Computational mechanisms underlying the impact of Pavlovian bias on instrumental learning in problematic social media users

Lu Liu; Yi-Xu Pang; Zhi-Hao Song; Si-Jia Chen; Ying-Yi Han; Yuan-Wei Yao

doi:10.1556/2006.2025.00026

Journal of Behavioral Addictions

Volume/Issue:: Accepted Manuscript / Online First

Computational mechanisms underlying the impact of Pavlovian bias on instrumental learning in problematic social media users

Authors:

Lu Liu

Lu Liu Department of Psychology, Sun Yat-sen University, Guangzhou, China

Search for other papers by Lu Liu in
Current site
Google Scholar
PubMed

https://orcid.org/0000-0001-5461-6706

Yi-Xu Pang

Yi-Xu Pang Department of Psychology, Sun Yat-sen University, Guangzhou, China

Search for other papers by Yi-Xu Pang in
Current site
Google Scholar
PubMed

Zhi-Hao Song

Zhi-Hao Song Department of Psychology, Sun Yat-sen University, Guangzhou, China

Search for other papers by Zhi-Hao Song in
Current site
Google Scholar
PubMed

Si-Jia Chen

Si-Jia Chen Department of Psychology, Sun Yat-sen University, Guangzhou, China

Search for other papers by Si-Jia Chen in
Current site
Google Scholar
PubMed

Ying-Yi Han

Ying-Yi Han Department of Psychology, Sun Yat-sen University, Guangzhou, China

Search for other papers by Ying-Yi Han in
Current site
Google Scholar
PubMed

, and

Yuan-Wei Yao

Yuan-Wei Yao Department of Psychology, The University of Hong Kong, Hong Kong SAR, China
State Key Laboratory of Brain and Cognitive Sciences, The University of Hong Kong, Hong Kong SAR, China

Search for other papers by Yuan-Wei Yao in
Current site
Google Scholar
PubMed

ywyao@hku.hk https://orcid.org/0000-0002-9635-7826

View More View Less

Online Publication Date:: 08 Apr 2025

Publication Date:: 08 Apr 2025

Article Category:: Research Article

DOI:: https://doi.org/10.1556/2006.2025.00026

Keywords:: problematic social media use; Pavlovian bias; instrumental learning; computational modeling; reward sensitivity

Open access

Download PDF

Get Permissions

Abstract

Background and aims

Problematic social media use (PSMU), a potential behavioral addiction, has become a worldwide mental health concern. An imbalanced interaction between Pavlovian and instrumental learning systems has been proposed to be central to addiction. However, it remains unclear whether individuals with PSMU also over-rely on the Pavlovian system when flexible instrumental learning is required.

Methods

To address this question, we used an orthogonalized go/no-go task that distinguished two axes of behavioral control during associative learning: valence (reward or punishment) and action (approach or avoidance). We compared the learning performance of 33 individuals with PSMU and 32 regular social media users in this task. Moreover, latent cognitive factors involved in this task, such as learning rate and reward sensitivity, were estimated using a computational modeling approach.

Results

The PSMU group showed worse learning performance when Pavlovian and instrumental systems were incongruent in the reward, but not the punishment, domain. Computational modeling results showed a higher learning rate and lower reward sensitivity in the PSMU group than in the control group.

Conclusions

This study elucidated the computational mechanisms underlying suboptimal instrumental learning in individuals with PSMU. These findings not only highlight the potential of computational modeling to advance our understanding of PSMU, but also shed new light on the development of effective interventions for this disorder.

Abstract

Background and aims

Methods

Results

Conclusions

Introduction

Social media use has become a near-ubiquitous part of our daily life over the last decade. As of 2023, there were more than 4.5 billion active social media users (https://www.statista.com/statistics/278414). Despite its popularity and convenience, the negative effects of problematic social media use (PSMU) on the physical health and social functioning have been widely reported (Sun & Zhang, 2021). PSMU has been considered a potential behavioral addiction given its similarities to other established addictions, although its classification is still under debate, partly because of the lack of empirical studies examining the cognitive and neurobiological mechanisms associated with this disorder (Brand et al., 2020; Cataldo, Billieux, Esposito, & Corazza, 2022; Montag, Marciano, Schulz, & Becker, 2023; Wegmann, Müller, Turel, & Brand, 2020).

Altered associative learning, such as automatic approach to addictive stimuli, has been hypothesized to play a critical role in PSMU and other addictions (Heinz et al., 2019; Lindström et al., 2021; Liu et al., 2024; Meshi, Elizarova, Bender, & Verdejo-Garcia, 2019; Redish & Johnson, 2007). Associative learning is primarily controlled by the Pavlovian and instrumental systems (Gershman, Guitart-Masip, & Cavanagh, 2021). The Pavlovian system supports the acquisition of intrinsic stimulus-outcome associations, whereas the instrumental system supports a more flexible learning of the optimal response to each stimulus based on related outcomes (Guitart-Masip, Duzel, Dolan, & Dayan, 2014; Queirazza, Steele, Krishnadas, Cavanagh, & Philiastides, 2023). These two systems often generate the same predictions to facilitate efficient learning. However, they may interfere with each other in some cases and lead to sub-optimal choices (Guitart-Masip et al., 2012). These sub-optimalities are usually driven by two interdependent axes of behavioral control: valence (i.e., reward or punishment) and action (i.e., approach or avoidance) (Guitart-Masip et al., 2014). Specifically, the Pavlovian system inherently favors approaching reward-related stimuli and avoiding punishment-related stimuli. The instrumental system, on the other hand, allows flexible action selection based on task rules, such as avoidance to obtain rewards (Queirazza, Steele, Krishnadas, Cavanagh, & Philiastides, 2023). Examining the influence of Pavlovian cues on instrumental behavior can advance our understanding of the formation of maladaptive behaviors (e.g., habitual drug seeking) in addictive disorders (Heinz et al., 2019). However, few experimental studies have directly tested whether and how this process goes awry in PSMU.

A popular paradigm that has been used to examine the effect of Pavlovian conditioning on instrumental behavior is the Pavlovian-to-instrumental transfer (PIT) task. Higher PIT effects suggest a greater impact of the Pavlovian system on instrumental behavior (Garbusow et al., 2022). Although studies using the PIT task have provided valuable information on how Pavlovian system biases instrumental behavior in addictive disorders (Qin et al., 2023; Vogel et al., 2018; Xu et al., 2024), this task separates Pavlovian and instrumental conditioning, and the interaction between the two systems is not tested directly during acquisition but in an extinction stage (Guitart-Masip et al., 2014). This is a crucial omission as the acquisition and extinction may rely on different neural and cognitive mechanisms (Sun et al., 2011). Furthermore, most PIT studies focused on the effects of Pavlovian cues on instrumental approach behavior (a few exceptions include Xu et al. (2024)), which limited a comprehensive investigation of the effects of valence and action on instrumental behavior (Chen et al., 2021). These issues can be addressed by the orthogonalized go/no-go task, in which valence and action are independently manipulated, such that the Pavlovian bias can interfere with or facilitate instrumental learning depending on the congruence of the two systems (Guitart-Masip et al., 2012).

More importantly, the orthogonalized go/no-go task allows for a computational analysis to examine the role of related latent cognitive factors (e.g., learning rate) during instrumental learning (Guitart-Masip et al., 2012). These computational models provide an effective approach for precisely investigating the cognitive processes that give rise to the behavioral characteristics of addiction (Gueguen, Schweitzer, & Konova, 2021; Perales, Flayelle, Verdejo-Garcia, Clark, & Billieux, 2024; Robinson, Chong, & Verdejo-Garcia, 2022). Although a few studies have reported learning and decision-making deficits in individuals with PSMU (Meshi et al., 2019; Müller et al., 2023), the computational modeling approach has been rarely used to explore whether and how different cognitive processes become dysfunctional during associative learning (Kato et al., 2023).

Therefore, we used an orthogonalized go/no-go paradigm to examine how Pavlovian biases influence instrumental learning in individuals with PSMU. We hypothesized that the PSMU group would show worse learning performance when Pavlovian and instrumental systems are incongruent in both reward and punishment domains. Using computational modeling, we further examined the latent cognitive processes underlying the between-group differences in learning performance and their associations with the model-free task performance and self-reported measures.

