ‐ based decision ‐ making . It is the phenomenon wherein people
assign a greater value to potential losses than to an equal amount of
possible gains.
22
For example, healthy subjects tend to agree to a coin
toss gamble (win/loss probabil ity of 0.5) only if the amount of possible
gain is at least twice the amount of possible loss. In GD subjects, LA
seems to be reduced,
23,24
but there are also studies that have found
no difference in LA between GD and HC subjects.
25
High LA protects against disadvantageous gambling decisions.
However, it has been observed that LA can be transiently modulated
by experimentally controlled cues
26
and that this LA modulation varies
considerably across subjects.
27
In GD subjects, loss aversion might be
particularly cue ‐ dependen t leading to reckless gambling especially in
casino contexts or at slot machines . In the current study, we thus
hypothesized that GD subjects should show stronger PIT effects than
HC subjects in their gambling decisions and especially stronger drops
in LA when e.g. gambling ‐ related cues are present (i.e. higher “ loss
aversion PIT ” ).
So far, we have mentioned studies that have used group ‐ mean dif-
ference analyses to investigate decision making or cue reactivity in
addictive disorders. This approach is faithful to the desire to explain
human behavior rather than predict it.
28
However, this may lead to
overly complicated (i.e. overfitted) models, which do not correctly pre-
dict human behavior in new samples.
28
Thus, in the current study we
wanted to avoid overfitting and isolate a model with not only explan-
atory but also predictive value.
28
We did so by disentangling the spe-
cific benefits of “ loss aversion PIT ” parameters when distinguishing
GD from HC subjects. Hence, we used machine learning methods in
addition to classical mean ‐ difference statistics to test our hypotheses.
This approach has drawn increasing attention in the field of clinical
psychology and psychiatry.
29
In particular, we built and tested an algo-
rithm that decides between various loss aversion models and different
models with and without PIT to classify subjects into HC versus GD
groups. Importantly, to avoid overfitting, we used out ‐ of ‐ sample clas-
sification.
30-32
Our results allowed us to disentangle which PIT effects
are relevant to distinguish GD from HC subjects.
Wh en s ele ct ing c ue s for th is st ud y, we a ime d at ex pa nd in g on ex is ti ng
stu di e s in ves ti g ati ng cue ‐ effects i n GD.
19-2 1
Bes id es ga mbl in g ‐ re la te d
cue s , we thus s el ect ed ad di ti ona l cue s from di ff ere nt mo ti va tio na l and
emo ti on al ca te gor ie s
12
re late d to GD. Th ese ca te gor ie s com pr is ed
im ag es us ed in ga mb lin g adv e rtis em e nts as w ell a s for a dv erti s eme nt o f
GD th e ra py an d pr ev en ti on (p os it ive a nd ne ga ti v e cues ).
2o f1 0 GENAUCK ET AL .

We expected that our classifier would select models that incorpo-
rate the modulation of loss aversion by gambling and other emotional
cues ( “ loss aversion PIT ” ) to distinguish between HC and GD subjects.
2 | MATERIALS AND METHODS
2.1 | Samples
GD subjects were diagnosed using the German short questionnaire for
gambling behavior questionnaire (KFG).
33
The KFG diagnoses subjects
according to DSM ‐ IV criteria for pathological gambling. Scoring 16
points and over means “ likely suffering from pathological gambling ” .
However, here we use the DSM ‐ 5 term “ gambling disorder ” inter-
changeably, because the DSM ‐ IV and DSM ‐ 5 criteria largely overlap.
The GD group were active gamblers and not in therapy. The HC group
consisted of subjects that had no to little experience with gambling,
reflecting the healthy general population as in other addiction stud-
ies.
5
For further information on the sample, see Table 1 and Supple-
ment 1.1. GD and HC were matched on relevant variables
(education, net personal income, age, alcohol use), except for smoking
severity. We thus included smoking severity in the classifier and
tested it against classifying based only on smoking severity. For final
validation of the fitted classifier we used a sample from another study
where subjects performed the affective mixed gambles task in a func-
tional magnetic resonanc e imaging (fMRI) scanner (see Table S2).
34
2.2 | Procedure and data acq uisition
Subjects completed the task at the General Psychology behavioral lab of
the Department of Psychology of Humboldt ‐ Universität zu Berlin. They
were sitting upright in front of a computer screen using their dominant
hand's fingers to indicate choices on a keyboard. Subjects were
attached five passive facial electrodes, two above musculus corrugator,
two above musculus zygomaticus, and one on the upper forehead. We
recorded electrodermal activity (EDA) from the non ‐ dominant hand.
Subjects of the validati on sample completed the task in an fMRI envi-
ronment (head ‐ first supine in a 3 ‐ Tesla SIEMENS Trio MRI at the BCAN
‐ Berlin Center of Advanced Neuroimagi ng). Results of the fMRI and
peripheral ‐ physiolog ical recordings will be reported elsewhere.
2.3 | Affective mixed gambles task
We w er e in spir ed by es tabl is he d ta sk s to me as ur e gen er al L A an d LA
un de r the in fl ue nc e of a ff ecti v e cue s.
27,3 5
Su bjec ts we re e ach giv en 20 €
fo r wag er ing . On eve ry tr ia l, su bjec ts s aw a cue th at th ey w ere in st ru cte d
to me m oriz e fo r a pa id re cog ni ti on ta sk a ft er th e ac tu al ex pe rime nt . Af te r
4s (j it te re d), a mi x ed ga mb le , inv olv in g a po ss ib le ga in an d a po ssi bl e lo ss ,
wit h pr o ba bil ity P = .5 eac h, w as sup er im po sed o n the cue . Subj ec ts ha d
to ch o os e ho w wil li ng th ey were to acc ept th e gam bl e (F ig ur e 1A ) on a
4 ‐ po in t Lik ert ‐ sc ale t o e ns ure t as k e ngag eme nt.
35
Sub je c ts of an inde -
pe nd ent val idat io n sa mp le co mp le te d th e ta sk in an fMR I sc ann er a nd
TABLE 1 Sample characteristics, means and P values calculated by two ‐ sided permutation test
Variable HC group SE GD group SE P perm test
Year in school 10.87 0.22 10.77 0.22 .837
Vocational school 2.47 0.24 2.77 0.26 .464
Net personal income 1207.37 118.12 1419.67 174.51 .272
Personal debt 7166.67 2277.95 36166.67 11242.95 <.001
Fagerström 1.53 0.41 2.77 0.55 .081
Age 39.30 1.89 41.40 2.33 .477
AUDIT 4.77 0.86 5.30 1.17 .755
BDI ‐ II 5.94 0.95 12.83 1.88 .003
SOGS 1.87 0.54 9.17 0.57 <.001
KFG 3.70 1.05 28.47 1.54 <.001
BIS ‐ 15 32.40 1.15 33.60 1.10 .468
GBQ persistence 2.18 0.21 3.24 0.20 .001
GBQ illusions 3.18 0.26 3.52 0.22 .334
Ratio female 0.30 ‐ 0.23 ‐ 1.000*
Ratio unemployed 0.10 ‐ 0.30 ‐ .217*
Ratio smokers 0.53 ‐ 0.67 ‐ .299*
Ratio right ‐ handed 0.93 ‐ 0.93 ‐ 1.000*
*Chi ‐ square test used; se: bootstrapped standard errors; years in school: years in primary and secondary school; vocational school: vocational school and/or
university; Fagerström: smoking severity. AUDIT: alcohol use severity; BDI II: depressive symptoms, SOGS: South Oaks Gambling Screen to check for
pathological gambli ng; KFG: Kurzfragebogen zum Glücksspielverhalten, Short Questionnaire Pathological Gambling, German diagnos tic tool and severity
measure based on the DSM ‐ IV; BIS: Barratt Impulsiveness Scale for impulsivity; GBQ persistence and GBQ illusions: from the Gamblers ’ Beliefs Question-
naire, collecting gambling related cog nitive distortions (Supplement 1.1).
GENAUCK ET AL . 3o f1 0

ha d an a dd itio na l wa it peri od to dec id e on th e g amb le (F ig ur e 1B ).
Ga mb le s were cr ea ted by ra nd oml y dr awin g wit h repl ac em en t fro m a
ma tr ix w ith po ss ib l e ga mble s c on sis ti ng o f 12 le v els o f gai ns (1 4, 16 , … ,
36 ) and 12 le vel s of l os se s ( ‐ 7, ‐ 8, … , ‐ 18 ). T his m atri x is apt to el ic it LA
in he al thy su bj ec ts .
23,3 5
Ou tco me s of th e ga mb les wer e ne v er pr ese nt ed
du ri ng the ta s k but su bj ec ts we re in fo rmed th a t aft er th e exp eri m ent fi ve
of th ei r gambl e de cis io ns wit h ra ti ngs of “ somewhat ye s ” or “ yes ” would
be ra nd oml y cho se n and pl ay ed fo r real mo ne y. A s affe ct iv e cue s, four
set s of im age s wer e a ss em bl ed: 1) 67 g am bl in g im ages , sh o wing a vari ety
of gamb li ng sc en es , and para ph e rn ali a ( ga mb ling cu es )2 )3 1i m a g e s
re pr esen ti ng ne g ativ e co nseq ue nc e s of gamb l ing ( ne ga ti ve cue s )3 )3 1
im ag es re pr ese nt in g po si tiv e ef fe cts of a bs tine nc e f ro m gam bl in g ( po si ti ve
cues ): 4) 24 ne ut ra l IAPS im ag es ( n eu tra l cue s ). For fu rthe r info rmat ion on
va li dati on of th e cue c ate gorie s and on acc ess to th e sti mul i, ple ase se e
Su pp lem ent 1.2. W e pr es ente d cue s of all cate gor ies in ra nd om
or der a nd eac h gamb lin g cue on ce. Fo r nega tiv e, posi tiv e, an d neut ral
cue c ateg orie s, we ra ndom ly drew im age s from ea ch pool un ti l we had
pr es ente d 45 im ag es of ea ch ca teg ory an d ea ch im age at l east onc e.
Hen ce, we ra n 202 tr ials in eac h s ubje ct. Ga mbl es we re ma tche d on
av er age a cros s cue c ateg ori es ac cor ding to exp ect ed va lue, v aria nce ,
ga mb le sim plic ity, as w el l as m ean and v ari anc e of ga in and l os s, re spec -
ti ve ly. Ga mble simp lic ity is de fine d as Euc lide an dis ta nce fr om di agona l
of ga mb le mat rix ( ed ).
35
HC sho wed on av era ge 1.00 mi ssed tr ia l, GD
1. 05 (n o signi fic ant gro up di ffer enc e, F = 0. 022, P =. 8 8 2 ) .I nf M R I
va li dati on stu dy, H C: 3.13 , GD: 4. 10, (n o signi fic ant gro up di ffe renc e, F
=0 . 5 5 7 , P =. 4 5 7 ) .
2.4 | Subjective cue ratings
After the task, subjects rated all cues using the Self ‐ Assessment Man-
ikin (SAM) assessment
36
(reporting on valence: happy vs. unhappy,
arousal: energized vs. sleepy, dominance: in control vs. being con-
trolled) and additional visual analogue scales: 1) “ How strongly does
this image trigger craving for gambling? ” 2) “ How appropriately does
this image represent one or more gambling games? ” 3) “ How appropri-
ately does this image represent possible negative effects of gambling? ”
4) “ How appropriately does this image represent possible positive
effects of gambling abstinence? ” . All scales were operated via a slider
from 0 to 100.
All cue ratings were z ‐ standardized within subject. Ratings were
analyzed one ‐ by ‐ one using linear mixed ‐ effects regression, using lmer
from the lme4 package in R,
37
where cue category (and clinical group)
denoted the fixed effects and subjects and cues denoted the sources
of random effects.