Material and methods

Participants

A total of 33 individuals with PSMU and 32 regular social media users were recruited from the local Universities to complete the experiment. PSMU severity was assessed using the Bergen Social Networking Addiction Scale (BSNAS; Andreassen, Pallesen, & Griffiths, 2017). All participants had normal or corrected-to-normal vision, no history of neurological or psychiatric disorders, and were free from current psychoactive medication use. The sample size is similar to previous studies using the orthogonalized go/no-go task and computational modeling (Guitart-Masip et al., 2012; Kim et al., 2023; Moutoussis et al., 2018) and gives us 80% power to detect a large between-group effect (e.g., Cohen's d = 0.75). We recruited regular social media users with low PSMU severity as the healthy control (HC) group, since we are particularly interested in the impact of PSMU, but not general time spent on social media, on associative learning. Information related to the internet use, including online video clip and live watching/listening time, online social communicating time, and online gaming time, was also measured. All participants were compensated 30 RMB for their time. Moreover, to motivate participants to perform the task carefully, they were told that they would receive a bonus (0-6 RMB) based on the accumulated points earned on the task.

Measures

The BSNAS was used to measure PSMU severity, and a cut-off threshold of 19 was used to separate the PSMU and HC groups as suggested by previous research (Bányai et al., 2017). The BSNAS is a six item self-report assessment of the usage of social media (Andreassen et al., 2017). Participants rated all items on a 5-point Likert scale. Total scores ranged from 6 to 30. The BSNAS has been widely used in previous PSMU studies and has shown good psychometric properties (Andreassen et al., 2017; Bányai et al., 2017).

Moreover, a subset of participants (n = 46, including 27 PSMU and 19 HC participants) completed the Barratt Impulsiveness Scale version 11 (BIS-11; Patton, Stanford, & Barratt, 1995) and the Behavioral Inhibition System/Behavioral Approach System (BIS/BAS; Carver & White, 1994) scales. The BIS-11 consists of 30 items on a 4-point Likert scale. The total score of the BIS-11 was used as an index of self-reported impulsivity. The BIS/BAS scales include 20 items on a 4-point Likert scale. The scores of the BIS and BAS scales were used as indices of self-reported sensitivity to reward and punishment, respectively.

Procedure

Orthogonalized go/no-go task

The orthogonalized go/no-go task was adapted from a probabilistic reinforcement learning paradigm (Cavanagh, Eisenberg, Guitart-Masip, Huys, & Frank, 2013; Guitart-Masip et al., 2012). This task included two orthogonal factors: valence (reward or punishment) and action (go or no-go), which formed four conditions: 1) go to win reward (GW), 2) no-go to win reward (NGW), 3) go to avoid punishment (GA), and 4) no-go to avoid punishment (NGA). On each trial (Fig. 1A), participants were presented with one of the four cues, each of which predicted a unique combination of the optimal action (go or no-go) and potential outcome (reward or punishment). Participants were not explicitly told the action contingencies for each cue. Thus, they need to explore both actions (i.e., respond to the target or withhold a response) to learn how to achieve the best outcome for each cue. Since the valence-action associations in the GW and NGA conditions were consistent with Pavlovian pre-programmed behavioral tendency, these two conditions were Pavlovian-congruent. In contrast, the GA and NGW conditions were Pavlovian-incongruent (Kim et al., 2023).

Fig. 1.

Task design overview. (A) A schematic illustration of the orthogonalized go/no-go task. Four types of stimuli were presented. Two conditions were Pavlovian-congruent: go action to win reward and no-go action to avoid punishment. The other two conditions were Pavlovian-incongruent: go action to avoid punishment and no-go action to win reward. (B) Valence-action contingencies in the four experimental conditions. The optimal action (go or no-go), possible outcomes, and their probabilities after a correct response to the target (in each cell)

Citation: Journal of Behavioral Addictions 2025; 10.1556/2006.2025.00026

As shown in Fig. 1A, each trial began with the display of a cue followed by a variable interval. Then a white circle as target appeared in the middle of the screen that required the participants to respond (go or no-go). After the offset of the circle, the feedback was presented indicating a reward outcome (green +$), a punishment outcome (red -$), or a neutral outcome (a yellow bar indicating no reward or punishment). A fixation cross with a variable interval was presented as the intertrial interval. Similar to many other RL tasks, the feedback of the current task was probabilistic. Specifically, on reward trials, 75% of correct responses and 25% of incorrect responses were rewarded (and the remaining 25% of correct and 75% of incorrect responses led to the neutral outcome). On punishment trials, 75% of correct responses and 25% of incorrect responses avoided losing money (Fig. 1B).

The task included three blocks, each with fifteen trials per experimental condition (GW, NGW, GA, NGA). Therefore, there were a total of 180 trials with 45 trials for each condition. The trials were randomly presented within each block. A practice session was conducted before the formal task to help participants to get familiar with the task.

Behavioral data analysis

The behavioral data were analyzed using R (version 4.3.2; https://cran.r-project.org/). Between-group comparisons of demographic and clinical characteristics were conducted using independent-sample t-tests or chi-squared tests.

For the task performance, a mixed ANOVA on response accuracy was conducted with valence (reward/punishment) and action (go/no-go) as within-subject factors, and group (PSMU/HC) as a between-subject factor.

Similar to previous studies (Kim et al., 2023), we have also calculated two Pavlovian bias indices (PBI), defined as the difference between the accuracy of Pavlovian-congruent and Pavlovian-incongruent conditions in the reward and punishment domains, respectively:

P B I (r e w a r d) = A c c u r a c y (g o t o w i n) - A c c u r a c y (n o g o t o w i n)

P B I (p u n i s h m e n t) = A c c u r a c y (n o g o t o a v o i d) - A c c u r a c y (g o t o a v o i d)

Computational modeling construction

Five nested reinforcement learning models used in previous research have been applied to examine latent parameters thought to underlie individual differences in behavioral performance (Guitart-Masip et al., 2012). For example, Model 5 (the best model of the current study) assigned each action

a

associated with a stimulus

s

on trial t a probability based on a weight function

W_{t} (a, s)

. A softmax choice function was used to predict the most likely action on each trial (Kim et al., 2023).

p (a_{1} | s) = {\frac{\exp (W_{t} (a_{1}, s))}{\exp (W_{t} (a_{1}, s)) + \exp (W_{t} (a_{2}, s))}} (1 - ξ) + \frac{ξ}{2}

(1)

Here, the probability of a go action was estimated using a squashed softmax function and based on the weights for go (

a

₁) and no-go (

a

₂). The parameter

ξ

was the irreducible noise (0 ≤

ξ

≤ 1), representing the randomness of action selection. The action weight function

W_{t} (a, s)

for the two reward conditions (GW and NGW) was defined as follows:

W_{t} (a, s) = {\begin{array}{c} Q_{t} (a, s) + b + {π_{r e w} V}_{t} (s) i f a = g o \\ Q_{t} (a, s) i f a = n o g o \end{array}

(2)

The parameter $b$ was a static bias for go action, which modified the action weight across valence. The parameter $π_{r e w}$ was the Pavlovian bias reflecting the behavioral approach to stimuli that had positive learned value. The action weight function $W_{t} (a, s)$ for the two punishment conditions (GA and NGA) was the same as Eq. (2) except that the parameter $π_{r e w}$ was replaced by $π_{p u n}$ , another Pavlovian bias parameter that reflecting behavioral inhibition for stimuli with negative learned value.

The

Q

values were updated according to a Rescorla-Wagner like rule. For the two reward conditions (GW and NGW), a reward sensitivity parameter

ρ_{r e w}

scaled the feedback value

r

and a learning rate parameter ε scaled the update term, as follows:

Q_{t} (a_{t}, s_{t}) = Q_{t - 1} (a_{t}, s_{t}) + ε (ρ_{r e w} r_{t} - Q_{t - 1} (a_{t}, s_{t}))

(3)

The expected value of each stimulus

V_{t}

was updated using a similar rule.