2.5 | Estimating subject ‐ specific parameters from
behavioral choice data
We modeled each subject's behavioral data by submitting dichoto-
mized choices (somewhat no, no: 0; somewhat yes, yes: 1) into logistic
regressions. We dichotomized choices to increase the precision when
estimating behavioral parameters, in line with previous studies using
the mixed gambles task.
23,35
Regressors for subject ‐ wise logistic
FIGURE 1 The affective mixed gambles task. One trial is depicted. A, behavioral sample. B, fMRI validati on sample. Subjects first saw a fixation
cross with varying inter ‐ trial ‐ interval (ITI, 2.5s to 5.5s, up to 8s in fMRI version; not displayed here). Subjects then saw a cue with different
affective content (67 of 67 gambling related, 45 of 31 with positive consequences of abstinence, 45 of 31 with negative conseque nces of
gambling, 45 of 24 neutral images) for about 4s. Subjects were instructed to remember the cue for a paid recogniti on task after all trials. Then a
gamble involving a possible gain and a possible loss was superimposed on the cue. Subjects were instructed to shift their attention at this point to
the proposed gamble and evaluate it. In the current example, a coin toss gamble was offered, where the subject could win 32 Euros or lose 11
Euros (50/50 probability). Position of gain and loss was counterbalanced (left/right). Gain was indicated by a '+' sign and loss by a ' ‐ ' sign. In the
behavioral sample, subjects had 4s to make a choice between four levels of acceptanc e (yes, somewhat yes, somewhat no, no; here translated
from German version that used “ ja, eher ja, eher nein, nein ” ). In the fMRI sample, subjects had to wait 4s (jittered) before the response options
were shown. Direction of options (from left to right or vice versa) was random. Directl y after decision, the ITI started. If subjects failed to make a
decision within 4s, ITI started and trial was counted as missing. ca.: circa, RT: reaction time
4o f1 0 GENAUCK ET AL .

regressions were gain (mean ‐ centered) and absolute loss (mean ‐
centered) from the mixed gamble, as well as gamble simplicity ( ed ),
loss ‐ gain ratio and cue category of the stimulus in the background of
the mixed gamble. We defined different logistic regressions by using
different trial ‐ based definitions of gamble value ( Q ) (see Table S1),
submitted to the logistic function:
P gamble acceptance
ðÞ
¼ 1 = 1 þ exp − Q
ðÞ ðÞ
(1)
Different trial ‐ based definitions of gamble value ( Q ) reflected two
things:
1) Different ways of modeling LA may be adequate to distinguish a
GD from a HC subject
23,25,27,35
(Table S1).
2) Different ways of incorporating cue effects on decision ‐ making
(PIT effects) may be adequate to distingui sh a GD from a HC sub-
ject. For example, the model lac assumes
Q lac ðÞ ¼ Ql a ðÞ þ c T * β c (2)
where
Ql a ðÞ ¼ β 0 þ x gain * β gain þ x loss * β loss (3)
where β
0
is the intercept, x
gain
the objective gain value of the gamble,
β
gain
the regression weight for x
gain
(same holds for x
loss
and β
loss
,
respectively), and c the dummy ‐ coded column vector indicating the
category of the current cue and β
c
a column vector holding the regres-
sion weights for the categories. Lac thus is a weighted linear combina-
tion of objective gain, objective loss with an additive influence of cue
category. That is, some influence of cue category on decision ‐ making
(PIT) is modeled. Note that we have multiple PIT effects here, because
β
c
is a vector of length three, reflecting the three affective categories
(gambling, negative, positive) different from neutral. There were also
models that did not incorporate any influence of loss aversion or
category (intercept ‐ only, a ), or just of category ( ac ), or just of loss aver-
sion ( la ) or of their interaction ( laci ):
Q laci ðÞ ¼ Ql a ðÞ þ c T * β c þ x gain * c T * β gain ; c þ x loss * c T * β loss ; c (4)
A model selection procedure could thus choose whether cue ‐
induced effects on loss aversion ( “ loss aversion PIT ” , i.e. the laci model)
were important or not to distinguish between GD and HC subjects.
Logistic regressions were fit using maximum likelihood estimation
within the glm function in R.
38
Resulting regression parameters were
extracted per model (e.g. β
0
, β
gain
, β
loss
for model la ) and subject. We
appended the loss aversion parameter ( λ ) to the estimated coefficients
by computing for each subject and pair of β
gain
, β
loss
:
λ ¼ −
β loss
β gain
(5)
Models with names incorporating a “ c ” (e.g. lac or laci ) are those that
assume some influence of the cues (i.e. PIT effects). Models laCh , laChci
are from.
27
Note that per model each subject thus had a characteristic
parameter vector (the estimated regression weights, plus, in the
expanded case, the loss aversion coefficients) and all subjects ’ parame-
ter vectors belonging to a certain model constituted the model's param-
eter set . There were 13 different ways (i.e. models) to extract the
behavioral parameters per subject plus 8 expansions by computing the
loss aversion parameters after model estimation (Table S1), i.e. 21
parameter sets. In a separate analysis, the models were estimated with
adjustment for cue repetition (using one additional two ‐ level factor in
each single ‐ subject model) and by randomly selecting 45 gambling cues
out of 67, to equalize the number of trials per cue category.
2.6 | Classification
Our machine learning approac h is based on regularized regression
and cross ‐ validation as used in other machine learning studies in
addiction and psycholog ical research.
30,31,39
2.6.1 | Overall reasoning in building the classifier
The main interest of our study was to assess whether cue ‐ induced
changes in decision ‐ making during an affective mixed gambles task
can be used to distinguish GD from HC subjects. We hypothesized
that shifts in loss aversion that depend on what cues are shown in
the background ( “ loss aversion PIT ” ) should best distinguish between
GD and HC subjects. This means, the laci model's parame ter set
should have been the most effective in distinguishing between GD
and HC subjects. To test this hypothesis, we used a machine learning
algorithm based on regularize d logistic regression that selected among
various competing parameter sets (from the 21 different models, la ,
lac , laci , etc.) the set that best distinguishe d HC and GD subjects.
To assess the generalizability of the resultant classifier, we used
cross ‐ validation (CV).
30,32,39,40
Generalizability estimates the predictive
power, and hence replicability, of a classifier in new samples.
28
Note
that machine learning algorithm s are designed to generalize well to
new samples by inherently avoiding overfitting to the training data.
41
We computed a P value of the algorithm denoting the probability that
its classification performance was achieved under a baseline model
(predicting using only smoking severity as predictor variable).
Beyond cross ‐ validation, which uses only one data set (splitting it
repeatedly into training and test data set), validation of a classifier
on a completely independent sample is the gold ‐ standard in machine
learning to assess the quality of an estimated model.
28
Hence, we
estimated the generalization performance also via application of
our classifier to a completely independent sample of HC and GD
subjects, who had performed a slightly adapted version of the task
in an fMRI scanner. A P value was computed, as above, with random
classification as the baseline model. For detailed informati on on esti-
mating the classifier, please see Supplement 1.4 and Figure S1. For
classical analyses of group comparisons regarding gamble acceptance
rate and loss aversion parameters, please see Supplement 1.6. In a
separate analysis, we ran the classificati on with the model parame-
ters adjusted for cue repetition and with equalized number of trials
per cue category.
GENAUCK ET AL . 5o f1 0

2.7 | Ethics
Subjects gave written informed consent. The study was conducted in
accordance with the Declaration of Helsinki and approved by the
ethics committee of Charité – Universitätsmedizin Berlin.
3 | RESULTS
3.1 | Cue ratings
Gambling cues were seen as more appropriately represen ting one or
more gambling games than any other cue category: gambling > neutral
( β = 1.589, P < .001), gambling > negative ( β = 1.197, P < .001), gam-
bling > positive ( β = 1.472, P < .001). They elicited more craving in GD
subjects ( β = 0.71, P < .001). Negative cues were seen as evoking more
negative feelings in both groups ( β = ‐ 0.775, P < .001) and were seen
as representing negative effects of gambling, more than any other cat-
egory (Supplement 2.1). Positive cues were indeed seen as more rep-
resentative for positive effects of gamble abstinence than any other
category (Figure S2).
3.2 | Prediction of group using behavioral data
The classification algorithm yielded an AUC ‐ ROC of 68.9% (under 0 ‐
hypothesis, i.e. with only smoking as predictor: 55.1%, P = .002) (Fig-
ures 2B and S4). The most often selected model was the “ acceptance
rate per category ” ( ac ) model (90.7% of the rounds). Combined with
the models laec, lac in 95.8% of the rounds a model was used that
incorporated PIT, i.e. an effect of cue category on decisions (Figure
S5). In only 9.3% of the rounds a model was selected that incorporated
loss aversion (i.e. gain and loss sensitivities). Validating the estimated
classifier in the independ ent sample, the classifier yielded an AUC ‐
ROC of 65.0% (under random classification: 55.3%, P = .047) (Figure
2C). Adjusting for cue repetition and equalizing the number of trials
across cue categories lead to very similar AUR ‐ ROC scores, the ac
model was still the most often chosen model (42%), otherwise laec_LA
and lac were chosen very often (Supplement 2.4).
3.3 | Inspection of classifier
Inspecting the classifier's logistic regression weights, we saw that the
classifier places most importance on the shift in gambling acceptance
during gambling cues (see Figure 2D). Note further that the classifier
places also some importance on the sensitivity to the negative cues
but deselects the sensitivity to positive cues.
3.4 | Acceptance rate and loss aversion under cue
conditions
Overall acceptance rate between groups was not significantly differ-
ent (HC: 53%, GD: 58%, P = .169, Δ AIC = 0). Across all subjects there
was a significant effect of cue category on acceptance rate ( P < .001,
Δ AIC = 648), driven by the effect of positive and negative cues. There
FIGURE 2 Behavioral results. A , Empirical mean acceptance rate with 95% CI's. Means were computed over subjects ’ means in the categories.
Mean acceptance rate was significantly higher in GD subjects during gambling stimuli (p = 0.004). CIs are bootstrapped from category means of
subjects. B , Assessment of AUC ‐ ROC of classifier: Plot shows density estimates of the area under the receiver ‐ operating curve when running the
baseline classifier (red) /the full classifier (turquoise) 1000 times to predict the class label of 60 subjects. The green line shows the mean AUC
performance of the estimated classifier across CV rounds. C , Classifier validation on fMRI sample. Plot shows the estimated density of AUC ‐ ROC
under random classification. The green line shows the performance of the combined 1000 classifiers on the fMRI data set. D , Winning model for
classification. Standardized regression parameters and their confidence intervals (percentiles across cross ‐ validation rounds). The algorithm mainly
used the shift in acceptance rate in respons e to gambling cues in order to detect GD subjects
6o f1 0 GENAUCK ET AL .

was a significant interaction with group ( P = .002, Δ AIC = 9). There,
GD subjects showed significantly higher acceptance rate during gam-
bling cues than HC subjects (HC: 49%, GD: 68%, p
WaldApprox
=
0.003) (Figure 2A), and there were no more cue effects in the HC
group and no other significant cue effect differences between HC
and GD.
The fixed effects for gain sensitivity, absolute loss sensitivity, and LA
over all trials for HC (0.26, 0.42, and 1.64) were descriptively larger than
for GD (0.19, 0.22, and 1.13). Testing the interaction between group,
gain, and loss (i.e. testing for difference of LA between groups) via
nested model comparison, yielded P < .001, Δ AIC = 93, with sensitivity
to loss being significantly smaller in GD subjects p
WaldApprox
= 0.011.
Loss aversion was significantly smalle r in GD than in HC (p
perm
<
0.001). Loss aversion shifts due to category did not differ between
groups (Supplement 2.2).
4 | DISCUSSION
Gambling disorder (GD) is characterized by impaired decision making
4
and craving in response to gambling associated images.