V_{t}

was updated according to the difference between current feedback

r_{t}

and previous expected value

V_{t - 1}

to the stimulus:

V_{t} (s_{t}) = V_{t - 1} (s_{t}) + ε (ρ_{r e w} r_{t} - V_{t - 1} (s_{t}))

(4)

For the two punishment conditions (GA and NGA), the

Q

and

V

values were updated in a similar manner, except that a punishment sensitivity parameter

ρ_{p u n}

scaled the feedback value

r

Q_{t} (a_{t}, s_{t}) = Q_{t - 1} (a_{t}, s_{t}) + ε (ρ_{p u n} r_{t} - Q_{t - 1} (a_{t}, s_{t}))

Eq(5)

V_{t} (s_{t}) = V_{t - 1} (s_{t}) + ε (ρ_{p u n} r_{t} - V_{t - 1} (s_{t}))

Eq(4)

The other four models are reduced versions of Model 5 (Fig. 3). For example, Model 4 used a single Pavlovian bias parameter $π$ for both the reward and punishment conditions; Model 3 used the same Pavlovian bias $π$ and feedback sensitivity $ρ$ for the reward and punishment conditions; Model 2 did not consider the Pavlovian bias; and Model 1 did not consider the Pavlovian bias and static bias for the go action.

Model parameter estimation

The model parameters were estimated using hierarchical Bayesian analysis (HBA), as it can generate more robust estimates and is less sensitive to outliers compared to other estimation methods, such as maximum likelihood estimation (Gelman & Shalizi, 2013). HBA was conducted based on Markov chain Monte Carlo (MCMC) algorithms as implemented using Stan (version 2.21; https://mc-stan.org/) and hBayesDM package in R (Ahn, Haines, & Zhang, 2017; Gelman & Shalizi, 2013). Similar to the models for the orthogonalized Go/no-go task in the hBayesDM library (Ahn et al., 2017), we used the weakly informative priors for all parameters (Table S2). The models for the PSMU and HC groups were fitted the separately to generate stable and reliable individual estimates (Kim et al., 2023). We used four independent chains and a sample size of 4,000, including 2,000 burn-in samples per chain. The convergence of the chains was confirmed by inspecting the Rhat values (Rhat <1.1) and the trace plots of four chains.

Model comparison

The leave-one-out information criterion (LOOIC) values were used to compare the models regarding their goodness-of-fit (Vehtari, Gelman, & Gabry, 2017). The LOOIC is a Bayesian criterion to evaluate the out-of-sample predictive performance of a model, and a lower LOOIC indicates a better model fit. The model weight, an index of the relative likelihood of a model calculated using Bayesian model averaging, was also calculated (Yao, Vehtari, Simpson, & Gelman, 2018). The loo package (version 2.7.0; https://cran.r-project.org/web/packages/loo/) in R was used to calculate the LOOIC and model weight for each model.

Since model 5 showed best model fitting for both PSMU and HC groups, we conducted the group comparison based on this model. Specifically, the posterior distribution of the HC group was subtracted from that of the PSMU group. Group differences were considered credible when the 90% highest density intervals (HDI) of posterior difference distributions did not include the value 0 (Kreis, Zhang, Moritz, & Pfuhl, 2022; Langley et al., 2023).

Ethics

The study procedures were carried out in accordance with the Declaration of Helsinki. The Institutional Review Board of the Sun Yat-sen University approved the study. All subjects were informed about the study and all provided informed consent.

Results

Demographic and behavioral results

The PSMU and HC groups were matched for age, gender, and education. As expected, the PSMU group showed significantly higher problematic social media use severity relative to the HC group. The PSMU group seemed to spend more time using social media than the HC group, although the difference was not significant (Table 1).

Table 1.

Demographic information, social media use characteristics, and task performance

Variables	Mean (SD)		t-test/χ²	p	Effect size
Variables	PSMU (n = 33)	HC (n = 32)	t-test/χ²	p	Effect size
Age	20.52 (1.60)	21.25 (1.77)	−1.75	0.09	d = 0.43
Gender (f/m)	26/7	20/12	1.37	0.24	V = 0.15
Education	14.91 (1.40)	15.09 (1.33)	−0.55	0.59	d = 0.13
BSMAS	22.91 (2.54)	13.19 (3.21)	13.57	<0.001	d = 3.36
Gaming time (hrs/week)	5.11 (7.45)	2.60 (2.90)	1.79	0.08	d = 0.44
Video and music use (hrs/week)	17.22 (13.37)	13.23 (12.56)	1.24	0.22	d = 0.39
Online social network (hrs/week)	19.32 (18.34)	14.28 (14.98)	1.47	0.21	d = 0.30
Impulsivity scale^a	72.07 (7.15)	67.42 (7.17)	2.17	0.04	d = 0.65
BIS scale^a	20.70 (2.80)	19.26 (3.35)	1.59	0.12	d = 0.48
BAS scale^a	40.33 (4.14)	41.74 (5.58)	−0.98	0.33	d = 0.30
PBI - reward domain	0.50 (0.36)	0.28 (0.41)	2.39	0.02	d = 0.57
PBI - punishment domain	−0.04 (0.22)	−0.02 (0.21)	−0.29	0.77	d = 0.09

BSMAS: Bergen social media addiction scale; BIS: Behavioral Inhibition System scale; BAS: Behavioral Approach System scale; PBI: Pavlovian bias index. PSMU: problematic social media use; HC: healthy control.

^a A subset of 46 participants (19 HC and 27 PSMU) completed the scales.

A mixed ANOVA on accuracy was also conducted with valence (win/lose) and action (go/no-go) as within-subject factors, and group (PSMU/HC) as a between-subject factor. We found that the interaction effect of valence by action by group (F_(1,63) = 3.97, p = 0.051, partial η² = 0.06) was marginal significant. Moreover, the interaction effects of valence by group (F_(1,63) = 4.25, p = 0.043, partial η² = 0.06), and action by group (F_(1,63) = 4.78, p = 0.032, partial η² = 0.07) were both significant. The simple effect analysis showed that the PSMU group performed significantly worse than the HC group in the no-go to win condition (F_(1,63) = 7.66, p = 0.007, partial η² = 0.11; Fig. 2A); however, the between-group difference was insignificant in the other three conditions.

Fig. 2.

Model-free results of the orthogonalized go/no-go task. (A) The PSMU group showed decreased accuracy in the no-go to win condition, reflecting a higher Pavlovian bias in the reward domain compared with the HC group. (B) The PSMU group showed a higher behavioral Pavlovian bias index (congruent accuracy - incongruent accuracy) in the reward domain, but not in the punishment domain, compared with the HC group. (C) Learning curves across trials indicated that the two groups performed similarly in most conditions, except for the no-go to win condition.

Error bars indicate standard error of mean

Citation: Journal of Behavioral Addictions 2025; 10.1556/2006.2025.00026

Similar to the ANOVA findings, we found that the PSMU group showed a significantly higher Pavlovian bias index for the reward domain, but not the punishment domain, compared with the HC group (Table 1, Fig. 2B). The learning curves across trials indicated that the two groups performed similarly in most conditions, except for the no-go to win condition (Fig. 2C). Therefore, we turned to a computational modeling analysis to further elucidate the latent processes underlying the condition-specific learning characteristics and between-group differences based on the reinforcement learning theory.

Computational modeling results

Reinforcement learning modeling was proceeded according to previous research (Guitart-Masip et al., 2012; Kim et al., 2023). We used five parameterized reinforcement learning models (Fig. 3A) to fit the behavioral data of the participants, and model comparison was conducted separately for each group. Based on the LOOIC values, Model 5 (with separate outcome sensitivity and Pavlovian bias parameters for the reward and punishment domains) showed the best fitting (lowest LOOIC values) for both groups (Fig. 3A).

Fig. 3.