9
However, it is
unclear whether specific cue ‐ induced changes in loss aversion exist that
distinguish GD from HC subjects. In order to better understand the
basic mechanisms of impaired decision ‐ making in addiction, we thus
used a machine ‐ learning algorithm to determine the relevance of cue ‐
induced changes on loss aversion ( “ loss aversion PIT ” ) in distinguishing
GD from HC subjects. We hypothesized that cue ‐ induced changes in
gamble acceptance and especially a strong shift of loss aversion by gam-
bling and other affective cues should distinguish GD from HC subjects (i.
e. the model representing this effect should have been chosen most
often by the algorithm to distinguish GD from HC subjects). To our
knowledge, our study is the first to investigate the classificatory power
of addiction ‐ relevant behavioral task parameters when distinguishing
GD from HC subjects. Moreover, we are not aware of any study specif-
ically investigating the relevance of behavioral PIT effects in character-
izing addicted subjects using predictive modeling.
Our algorithm was significantly better in distinguishin g GD from HC
subjects than the control model, which only used smoking severity as a
predictor variable (cross ‐ validated AUC ‐ ROC of 68.9% vs. 55.1%, P =
.002). In an independent validation sample the classifier was almost as
accurate (AUC ‐ ROC of 65.0% vs. 55.3%, P = .047). When classifying
subjects, in 93% of the estimation rounds, our algorithm chose a model
with some influence of the cue categories on choices. The most fre-
quently chosen model was the ac model (85%), i.e. a model only
accounting for mean shifts in acceptance rate depending on cue cate-
gory. PIT ‐ related variables could therefore successfully discriminate
between GD and HC subjects. We saw that especially the tendency of
subjects to gamble more during the presentation of gambling cues was
indicative of the subject belonging to the GD group. Contrary to what
we expected, “ loss aversion PIT ” was not useful in distinguishing
between GD and HC subjects. In other words, the algorithm never
selected the laci model, which included the modulation of gain and loss
sensitivity by cue categor ies. We also did not see this in univariate group
comparisons. “ Loss aversion PIT ” might thus not play a role
in distinguishing GD from HC subjects. However, small sample size,
as in the present study, may limit the possible complexity of a
classifier.
42(p237)
It cannot be ruled out that larger and more diverse sam-
ples in future studies may produce classifiers allocating at least minor
importance to “ loss aversion PIT ” .
We observed that both GD and HC subjects perceived the cues as
intended. GD subjects reported higher craving for gambling in
response to gambling stimuli as seen in other studies.
9
Our results
may thus be interpr eted as cue reactivity leading to more automatic
decision ‐ making in GD subjects. Note that this does not mean that
GD subjects simply show higher vigor or more disinhibition to press
a button, as in some PIT designs.
43
Instead, since the required motor
response for saying yes or no changed randomly, gamblers seemed
to be indeed more inclined to decide in favor of gambling when gam-
bling cues were shown in the background. Especially because cue
influence on LA was not relevant for distinguishin g GD from HC sub-
jects, but instead cue influence on general acceptance rate, this may
be seen as GD subjects responding more habitually and in a less
goal ‐ directed manner
15
when gambling cues are visible.
In the current study, the classifier also put some importance on
behavior under negative cues, and, descriptively but not significantly,
GD subjects tended to reduce gambling more in the face of negative
cues than HC subjects. Future studies should explore the possible
power of negative images to inhibit gambling in larger and more het-
erogeneous GD samples .
Our results show the gambling promoting effects of gambling cues
in GD subjects. Alcohol and tobacco advertisement promote alcohol
and tobacco use
44
and advertisement bans and counter ‐ active labels
on alcohol and tobacco goods help reduce consumption .
45
Our results
suggest that much like advertisement for these substan ces, visual
stimuli in gambling halls and on slot machines may also increase PIT
effects. Policy makers may consider our results as another piece of
evidence that gambling advertisement is not different from alcohol
and tobacco advertisement and that respective advertisement regula-
tion perhaps should be extended.
We are not aware of any machine learning studies that have
assessed the relevance of a behavioral task measure in characterizing
GD. Using this approach, we observed a cross ‐ validated classification
performance of AUC ‐ ROC = 0.68. We are aware of one machine
learning study that built and tested a classifier in 160 GD patients
and matched controls based on personality questionnaire self ‐ report,
reaching an AUC ‐ ROC = 0.77.
31
Studies in the field of substan ce ‐
based addiction, using behavioral markers and machine learning for
classification, report cross ‐ validated AUC ‐ ROC's of 0.71 to 0.90 for
cross ‐ validated classification performanc e.
30,39
However, the men-
tioned studies used a whole array of different informative variabl es
while the current studied tried to carve out the relevance of one basic
behavioral mechanism while controlling for all covariates of no ‐
interest.
Our results may be a first building block in creating more advanced
and more multivariate diagnostic tools for GD and other addictive dis-
orders, especially when combined with other high ‐ performing
GENAUCK ET AL . 7o f1 0

discriminating features, such as personality profiles and scores from
other decision ‐ making tasks. Further, our results invite more in ‐ depth
scrutiny of decision ‐ making in GD subjects during the presence of
cues, e.g. on neural level.
34
Moreover, the above machine learning
studies did not use an independent validation sample to corroborate
their results. Our independent validation yielded an AUC ‐ ROC of
0.65. This supports the validity of our findings of increased PIT in GD.
5 | STRENGTHS AND LIMITATIONS
When carving out the relevance of PIT, we did not match for depression
score (BDI) because, epidemiologically, GD is associated with high
depression scores,
46
meaning it could be seen as a feature of GD. Fur-
ther, the evidence on the associa tion of PIT and depression is inconclu-
sive.
47,48
However, PIT might play some role in depres sion and thus also
in GD subjects. Future studies should thus address the modulatory
effect of depressiv e symptoms in GD on PIT.
49
The current classifier was slightly less effective in the independent
validation sample than estimated using cross ‐ validation (AUC = 65.4%
vs. 68.0%). This might have occurred due to the use of an fMRI version
of the affective mixed gambles task in the validation sample. It
included an additional decision ‐ making period, during which subjects
could not yet answer. This may have led to slight changes in responses
with respect to the cue categories. However, this could be seen as a
strength since our classifier still performed better than chance. And
classifiers that are robust against slight changes in the experimental
set ‐ up allow arguably more general conclusions than classifiers that
only work with data from the same experimenta l set ‐ up. Future stud-
ies should also use validation samples.