Computational modeling results of the orthogonalized go/no-go task. (A) Model comparison showed that Model 5 was the best for both the HC and PSMU groups. Lower LOOIC values indicated better model performance. ξ: irreducible noise; ε: learning rate; ρ: sensitivity to outcomes; ρ_rew: sensitivity to reward outcomes; ρ_pun: sensitivity to punishment outcomes; b: go bias; π: Pavlovian bias. π_rew: Pavlovian bias in the reward domain; π_pun: Pavlovian bias in the punishment domain. (B–C) The PSMU group showed a higher learning rate and lower reward sensitivity compared with the HC group. Posterior distributions of the group-level parameters were taken from Model 5

Citation: Journal of Behavioral Addictions 2025; 10.1556/2006.2025.00026

Since the learning characteristics of the PSMU and HC groups during the orthogonalized go/no-go task were best captured by the same model, we thus calculated the posterior distributions and performed between-group comparison on all group-level parameters of Model 5. As shown in Fig. 3C, the PSMU group showed higher learning rate (ε) and lower reward sensitivity (ρ_rew). No credible between-group differences were found in the other five parameters (Table 2). The posterior distributions of these five group-level parameters can be found in Fig. S1.

Table 2.

Posterior mean (90% highest density interval) of group-level parameters in the PSMU and HC groups

Parameters	PSMU	HC	Difference
Irreducible noise (ξ)	0.065 [0.041, 0.094]	0.079 [0.051, 0.115]	−0.016 [−0.057, 0.026]
Learning rate (ε)	0.127 [0.087, 0.178]	0.065 [0.037, 0.106]	0.061 [0.003, 0.118]
Go bias (b)	0.894 [0.697, 1.110]	0.854 [0.620, 1.120]	0.037 [−0.300, 0.345]
Pavlovian bias - reward (π_rew)	0.202 [0.039, 0.401]	0.053 [−0.230, 0.362]	0.151 [−0.202, 0.494]
Pavlovian bias - punishment (π_pun)	0.431 [0.268, 0.595]	0.428 [0.234, 0.636]	0.002 [−0.272, 0.243]
Sensitivity - reward (ρ_rew)	8.770 [6.490, 12.200]	22.400 [16.000, 32.800]	−14.170 [−22.756, −5.182]
Sensitivity - punishment (ρ_pun)	5.460 [4.230, 7.100]	8.380 [6.160, 11.800]	−3.077 [−6.209, 0.070]

The credible differences between the PSMU and HC groups are indicated in bold font.

The current study focused on the between-group comparison on group-level parameters. Another analytical approach is to treat all participants as a single group and examine the relationship between the parameter values and PSMU severity (i.e., BSNAS scores) across all participants. To examine whether our findings were generally consistent across these two analytical approaches, we also fitted Model 5 to all participants and performed post-hoc correlation analyses between the PSMU severity and model parameter values and model-free measures. We found that the PSMU severity was negatively associated with reward sensitivity (Spearman's $ρ$ ₍₆₃₎ = −0.66, p < 0.001) and punishment sensitivity ( $ρ$ ₍₆₃₎ = −0.27, p = 0.03). The relationship between PSMU severity and learning rate was insignificant ( $ρ$ ₍₆₃₎ = 0.15, p = 0.22). All other associations were negligible (Table S3). For the model-free indices, the PSMU severity was marginally associated with the accuracy of the NGW condition ( $ρ$ ₍₆₃₎ = −0.23, p = 0.07; Table S4).

Correlation between computational, behavioral, and self-reported measures

Finally, we explored the relationship between the key parameter values (i.e., learning rate, reward sensitivity, and punishment sensitivity), model-free task performance (i.e., accuracy of the two incongruent conditions), and self-reported measures in all participants. As shown in Table S5, we found that the reward sensitivity parameter values correlated negatively with impulsivity ( $ρ$ ₍₄₄₎ = −0.30, p = 0.04) and positively with BAS scores ( $ρ$ ₍₄₄₎ = 0.31, p = 0.04). The associations between the accuracy of the two incongruent conditions and the self-reported measures were not significant (Table S6). Moreover, we found that the reward sensitivity was marginally correlated with the accuracy of the NGW condition ( $ρ$ ₍₆₃₎ = 0.24, p = 0.05). The learning rate was positively associated with the accuracy of the NGW condition ( $ρ$ ₍₆₃₎ = 0.26, p = 0.04) and the GA condition ( $ρ$ ₍₆₃₎ = 0.43, p < 0.001), although these effects were mainly driven by the HC group ( $ρ$ ₍₃₀₎ = 0.51, p = 0.003 for NGW and $ρ$ ₍₃₀₎ = 0.46, p = 0.008 for GA) instead of the PSMU group ( $ρ$ ₍₃₁₎ = 0.12, p = 0.49 for NGW and $ρ$ ₍₃₁₎ = 0.39, p = 0.03 for GA). More details on the associations between the key parameter values and task performance can be found in Table S6.

Discussion

This study used a combination of an orthogonalized go/no-go task and computational modeling to examine how individuals with PSMU were influenced by Pavlovian bias during instrumental learning. We found that: 1) the PSMU group exhibited a higher Pavlovian bias index in the reward domain than the HC group; 2) the computational modeling results showed an increased learning rate and a decreased reward sensitivity in the PSMU group than the HC group. Taken together, the study provides new insights into how Pavlovian bias influences instrumental learning and the underlying computational mechanisms in individuals with PSMU.

In the orthogonalized go/no-go task, two interdependent axes of behavioral control, valence and action, were independently manipulated, which allowed us to comprehensively examine the interaction between Pavlovian and instrumental systems during associative learning when they are congruent or in conflict (Guitart-Masip et al., 2014). Our findings suggest that the instrumental learning in the reward domain was overly impacted by the Pavlovian pre-programmed behavioral tendency in the PSMU group, in parallel with studies using the PIT task that found enhanced specific PIT effects in IGD participants (Qin et al., 2023; Vogel et al., 2018; Xu et al., 2024). For example, a recent study revealed that individuals with IGD showed fewer button presses to cues with negative values and more button presses to cues with positive values in Go trials (Xu et al., 2024). Although the evidence based on the same paradigm is still needed, these findings suggest that individuals with different types of problematic usage of the internet (e.g., PSMU and IGD) may overly rely on the Pavlovian system, which may interfere with the use of flexible instrumental learning.

Furthermore, previous studies in SUDs have also shown a higher Pavlovian bias or habitual learning in general at the expense of goal-directed instrumental learning (Ersche et al., 2016; Sebold et al., 2014). For example, a stronger PIT effect has been reported to be closely associated with alcohol use disorder severity (Garbusow et al., 2022), and the behavioral deficits may be linked to a decreased lateral prefrontal cortex activity and an increased ventral striatum activity (Chen et al., 2021). Taken together, these findings suggest that an over-reliance of Pavlovian system during associative learning seems to be a key characteristic of addiction in general (Chen et al., 2021; Garbusow et al., 2022). However, it should be noted that demographic (e.g., age and medication status) and methodological (e.g., paradigms and analyses) differences between SUD and PSMU studies warrant careful consideration. Direct comparisons based on matched samples and the same paradigm are still needed for future research.

To more precisely investigate latent factors that may contribute to associative learning during the orthogonalized go/no-go task, we tested five nested computational models (Guitart-Masip et al., 2012; Kim et al., 2023). The model comparison results indicated that the same model showed the best fit for both the PSMU and HC groups. Except for common parameters, such as learning rate and static action bias, this model also distinguished between outcome sensitivity and Pavlovian bias parameters related to reward and punishment. Altered reward processing has been highlighted as a central feature of different types of addictive disorders (Brand et al., 2019; Dong & Potenza, 2014; Li et al., 2020; Yao et al., 2017, 2022). The decreased reward sensitivity and blunted reward prediction error signals in the reward system at the neural level have been reported in SUDs and IGD (Lei et al., 2022; Rose et al., 2014; Tanabe et al., 2013). Our study extends this important question to the context of PSMU and suggests that insensitivity to monetary rewards may play a critical role in associative learning deficits in PSMU.