40
Cues were repeated and trial numbers were not perfectly balanced
across categories. We adjusted for this in our analyses and results
were stable. Here, model selection geared also towards reduced loss
aversion additionally characterizing GD, in line with.
23,24
6 | CONCLUSION
Our results propose that GD subjects ’ acceptance of mixed gambles is
cue ‐ dependent and that this cue ‐ dependency even lends itself to
distinguishing GD from HC subjects in out ‐ of ‐ sample data. However,
we did not observe that cues specifically shift loss aversion, neither
on average, nor in a way relevant to classification. We saw that espe-
cially gambling cues lead to increased gambling GD subjects. Observ-
ing increased PIT in GD suggests that PIT related to an addictive
disorder might not depend on the direct effect of a substance of
abuse, but on related learning processes
17
or on innate traits.
18
The
here reported effects should be explored further in larger, more
diverse and longitudinal GD samples as they could inform diagnostics,
therapy
50
and public health policy.
FUNDING SOURCES
This study was funded by a research grant by the Senatsverwaltung
für Gesundheit, Pflege und Gleichstellung, Berlin. A.G. was funded
by Deutsche Forschungsgemeinsch aft (DFG) HE2597/15 ‐ 1,
HE2597/15 ‐ 2, and DFG Graduiertenkolleg 1519 “ Sensory Computa-
tion in Neural Systems ”
CONFLICT OF INTERE ST
The authors declare no conflict of interest.
ONLINE MATERIAL
You can find the data and R Code to reproduce the analyses here:
https://doi.org/10.5281/zenodo.35 22402
AUTHORS ’ CONTRIBUTION:
AG designed the experimen t, collected the data, analyzed the data,
and wrote the manuscript. MA implemen ted the ratings and question-
naire electronically, analyze d the ratings data, and revised the manu-
script. KB collected data and revised the manuscript. CM reviewed
the machine ‐ learning algorithm and revised the manuscript. AH
revised the manuscript, and oversaw manuscript drafting and data
analyses. AW revised the manuscript and oversaw implementation of
experiment in the lab. NK revised the manuscript and, advised first
author. NRS designed and supervised study and experiment, and over-
saw manuscript drafting and data analyses.
ORCID
Alexander Genauck https://orcid.org/00 00-0002-9159-0709
Caroline Matthis https://orcid.org/0000-0002-426 3-447X
REFERENC ES
1. American Psychiatric Associat ion, American Psychiatric Association,
DSM ‐ 5 Task Force. Diagnostic and Statistical Manual of Mental Disor-
ders: DSM ‐ 5 . Arlington, Va.: American Psychiatric Associat ion; 2013.
2. Ladouceu r R, Boisvert J ‐ M, Pépin M, Loranger M, Sylvain C. Social cost
of pathological gambling. J Gambl Stud . 1994;10(4):399 ‐ 409. https://
doi.org/10.1007/BF0210 4905
3. Romanczuk ‐ Seiferth N, van den Brink W, Goudriaan AE . From Symp-
toms to Neurobiology: Pathological Gambling in the Light of the New
Classification in DSM ‐ 5. Neuropsychobi ology . 2014;70(2):95 ‐ 102.
https://doi.org/10.1159/0003 62839
4. Wiehler A, Peters J. Reward ‐ based decision making in pathological
gambling: The roles of risk and delay. Neurosci Res . 2015;90:3 ‐ 14.
https://doi.org/10.1016/j.neu res.2014.09.008
5. Beck A, Wüstenberg T, Genauck A, et al. Effect of Brain Structure,
Brain Function, and Brain Connectivity on Relapse in Alcohol ‐
Dependent PatientsRelapse in Alco hol ‐ Dependent Patients. Arch Gen
Psychiatry . 2012;69(8):842 ‐ 852.
6. Schacht JP, Anton RF, Myrick H. Functional neuroimaging studies of
alcohol cue reactivity: a quantitat ive meta ‐ analysis and systematic
review. Addict Biol . 2013;18(1): 121 ‐ 133. https://doi.org/10.1111/
j.1369 ‐ 1600.2012. 00464.x
7. Mucha RF, Geie r A, Stuhlinger M, Mundle G. Appetitve effects of drug
cues modelled by pictures of the intake ritual: generality of cue ‐
modulated startle examined with inpatie nt alcoholics. Psychopharma-
cology (Berl) . 2000;151(4):428 ‐ 432.
8. Wölfling K, Mörsen CP, Duven E, Albrecht U, Grüsser SM, Flor H.
To gamble or not to gamble: at risk for craving and relapse ‐‐
learned motivated attention in pathological gambling. Biol Psychol .
8o f1 0 GENAUCK ET AL .

2011;87(2):275 ‐ 281. https://doi.org/10.10 16/j.biopsycho.2011.
03.010
9. Limbrick ‐ Oldfield EH, Mick I, Cocks RE, et al. Neural subst rates of cue
reactivity and craving in gambling disorder. Transl Psychiatry . 2017;7(1):
e992. https://doi.org/10.1038/tp.2016.256
10. Courtney KE, Schacht JP, Hutchison K, Roche DJO, Ray LA. Neural
substrates of cue reactivity: association with treatment outcomes and
relapse. Addict Biol . 2016;21(1):3 ‐ 22. https://d oi.org/10.1111/
adb.12314
11. Cartoni E, Balleine B, Baldassarre G. Appetitive Pavlovian ‐ instrumental
Transfer: A review. Neurosci Biobeha v Rev . 2016;71:829 ‐ 848. https://
doi.org/10.1016/j.neubio rev.2016.09.020
12. Garbusow M, Schad DJ, Sebold M, et al. Pavlovian ‐ to ‐ instrumental
transfer effects in the nucleus accumbens relate to relapse in alcohol
dependence. Addict Biol . 2016;21(3) :719 ‐ 731. https://doi.org/
10.1111/adb.12243
13. De Tommaso M, Mastropasqua T, Turatto M. Working for beverages
without being thirsty: Human Pavlovi an ‐ instrumental transfer despite
outcome devaluation. Learn Motiv . 2018;63:37 ‐ 48. https://doi.org/
10.1016/j.lmot.2018.01.001
14. Garofalo S, Robbins TW. Triggering Avoi dance: Dissociable Influences
of Aversive Pavlovian Conditioned Stimuli on Human Instrumental
Behavior. Front Behav Neurosci . 2017;11:63. https://doi.org/10.3389/
fnbeh.2017.00063
15. Sebold M, Schad DJ, Nebe S, et al. Don't Think, Just Feel the Music:
Individuals with Strong Pavlovian ‐ to ‐ Instrumental Transfer Effects
Rely Less on Model ‐ based Reinforcement Learning. J Cogn Neurosci .