In addition to reward sensitivity, our modeling results also showed an increased learning rate in the PSMU group. Learning rate is critical for goal-directed behavior because it needs to be adapted to environmental changes (O’Reilly, 2013). As a computational measure, the goodness of learning can be considered as an index of how well people adapt their learning rate to the changing environment (Lockwood et al., 2020; Scholl & Klein-Flügge, 2018). Importantly, another recent reinforcement learning study in other addictions (i.e., IGD and SUDs) also identified an increased learning rate (Kwon, Choi, Park, Ahn, & Jung, 2024). It should be noted, however, that the optimal learning rate depends on the nature of the task and may vary between studies (Lockwood et al., 2020). According to the winning model of the current study, the reward sensitivity and learning rate together control learning from the previous feedback (Eq. (3)). Therefore, one possibility is that the PSMU group increased the learning rate to counteract the negative effects of the blunted reward sensitivity on feedback learning. However, such compensation may be ineffective in situations that require more flexible behavior (e.g., the NGW condition).

While researchers have devoted much effort to applying theoretical models of addiction to PSMU, few empirical studies have directly tested whether the hypotheses from these theories still hold in PSMU (Wegmann et al., 2020). The current study partially addressed this research gap by examining several critical factors included in the theoretical models of addiction. For example, in line with the influential I-PACE model (Brand et al., 2019; Wegmann et al., 2020), we found that individuals with PSMU showed poorer performance in a task that requires the interaction of several affective, cognitive, and executive processes (e.g., reward expectation, decision-making, and feedback learning). Beyond the framework of the original I-PACE model, our study further elucidated how some latent factors, including reward sensitivity and learning rate, interact with each other using a computational modeling approach. Moreover, the poor performance of the PSMU group in the NGW condition may not be due to a general impairment of inhibitory control, since they showed similar performance in the GA and NGA conditions and three bias values from the computational model compared with the HC group. However, it should be noted that this evidence is rather indirect, as this task is not designed to specifically measure inhibitory control. Therefore, a promising future direction is to incorporate multiple paradigms to investigate whether and how the general or PSMU-specific inhibitory control may influence the decision-making and learning, as suggested by the I-PACE model (Brand et al., 2019).

The findings of the current study may also shed new light on the development of tailored interventions for PSMU. For example, the motor response training (also based on a go/no-go paradigm) may be used to reduce the impact of stimulus-outcome associations preprogrammed in the Pavlovian system on instrumental learning (Chen, Veling, Dijksterhuis, & Holland, 2016), although its efficacy in alleviating PSMU symptoms has not been tested yet. In addition, neuromodulatory methods, such as non-invasive brain stimulation or real-time neurofeedback, may be used to change instrumental learning and related cognitive processes (Kahnt, 2023). For example, anodal transcranial direct current stimulation (tDCS) over the dorsolateral prefrontal cortex has been shown to reduce Pavlovian bias in the orthogonalized go/no-go task (Kim et al., 2023). However, the effect seems to be stronger in the punishment domain, but not in the reward domain. Therefore, future studies should further explore interventions that can reduce reward-related Pavlovian bias and PSMU severity.

Despite the above advantages, this study has some limitations. First, we only assessed the learning performance at the behavioral level. The interaction between valence and action has been suggested to be under the modulation of the dopaminergic system (Guitart-Masip et al., 2012). Future work to link these findings to potential neuromodulatory measures is needed. Second, although the sample size of the current study is similar to many other computational modeling studies (Hoven, Luigjes, & van Holst, 2024; Kwon et al., 2024; Wiehler, Chakroun, & Peters, 2021) and meets the basic requirement for most parametric analyses, future studies with large sample sizes are still needed to confirm the findings of this study. Particularly, only a subset of participants completed the questionnaires, so the related correlation results should be considered preliminary. Third, we did not find between-group differences related to reinforcement learning in the punishment domain, possibly because the monetary losses used in our task did not induce strong punishment effects. Studies using more effective punishment manipulation (e.g., an electric shock) are still needed to investigate the influence of Pavlovian bias on instrumental learning in the punishment domain in PSMU. Moreover, we did not include other cognitive tasks in the current study. It would be interesting to examine the relationship between performance on the orthogonalized go/no-go task (including modeling results) and basic executive functions such as working memory and general inhibitory control in individuals with PSMU. Finally, this study used young adults with high severity of PSMU. Future studies are encouraged to test if our findings could be generalized to clinical samples.

Conclusions

In summary, this study systematically investigated the impact of Pavlovian pre-programmed behavioral tendencies on instrumental learning in PSMU by dissociating the roles of valence and action. Individuals with PSMU showed worse learning performance, especially when the Pavlovian and instrumental systems were in conflict. Computational modeling further showed an increased learning rate and a blunted reward sensitivity in the PSMU group. These findings advance our understanding of how Pavlovian pre-programmed behavioral tendencies bias instrumental learning in PSMU, shedding new light on extending classical addiction theory to the field of PSMU and developing more effective interventions for this disorder.

Funding sources

This work was supported by the National Science Foundation of China (No. 32200910 and No. 32400918); the Humanities and Social Sciences Research Project from the Ministry of Education of China (No. 22YJC190016); the STI 2030-Major Projects (No. 2021ZD0200500); the Funding by Science and Technology Projects in Guangzhou (No. SL2023A04J00648); and the Early Career Scheme of the Hong Kong Research Grants Council (No. 27618524).

Authors' contribution

LL and Y-WY designed the study. LL, Y-XP, Z-HS, S-JC, and Y-YH collected the data. LL and Y-WY performed analysis. LL administrated the project. LL and Y-WY wrote the manuscript with the input from all the other authors.

Conflict of interest

The authors have no conflict of interests.

Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1556/2006.2025.00026.

References

Ahn, W.-Y., Haines, N., & Zhang, L. (2017). Revealing neurocomputational mechanisms of reinforcement learning and decision-making with the hBayesDM package. Computational Psychiatry, 1(0). https://doi.org/10.1162/cpsy_a_00002.
- Search Google Scholar
- Export Citation
Andreassen, C. S., Pallesen, S., & Griffiths, M. D. (2017). The relationship between addictive use of social media, narcissism, and self-esteem: Findings from a large national survey. Addictive Behaviors, 64, 287–293. https://doi.org/10.1016/j.addbeh.2016.03.006.
- Search Google Scholar
- Export Citation
Bányai, F., Zsila, Á., Király, O., Maraz, A., Elekes, Z., Griffiths, M. D., … Demetrovics, Z. (2017). Problematic social media use: Results from a large-scale nationally representative adolescent sample. PloS One, 12(1), e0169839. https://doi.org/10.1371/journal.pone.0169839.
- Search Google Scholar
- Export Citation
Brand, M., Rumpf, H. J., Demetrovics, Z., Müller, A., Stark, R., King, D. L., … Potenza, M. N. (2020). Which conditions should be considered as disorders in the International Classification of Diseases (ICD-11) designation of “other specified disorders due to addictive behaviors”. Journal of Behavioral Addictions, 11(2), 150–159. https://doi.org/10.1556/2006.2020.00035.
- Search Google Scholar
- Export Citation
Brand, M., Wegmann, E., Stark, R., Müller, A., Wölfling, K., Robbins, T. W., & Potenza, M. N. (2019). The Interaction of Person-Affect-Cognition-Execution (I-PACE) model for addictive behaviors: Update, generalization to addictive behaviors beyond internet-use disorders, and specification of the process character of addictive behaviors. Neuroscience and Biobehavioral Reviews, 104, 1–10. https://doi.org/10.1016/j.neubiorev.2019.06.032.
- Search Google Scholar
- Export Citation
Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: the BIS/BAS scales. Journal of Personality and Social Psychology, 67(2), 319. https://doi.org/10.1037/0022-3514.67.2.319.
- Search Google Scholar
- Export Citation
Cataldo, I., Billieux, J., Esposito, G., & Corazza, O. (2022). Assessing problematic use of social media: Where do we stand and what can be improved? Current Opinion in Behavioral Sciences, 45, 101145. https://doi.org/10.1016/j.cobeha.2022.101145.
- Search Google Scholar
- Export Citation
Cavanagh, J. F., Eisenberg, I., Guitart-Masip, M., Huys, Q., & Frank, M. J. (2013). Frontal theta overrides pavlovian learning biases. Journal of Neuroscience, 33(19), 8541–8548. https://doi.org/10.1523/JNEUROSCI.5754-12.2013.
- Search Google Scholar
- Export Citation
Chen, H., Nebe, S., Mojtahedzadeh, N., Kuitunen-Paul, S., Garbusow, M., Schad, D. J., … Smolka, M. N. (2021). Susceptibility to interference between Pavlovian and instrumental control is associated with early hazardous alcohol use. Addiction Biology, 26(4), e12983. https://doi.org/10.1111/adb.13263.
- Search Google Scholar
- Export Citation
Chen, Z., Veling, H., Dijksterhuis, A., & Holland, R. W. (2016). How does not responding to appetitive stimuli cause devaluation: Evaluative conditioning or response inhibition? Journal of Experimental Psychology: General, 145(12), 1687. https://doi.org/10.1037/xge0000236.
- Search Google Scholar
- Export Citation
Dong, G., & Potenza, M. N. (2014). A cognitive-behavioral model of Internet gaming disorder: Theoretical underpinnings and clinical implications. Journal of Psychiatric Research, 58, 7–11. https://doi.org/10.1016/j.jpsychires.2014.07.005.
- Search Google Scholar
- Export Citation
Ersche, K. D., Gillan, C. M., Jones, P. S., Williams, G. B., Ward, L. H. E., Luijten, M., … Robbins, T. W. (2016). Carrots and sticks fail to change behavior in cocaine addiction. Science, 352(6292), 1468–1471. https://doi.org/10.1126/science.aaf3700.
- Search Google Scholar
- Export Citation
Garbusow, M., Ebrahimi, C., Riemerschmid, C., Daldrup, L., Rothkirch, M., Chen, K., … Smolka, M. N. (2022). Pavlovian-to-instrumental transfer across mental disorders: A review. Neuropsychobiology, 81(5), 418–437. https://doi.org/10.1159/000525579.
- Search Google Scholar
- Export Citation
Gelman, A., & Shalizi, C. R. (2013). Philosophy and the practice of Bayesian statistics. British Journal of Mathematical and Statistical Psychology, 66(1), 8–38. https://doi.org/10.1111/j.2044-8317.2011.02037.x.
- Search Google Scholar
- Export Citation
Gershman, S. J., Guitart-Masip, M., & Cavanagh, J. F. (2021). Neural signatures of arbitration between Pavlovian and instrumental action selection. PLoS Computational Biology, 17(2), e1008553. https://doi.org/10.1371/journal.pcbi.1008553.
- Search Google Scholar
- Export Citation
Gueguen, M. C. M., Schweitzer, E. M., & Konova, A. B. (2021). Computational theory-driven studies of reinforcement learning and decision-making in addiction: What have we learned? Current Opinion in Behavioral Sciences, 38, 40–48. https://doi.org/10.1016/j.cobeha.2020.08.007.
- Search Google Scholar
- Export Citation
Guitart-Masip, M., Duzel, E., Dolan, R., & Dayan, P. (2014). Action versus valence in decision making. Trends in Cognitive Sciences, 18(4), 194–202. https://doi.org/10.1016/j.tics.2014.01.003.
- Search Google Scholar
- Export Citation
Guitart-Masip, M., Huys, Q. J. M., Fuentemilla, L., Dayan, P., Duzel, E., & Dolan, R. J. (2012). Go and no-go learning in reward and punishment: Interactions between affect and effect. Neuroimage, 62(1), 154–166. https://doi.org/10.1016/j.neuroimage.2012.04.024.
- Search Google Scholar
- Export Citation
Heinz, A., Beck, A., Halil, M. G., Pilhatsch, M., Smolka, M. N., & Liu, S. (2019). Addiction as learned behavior patterns. Journal of Clinical Medicine, 8(8), 1086. https://doi.org/10.3390/jcm8081086.
- Search Google Scholar
- Export Citation
Hoven, M., Luigjes, J., & van Holst, R. J. (2024). Learning and metacognition under volatility in GD: Lower learning rates and distorted coupling between action and confidence. Journal of Behavioral Addictions, 13(1), 226–235. https://doi.org/10.1556/2006.2023.00082.
- Search Google Scholar
- Export Citation
Kahnt, T. (2023). Computationally informed interventions for targeting compulsive behaviors. Biological Psychiatry, 93(8), 729–738. https://doi.org/10.1016/j.biopsych.2022.08.028.
- Search Google Scholar
- Export Citation
Kato, A., Shimomura, K., Ognibene, D., Parvaz, M. A., Berner, L. A., Morita, K., & Fiore, V. G. (2023). Computational models of behavioral addictions: State of the art and future directions. Addictive Behaviors, 140, 107595. https://doi.org/10.1016/j.addbeh.2022.107595.
- Search Google Scholar
- Export Citation
Kim, H., Hur, J. K., Kwon, M., Kim, S., Zoh, Y., & Ahn, W.-Y. (2023). Causal role of the dorsolateral prefrontal cortex in modulating the balance between Pavlovian and instrumental systems in the punishment domain. Plos One, 18(6), e0286632. https://doi.org/10.1371/journal.pone.0295458.
- Search Google Scholar
- Export Citation
Kreis, I., Zhang, L., Moritz, S., & Pfuhl, G. (2022). Spared performance but increased uncertainty in schizophrenia: Evidence from a probabilistic decision-making task. Schizophrenia Research, 243, 414–423. https://doi.org/10.1016/j.schres.2021.06.038.
- Search Google Scholar
- Export Citation
Kwon, M., Choi, H., Park, H., Ahn, W.-Y., & Jung, Y.-C. (2024). Neural correlates of model-based behavior in internet gaming disorder and alcohol use disorder. Journal of Behavioral Addictions, 13(1), 236–249. https://doi.org/10.1556/2006.2024.00006.
- Search Google Scholar
- Export Citation
Langley, C., Armand, S., Luo, Q., Savulich, G., Segerberg, T., Søndergaard, A., … Johansen, A. (2023). Chronic escitalopram in healthy volunteers has specific effects on reinforcement sensitivity: A double-blind, placebo-controlled semi-randomised study. Neuropsychopharmacology, 48(4), 664–670. https://doi.org/10.1038/s41386-022-01523-x.
- Search Google Scholar
- Export Citation
Lei, W., Liu, K., Chen, G., Tolomeo, S., Liu, C., Peng, Z., … Xiang, B. (2022). Blunted reward prediction error signals in internet gaming disorder. Psychological Medicine, 52(11), 2124–2133. https://doi.org/10.1017/S003329172000402X.
- Search Google Scholar
- Export Citation
Li, Q., Wang, Y., Yang, Z., Dai, W., Zheng, Y., Sun, Y., & Liu, X. (2020). Dysfunctional cognitive control and reward processing in adolescents with Internet gaming disorder. Psychophysiology, 57(2), e13469. https://doi.org/10.1111/psyp.13469.
- Search Google Scholar