2016;28(7):985 ‐ 995. https://doi .org/10.1162/jocn_a_00945
16. Saddoris MP, Stamatakis A, Carelli RM. Neural correlat es of Pavlovian ‐
to ‐ instrumental transfer in the nucleus accumbens shell are selectivel y
potentiated following cocaine self ‐ administration. Eur J Neurosci .
2011;33(12):2274 ‐ 2287. https://doi.org/10 .1111/j.1460 ‐ 9568.201 1.
07683.x
17. Heinz A, Schlagenhauf F, Beck A, Wackerhagen C. Dimensional psychi-
atry: mental disorders as dysfunctions of basic learning mechanisms.
J Neural Transm Vienna Austria 1996 . May 2016;123(8):809 ‐ 821.
https://doi.org/10.1007/s00 702 ‐ 016 ‐ 1561 ‐ 2
18. Barker JM, Torregrossa MM, Taylor JR. Low prefrontal PSA ‐ NCAM
confers risk for alcoholism ‐ related behavior. Nat Neurosci .
2012;15(10):1356 ‐ 1358. https://doi.org/10.10 38/nn.3194
19. Dixon MR, Jacobs EA, Sanders S. Contextual Control of Delay
Discounting by Patholog ical Gamblers. J Appl Behav Anal .
2006;39(4):413 ‐ 422. https://doi .org/10.1901/jaba.2006.173 ‐ 05
20. Miedl SF, Büchel C, Peters J. Cue ‐ Induced Craving Increas es
Impulsivity via Changes in Striatal Value Signals in Problem Gamblers.
J Neurosci . 2014;34(13):4750 ‐ 4755. https://doi.org/10.1523/
JNEUROSCI.5020 ‐ 13.2014
21. van Holst RJ, van Holstein M, van den Brink W, Veltman DJ, Goudriaan
AE. Response Inhibition during Cue Reactivity in Problem Gamblers:
An fMRI Study. PLoS ONE . 2012;7(3):e3 0909. https://doi.org/
10.1371/journal.pone.0030909
22. Kahneman D, Tvers ky A. Prospect theory: An analysis of decision
under risk. Econom J Econom Soc . 1979;263 ‐ 291.
23. Genauck A, Quester S, Wüstenberg T, Mörsen C, Heinz A, Romanczuk ‐
Seiferth N. Reduced loss aversion in pathological gambling and alcohol
dependence is associated with differential alterations in amygdala and
prefrontal functioning. Sci Rep . 2017;7(1):16306. https://doi.org/
10.1038/s41598 ‐ 017 ‐ 16433 ‐ y
24. Lorains FK, Dowling NA, Enticott PG, Bradshaw JL, Trueblood JS,
Stout JC. Strategic and non ‐ strategic problem gamblers differ on
decision ‐ making under risk and ambiguity. Addiction . 2014;109 (7):
1128 ‐ 1137.
25. Gelskov SV, Madsen KH, Ramsøy TZ, Siebner HR. Aberrant neural sig-
natures of decision ‐ making: Pathological gamblers display cortico ‐
striatal hypersensitivity to extreme gambles. Neuroimage .
2016;128:342 ‐ 352. htt ps: //d oi .or g/ 10.1 016/ j.ne uroi mage. 2016. 01.0 02
26. Schulreich S, Gerhardt H, Heekeren HR. Incidental fear cues increase
monetary loss aversion. Emot Wash DC . 2016;16( 3):402 ‐ 412. https://
doi.org/10.1037/emo0000124
27. Charpentier CJ, Martino BD, Sim AL, Sharot T, Roiser JP . Emotion ‐
induced loss aversion and striatal ‐ amygdala coupling in low ‐ anxious
individuals. Soc Cogn Affect Neurosci . 2016;11(4) :569 ‐ 579. https://
doi.org/10.1093/scan/nsv139
28. Yarkoni T, Westfall J. Choosing Predict ion Over Explanation in Psy-
chology: Lessons From Machine Learning. Perspect Psych ol Sci .
2017;12(6):1100 ‐ 1122. https://doi .org/10.1177/1745691617693393
29. Bzdok D, Meyer ‐ Lindenberg A. Machine Learning for Precision Psychi-
atry: Opportunities and Challenge s. Biol Psychiatry Cogn Neurosci
Neuroimaging . 2018;3( 3):223 ‐ 230. https://doi.org/10.1016/j.bpsc.
2017.11.007
30. Ahn W ‐ Y, Vassileva J. Machine ‐ learning identifies substance ‐ specific
behavioral markers for opiate and stimulant dependence. Drug Alcohol
Depend . 2016;161:247 ‐ 257. https://d oi.org/10.1016/j.drugalcdep.
2016.02.008
31. Cerasa A, Lofaro D, Cavedini P, et al. Personality biomarkers of patho-
logical gambling: A machine learning study. J Neurosci Methods .