- Export Citation
Lindström, B., Bellander, M., Schultner, D. T., Chang, A., Tobler, P. N., & Amodio, D. M. (2021). A computational reward learning account of social media engagement. Nature Communications, 12(1), 1311. https://doi.org/10.1038/s41467-020-19607-x.
- Search Google Scholar
- Export Citation
Liu, L., Yao, Y. W., Fang, X. Y., Xu, L. X., Hu, M. J., Zhang, J. T., & Potenza, M. N. (2024). Compulsivity-related behavioral features of problematic usage of the internet: A scoping review of paradigms, progress, and perspectives. Journal of Behavioral Addictions, 13(2), 429–449. https://doi.org/10.1556/2006.2024.00023.
- Search Google Scholar
- Export Citation
Lockwood, P. L., & Klein-Flügge, M. C. (2020). Computational modelling of social cognition and behaviour: A reinforcement learning primer. Social Cognitive and Affective Neuroscience, 16(8), 761–771. https://doi.org/10.1093/scan/nsaa040.
- Search Google Scholar
- Export Citation
Meshi, D., Elizarova, A., Bender, A., & Verdejo-Garcia, A. (2019). Excessive social media users demonstrate impaired decision making in the Iowa Gambling Task. Journal of Behavioral Addictions, 8(1), 169–173. https://doi.org/10.1556/2006.7.2018.138.
- Search Google Scholar
- Export Citation
Montag, C., Marciano, L., Schulz, P. J., & Becker, B. (2023). Unlocking the brain secrets of social media through neuroscience. Trends in Cognitive Sciences, 27(12), 1102–1104. https://doi.org/10.1016/j.tics.2023.09.005.
- Search Google Scholar
- Export Citation
Moutoussis, M., Rutledge, R. B., Prabhu, G., Hrynkiewicz, L., Lam, J., Ousdal, O.-T., … Dolan, R. J. (2018). Neural activity and fundamental learning, motivated by monetary loss and reward, are intact in mild to moderate major depressive disorder. PLoS One, 13(8), e0201451. https://doi.org/10.1371/journal.pone.0201451.
- Search Google Scholar
- Export Citation
Müller, S. M., Antons, S., Wegmann, E., Ioannidis, K., King, D. L., Potenza, M. N., … Brand, M. (2023). A systematic review and meta-analysis of risky decision-making in specific domains of problematic use of the internet: Evidence across different decision-making tasks. Neuroscience and Biobehavioral Reviews, 152, 105271. https://doi.org/10.1016/j.neubiorev.2023.105271.
- Search Google Scholar
- Export Citation
O’Reilly, J. X. (2013). Making predictions in a changing world - inference, uncertainty, and learning. Frontiers in Neuroscience, 7, 33773. https://doi.org/10.3389/fnins.2013.00105.
- Search Google Scholar
- Export Citation
Patton, J. H., Stanford, M. S., & Barratt, E. S. (1995). Factor structure of the Barratt impulsiveness scale. Journal of Clinical Psychology, 51(6), 768–774. https://doi.org/10.1002/1097-4679(199511)51:6<768::AID-JCLP2270510607>3.0.CO;2-1.
- Search Google Scholar
- Export Citation
Perales, J. C., Flayelle, M., Verdejo-Garcia, A., Clark, L., & Billieux, J. (2024). Cognitive, computational and learning processes involved in behavioral addiction: Current views and future directions. Addictive Behaviors, 148(107874). https://doi.org/10.1016/j.addbeh.2023.107874.
- Search Google Scholar
- Export Citation
Qin, C., Feng, S., Chen, Y., Liao, X., Cheng, X., Tian, M., … Gong, K. (2023). Enhanced Pavlovian-to-instrumental transfer in internet gaming disorder. Journal of Behavioral Addictions, 12(2), 471–479. https://doi.org/10.1556/2006.2023.00023.
- Search Google Scholar
- Export Citation
Queirazza, F., Steele, J. D., Krishnadas, R., Cavanagh, J., & Philiastides, M. G. (2023). Functional MRI signatures of Pavlovian and instrumental valuation systems during a modified orthogonalized go/no-go task. Journal of Cognitive Neuroscience, 35(12), 2089–2109. https://doi.org/10.1162/jocn_a_02062.
- Search Google Scholar
- Export Citation
Redish, A. D., & Johnson, A. (2007). A computational model of craving and obsession. Annals of the New York Academy of Sciences, 1104, 324–339. https://doi.org/10.1196/annals.1390.014.
- Search Google Scholar
- Export Citation
Robinson, A. H., Chong, T. T.-J., & Verdejo-Garcia, A. (2022). Computational models of exploration and exploitation characterise onset and efficacy of treatment in methamphetamine use disorder. Addiction Biology, 27(3), e13172. https://doi.org/10.1111/adb.13172.
- Search Google Scholar
- Export Citation
Rose, E. J., Salmeron, B. J., Ross, T. J., Waltz, J., Schweitzer, J. B., McClure, S. M., & Stein, E. A. (2014). Temporal difference error prediction signal dysregulation in cocaine dependence. Neuropsychopharmacology, 39(7), 1732–1742. https://doi.org/10.1038/npp.2014.21.
- Search Google Scholar
- Export Citation
Scholl, J., & Klein-Flügge, M. (2018). Understanding psychiatric disorder by capturing ecologically relevant features of learning and decision-making. Behavioural Brain Research, 355, 56–75. https://doi.org/10.1016/j.bbr.2017.09.050.
- Search Google Scholar
- Export Citation
Sebold, M., Deserno, L., Nebe, S., Schad, D. J., Garbusow, M., Hägele, C., … Smolka, M. (2014). Model-based and model-free decisions in alcohol dependence. Neuropsychobiology, 70(2), 122–131. https://doi.org/10.1159/000362840.
- Search Google Scholar
- Export Citation
Sun, N., Chi, N., Lauzon, N., Bishop, S., Tan, H., & Laviolette, S. R. (2011). Acquisition, extinction, and recall of opiate reward memory are signaled by dynamic neuronal activity patterns in the prefrontal cortex. Cerebral Cortex, 21(12), 2665–2680. https://doi.org/10.1093/cercor/bhr031.
- Search Google Scholar
- Export Citation
Sun, Y., & Zhang, Y. (2021). A review of theories and models applied in studies of social media addiction and implications for future research. Addictive Behaviors, 114, 106699. https://doi.org/10.1016/j.addbeh.2020.106699.
- Search Google Scholar
- Export Citation
Tanabe, J., Reynolds, J., Krmpotich, T., Claus, E., Thompson, L. L., Du, Y. P., & Banich, M. T. (2013). Reduced neural tracking of prediction error in substance-dependent individuals. American Journal of Psychiatry, 170(11), 1356–1363. https://doi.org/10.1176/appi.ajp.2013.12091257.
- Search Google Scholar
- Export Citation
Vehtari, A., Gelman, A., & Gabry, J. (2017). Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Statistics and Computing, 27(5), 1413–1432. https://doi.org/10.1007/S11222-016-9696-4.
- Search Google Scholar
- Export Citation
Vogel, V., Kollei, I., Duka, T., Snagowski, J., Brand, M., Müller, A., & Loeber, S. (2018). Pavlovian-to-instrumental transfer: A new paradigm to assess pathological mechanisms with regard to the use of Internet applications. Behavioural Brain Research, 347, 8–16. https://doi.org/10.1016/j.bbr.2018.03.009.
- Search Google Scholar
- Export Citation
Wegmann, E., Müller, S. M., Turel, O., & Brand, M. (2020). Interactions of impulsivity, general executive functions, and specific inhibitory control explain symptoms of social-networks-use disorder: An experimental study. Scientific Reports, 10(1), 3866. https://doi.org/10.1038/s41598-020-60819-4.
- Search Google Scholar
- Export Citation
Wiehler, A., Chakroun, K., & Peters, J. (2021). Attenuated directed exploration during reinforcement learning in gambling disorder. Journal of Neuroscience, 41(11), 2512–2522. https://doi.org/10.1523/JNEUROSCI.1607-20.2021.
- Search Google Scholar
- Export Citation
Xu, L., Zhang, J., Geng, X., Song, K., Zeng, P., Potenza, M. N., … Zhang, J. (2024). Pavlovian-to-instrumental transfer and outcome-devaluation effects in individuals with gaming experience. Computers in Human Behavior, 155, 108188. https://doi.org/10.1016/j.chb.2024.108188.
- Search Google Scholar
- Export Citation
Yao, Y.-W., Liu, L., Ma, S.-S., Shi, X.-H., Zhou, N., Zhang, J.-T., & Potenza, M. N. (2017). Functional and structural neural alterations in Internet gaming disorder: A systematic review and meta-analysis. Neuroscience & Biobehavioral Reviews, 83, 313–324. https://doi.org/10.1016/j.neubiorev.2017.10.029.
- Search Google Scholar
- Export Citation
Yao, Y., Vehtari, A., Simpson, D., & Gelman, A. (2018). Using stacking to average Bayesian predictive distributions (with discussion). Bayesian Analysis, 13(3), 917–1003. https://doi.org/10.1214/17-BA1091.
- Search Google Scholar
- Export Citation
Yao, Y.-W., Zhang, J.-T., Fang, X.-Y., Liu, L., & Potenza, M. N. (2022). Reward-related decision-making deficits in internet gaming disorder: A systematic review and meta-analysis. Addiction, 117(1), 19–32. https://doi.org/10.1111/add.15518.
- Search Google Scholar
- Export Citation

Supplementary Materials

ProCite

RefWorks

Reference Manager

BibTeX

Zotero

EndNote

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://akjournals.com/view/journals/2006/aop/article-10.1556-2006.2025.00026/article-10.1556-2006.2025.00026.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Collapse
Expand

Journal of Behavioral Addictions

Print ISSN:: 2062-5871

Online ISSN:: 2063-5303

Dr. Zsolt Demetrovics
Institute of Psychology, ELTE Eötvös Loránd University
Address: Izabella u. 46. H-1064 Budapest, Hungary
Phone: +36-1-461-2681
E-mail: jba@ppk.elte.hu

Indexing and Abstracting Services:

Web of Science [Science Citation Index Expanded (also known as SciSearch®)
Journal Citation Reports/Science Edition
Social Sciences Citation Index®
Journal Citation Reports/ Social Sciences Edition
Current Contents®/Social and Behavioral Sciences
EBSCO
GoogleScholar
PsycINFO
PubMed Central
SCOPUS
Medline
CABI
CABELLS Journalytics

2024
Web of Science
Journal Impact Factor	6.2
Rank by Impact Factor	Q1 (Psychiatry)
Journal Citation Indicator	1.63
Rank by Journal Citation Indicator	Q1 (Psychiatry)
Scopus
CiteScore	11.6
CiteScore rank	Q1 (Clinical Psychology)
SNIP	1.581
Scimago
SJR index	2.260
SJR Q rank	Q1

2023
Web of Science
Journal Impact Factor	6.6
Rank by Impact Factor	Q1 (Psychiatry)
Journal Citation Indicator	1.59
Scopus
CiteScore	12.3
CiteScore rank	Q1 (Clinical Psychology)
SNIP	1.604
Scimago
SJR index	2.188
SJR Q rank	Q1

Journal of Behavioral Addictions
Publication Model	Gold Open Access
Submission Fee	none
Article Processing Charge	990 EUR/article Effective from 1st Feb 2025: 1400 EUR/article
Regional discounts on country of the funding agency	World Bank Lower-middle-income economies: 50% World Bank Low-income economies: 100%
Further Discounts	Corresponding authors, affiliated to an EISZ member institution subscribing to the journal package of Akadémiai Kiadó: 100%.
Subscription Information	Gold Open Access

Journal of Behavioral Addictions
Language	English
Size	A4
Year of Foundation	2011
Volumes per Year	1
Issues per Year	4
Founder	Eötvös Loránd Tudományegyetem
Founder's Address	H-1053 Budapest, Hungary Egyetem tér 1-3.
Publisher	Akadémiai Kiadó
Publisher's Address	H-1117 Budapest, Hungary 1516 Budapest, PO Box 245.
Responsible Publisher	Chief Executive Officer, Akadémiai Kiadó
ISSN	2062-5871 (Print)
ISSN	2063-5303 (Online)

Senior editors

Editor(s)-in-Chief: Zsolt DEMETROVICS

Assistant Editor(s):

Csilla ÁGOSTON

Dana KATZ

Associate Editors

Stephanie ANTONS (Universitat Duisburg-Essen, Germany)
Joel BILLIEUX (University of Lausanne, Switzerland)
Beáta BŐTHE (University of Montreal, Canada)
Matthias BRAND (University of Duisburg-Essen, Germany)
Daniel KING (Flinders University, Australia)
Gyöngyi KÖKÖNYEI (ELTE Eötvös Loránd University, Hungary)
Ludwig KRAUS (IFT Institute for Therapy Research, Germany)
Marc N. POTENZA (Yale University, USA)
Hans-Jurgen RUMPF (University of Lübeck, Germany)
Ruth J. VAN HOLST (Amsterdam UMC, The Netherlands)

Editorial Board

Sophia ACHAB (Faculty of Medicine, University of Geneva, Switzerland)
Alex BALDACCHINO (St Andrews University, United Kingdom)
Judit BALÁZS (ELTE Eötvös Loránd University, Hungary)
Maria BELLRINGER (Auckland University of Technology, Auckland, New Zealand)
Henrietta BOWDEN-JONES (Imperial College, United Kingdom)
Damien BREVERS (University of Luxembourg, Luxembourg)
Julius BURKAUSKAS (Lithuanian University of Health Sciences, Lithuania)
Gerhard BÜHRINGER (Technische Universität Dresden, Germany)
Silvia CASALE (University of Florence, Florence, Italy)
Luke CLARK (University of British Columbia, Vancouver, B.C., Canada)
Jeffrey L. DEREVENSKY (McGill University, Canada)
Geert DOM (University of Antwerp, Belgium)
Nicki DOWLING (Deakin University, Geelong, Australia)
Hamed EKHTIARI (University of Minnesota, United States)
Jon ELHAI (University of Toledo, Toledo, Ohio, USA)
Ana ESTEVEZ (University of Deusto, Spain)
Fernando FERNANDEZ-ARANDA (Bellvitge University Hospital, Barcelona, Spain)
Naomi FINEBERG (University of Hertfordshire, United Kingdom)
Sally GAINSBURY (The University of Sydney, Camperdown, NSW, Australia)
Belle GAVRIEL-FRIED (The Bob Shapell School of Social Work, Tel Aviv University, Israel)
Biljana GJONESKA (Macedonian Academy of Sciences and Arts, Republic of North Macedonia)
Marie GRALL-BRONNEC (University Hospital of Nantes, France)
Jon E. GRANT (University of Minnesota, USA)
Mark GRIFFITHS (Nottingham Trent University, United Kingdom)
Joshua GRUBBS (University of New Mexico, Albuquerque, NM, USA)
Anneke GOUDRIAAN (University of Amsterdam, The Netherlands)
Susumu HIGUCHI (National Hospital Organization Kurihama Medical and Addiction Center, Japan)
David HODGINS (University of Calgary, Canada)
Eric HOLLANDER (Albert Einstein College of Medicine, USA)
Zsolt HORVÁTH (Eötvös Loránd University, Hungary)
Susana JIMÉNEZ-MURCIA (Clinical Psychology Unit, Bellvitge University Hospital, Barcelona, Spain)
Yasser KHAZAAL (Geneva University Hospital, Switzerland)
Orsolya KIRÁLY (Eötvös Loránd University, Hungary)
Chih-Hung KO (Faculty of Medicine, College of Medicine, Kaohsiung Medical University, Taiwan)
Shane KRAUS (University of Nevada, Las Vegas, NV, USA)
Hae Kook LEE (The Catholic University of Korea, Republic of Korea)
Bernadette KUN (Eötvös Loránd University, Hungary)
Katerina LUKAVSKA (Charles University, Prague, Czech Republic)
Giovanni MARTINOTTI (‘Gabriele d’Annunzio’ University of Chieti-Pescara, Italy)
Gemma MESTRE-BACH (Universidad Internacional de la Rioja, La Rioja, Spain)
Astrid MÜLLER (Hannover Medical School, Germany)
Daniel Thor OLASON (University of Iceland, Iceland)
Ståle PALLESEN (University of Bergen, Norway)
Afarin RAHIMI-MOVAGHAR (Teheran University of Medical Sciences, Iran)
József RÁCZ (Hungarian Academy of Sciences, Hungary)
Michael SCHAUB (University of Zurich, Switzerland)
Marcantanio M. SPADA (London South Bank University, United Kingdom)
Daniel SPRITZER (Study Group on Technological Addictions, Brazil)
Dan J. STEIN (University of Cape Town, South Africa)
Sherry H. STEWART (Dalhousie University, Canada)
Attila SZABÓ (Eötvös Loránd University, Hungary)
Hermano TAVARES (Instituto de Psiquiatria do Hospital das Clínicas da Faculdade de Medicina, Universidade de São Paulo, São Paulo, Brazil)
Wim VAN DEN BRINK (University of Amsterdam, The Netherlands)
Alexander E. VOISKOUNSKY (Moscow State University, Russia)
Aviv M. WEINSTEIN (Ariel University, Israel)
Anise WU (University of Macau, Macao, China)
Ágnes ZSILA (ELTE Eötvös Loránd University, Hungary)