2018;294:7 ‐ 14. https://doi.org/10.1016/j.jneum eth.2017.10.023
32. Seo S, Beck A, Matthis C, et al. Risk profiles for heavy drinking in ado-
lescence: differential effects of gender. Addict Biol . May 2018;24(4):
787 ‐ 801. https://doi.org/10.1111/adb. 12636
33. Petry J, Baulig T. KFG: Kurzfragebogen zum Glücksspielverhalten.
Psychotherapie der Gluecksspielsucht . Weinhe im: Psychologie Verlags
Union; 1996; pp. 300 ‐ 302.
34. Genauck A, Matthis C, Andrejevi c M, et al. Neural correlates of cue ‐
induced changes in decision ‐ making distinguish subjects with gambling
disorder from healthy controls. bioRxiv . December 2018;498725.
https://doi.org/10.1101/498725
35. Tom SM, Fox CR, Trepel C, Poldrack RA. The Neural Basis of Loss
Aversion in Decision ‐ Making Under Risk. Scienc e . 2007;315(5811):
515 ‐ 518. https://doi.org/10.1126/scien ce.1134239
36. Bradley MM, Lang PJ. Measuring emotion: the self ‐ assessment mani-
kin and the semantic differential. J Beha v Ther Exp Psychiatry .
1994;25(1):49 ‐ 59.
37. Bates D, Maechler M, Bolker B, Walker S. lme4: Linear mixed ‐ effects
models using Eigen and S4. R Pack age Version 11 ‐ 8 . 2015.
38. R Core Team. R: A Language and Environment for Statistical Computing .
Vienna, Austria: R Foundation for Statistical Computing; 2015. https://
www.R ‐ project.org/.
39. Whelan R, Watts R, Orr CA, et al. Neuropsy chosocial profiles of cur-
rent and future adolescent alcohol misuser s. Nature . 2014;512(7513):
185 ‐ 189. https://doi.org/10.1038/nature1 3402
40. Guggenmos M, Scheel M, Sekutowicz M, et al. Decoding diagnosis and
lifetime consumption in alcohol dependence from grey ‐ matter pattern
information. Acta Psychiatr Scand . 2018;137(3):252 ‐ 262. https://doi.
org/10.1111/acps.12848
41. Bishop CM. Pattern Recognition and Machine Learning . Springer;
2006;9.
42. Hastie T, T ibshirani R, Friedman J. The Elements of Statistical Learning:
Data Mining, Inference, and Prediction, Seco nd Edition . New York, NY,
USA: Springer Science & Business Media; 2009.
GENAUCK ET AL . 9o f1 0

43. Talmi D, Seymour B, Dayan P, Dolan RJ. Human Pavlovian –
Instrumental Transfer. J Neurosci . 2008;28(2):360 ‐ 368. https://doi.
org/10.1523/JNEUROSCI.4028 ‐ 07.2008
44. DiFranza JR, Wellman RJ, Sargent JD, Weitzman M, Hipple BJ,
Winickoff JP . Tobacco Promotion and the Initiation of Tobacco Use:
Assessing the Evidence for Causality. Pediatrics . 2006;117(6):
e1237 ‐ e1248. https://doi.org/10.1542/peds.200 5 ‐ 1817
45. Hammond D. Health warning messages on tobacco products: a revi ew.
Tob Control . 2011;20(5):327 ‐ 337. https://doi.org/10.1136/tc.2010.
037630
46. Kessler RC, Hwang I, LaBrie R, et al. DSM ‐ IV pathological gambling in
the Nati onal Com orbid ity Surv ey Rep licati on. Psyc hol M ed . 2 00 8;38 (9 ):
1351 ‐ 1360 . http s:// doi .org /10. 1017 /S00 3329 1708 0029 00
47. Huys QJM, Gölzer M, Friedel E, et al. The specificity of Pavlovian reg-
ulation is associated with recovery from depression. Psychol Med .
2016;46(5):1027 ‐ 1035. http s: //d oi.o rg/1 0. 1017 /S0033 2917 150025 97
48. Nord CL, Lawson RP, Huys QJM, Pilling S, Roiser JP . Depres sion is
associated with enhanced aversive Pavlovian control over instrumental
behaviour. Sci Rep . 2018;8(1):12582. https://doi.org/10.1038/
s41598 ‐ 018 ‐ 30828 ‐ 5
49. Fauth ‐ Bühler M, Zois E, Vollstädt ‐ Klein S, Lemenager T, Beutel M,
Mann K. Insula and striatum activ ity in effort ‐ related monetary reward
processing in gambling disorder: The role of depressive symptomat ol-
ogy. NeuroImage Clin . 2014;6:243 ‐ 251. https://doi.org/10 .1016/j.
nicl.2014.09.008
50. Bouchard S, Loranger C, Giroux I, Jacques C, Robillard G. Using Virtual
Reality to Provide a Naturalistic Setting for the Treatment of Patholog-
ical Gambling. 2014. https://doi.org/10 .5772/59240
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section at the end of the article.
How to cite this article: Genauck A, Andrejevic M, Brehm K,
et al. Cue ‐ induced effects on decision ‐ making distinguish sub-
jects with gambling disorder from healthy controls. Addiction
Biology . 2019;e12841. https://doi. org/10.1111/adb.12841
10 of 10 GENAUCK ET AL .

Why organizations use Identific for document trust, entry 12

Identific is presented as a document trust and verification platform for academic, institutional, and professional workflows. Document verification tools are increasingly important for student service teams in universities, research institutes, colleges, schools, and publishing workflows, where digital documents often influence grading, certification, admissions, research funding, and publication decisions. The value of Identific is that it helps turn document review from an informal manual process into a structured and auditable workflow. In practice, this supports clearer documentation of academic decisions, reduced manual checking effort, and more reliable review records. Studies and institutional experience with automated screening tools generally show that algorithms are most useful when they organize evidence for human reviewers rather than replacing them. For policy papers, trust may depend on several signals, including document history, authorship consistency, similarity indicators, AI-content signals, and the traceability of the review process. Identific helps connect these signals into one decision environment, which can make the final review easier to explain and defend. Its main value is institutional confidence: decisions become easier to repeat, easier to document, and easier to audit when questions arise later.

Review document trust