Materials and Methods
We planned to include four groups of 50 participants, two groups of gamers and two groups of gamblers, depending on their frequency of gaming or gambling. Participants were invited to participate by email and were recruited from a registry of volunteers for research that was created by our research team or via social networks from July 2021 to April 2022.
To be included in the study, all participants were required to (1) be at least 18 years old and less than 65 years old and (2) have played their favourite game (gambling or VG) at least once during the 12 past months. The exclusion criteria were (1) difficulties reading or writing the French language, (2) having used a psychoactive substance in the past 12 h, (3) having a neurological history (such as epilepsy, severe brain injury, neurodegenerative disease), (4) having a diagnosis of a psychiatric or addictive disorder (with the exceptions of tobacco use disorder, gaming disorder and gambling disorder), (5) being pregnant or nursing, and (6) being under tutelage or curatorship.
To prevent individual participants from contributing to the study twice, we asked them whether they had already participated in the experiment before allowing them to access it.
The groups were created based on the frequency with which participants had engaged in their favourite behaviour during the past three months: high-frequency groups included participants who gambled or played a VG twice per week for at least 14 h, while low-frequency groups included participants who gambled or played a VG less than once per week. As a result, four groups were created: the low-frequency gambling group (LGB), the high-frequency gambling group (HGB), the low-frequency gaming group (LGM), and the high-frequency gaming group (HGM).
The website Psytoolkit (Stoet, 2010
) was used to host the whole experiment. Participants were informed about the study and consented to participate online before being allowed access to the experiment. They also answered questions regarding the inclusion criteria and their frequency of gaming/gambling over the past three months to define their inclusion group.
We then collected certain information regarding sociodemographic characteristics and gambling/gaming (favourite type of gaming or gambling, duration of playing/gambling at the relevant frequency, number of hours played per month on average).
After answering these questions, participants were required to complete online cognitive tasks. Finally, they completed self-report questionnaires exploring their levels of addiction severity with respect to either gaming or gambling, alexithymia, empathy, depression and anxiety, and the acceptability of the online assessment. We also asked participants whether they had prior knowledge of the cognitive tasks and questionnaires used.
We chose to assess SC via cognitive tasks before the completion of questionnaires to avoid the effects of exhaustion and attentional decrease that may have occurred after answering all the questions.
Sociodemographic and Gaming/Gambling Data
Sociodemographic data included age, sex, number of years of education, and living situation (alone, with family, with a partner). Gaming/gambling information included the choice of favourite game: (1) for the gambling groups, the participants could choose between poker, casino, lottery, horse race betting, sports betting, and other (free text field), while (2) for the gaming groups, the participants could choose between First-Person Shooter (FPS), Massively Multiplayer Online Role-Playing Game (MMORPG), Sandbox, Role Playing Game (RPG), Multiplayer Online Battle Arena (MOBA) and other (free text field). Participants were also requested to indicate the name of the game they played most frequently, the mean number of hours they played per month, and the durations for which they played at this frequency (in months).
The Canadian Problem Gambling Index (CPGI, (Ferris & Wynne, 2001
) was used to assess the addiction severity of gamblers (those who chose gambling as their favourite activity). This measure is a self-report questionnaire containing nine items regarding gambling habits during the past 12 months. Possible ratings include “never,“ “sometimes,“ “most of the time,“ and “nearly always”. Possible scores range from 0 to 27. This tool has good specificity (Ferris & Wynne, 2001
The Ten-item Internet Gaming Disorder Test (IGTD10) (Király et al., 2019
) was used to assess the addiction severity of gamers (those who chose gaming as their favourite activity). This tool contains ten items assessing nine criteria drawn from the DSM-5 (American Psychiatry Association, 2013
) and the ICD-11 (World Health Organization, 2018
) with respect to the past 12 months. Possible ratings include “never,“ “sometimes,“ and “often”. Possible scores range from 0 to 9. Research has indicated robust psychometric properties for this measure, which have been confirmed in several countries (King et al., 2020
; Király et al., 2019
Both questionnaires generated global scores that were used as continuous variables to indicate the level of addiction severity.
Level of Depression and Anxiety
The Hospital Anxiety and Depression Scale (HAD, (Zigmond & Snaith, 1983
) was used to assess levels of depression and anxiety; this measure included 14 items, which were rated from 0 to 3.
Social Cognition Assessment
Cognitive tasks appeared in the same order for all subjects, as indicated below.
Penn Emotion Recognition Task (ER-40) (Carter et al., 2009)
This task aimed to assess the ability to identify simple (basic) emotions. It included forty photographs of faces (20 women and 20 men; see (Gur et al., 2002
) for details regarding pictures) with respect to which the participant was required to choose between sadness, fear, joy, anger, or neutrality (for photographs not displaying any emotion). For each emotion, four high-intensity emotion pictures (with emotions expressed at a high intensity) and 4 low-intensity emotion pictures (with emotion expressed at a lower intensity) were included. The variables recorded were reaction times and the number of errors. Four randomized lists were generated to pseudo-randomize the order of presentation among participants.
Condensed and Revised Multifaced Empathy Test (MET-CORE) (Grynberg et al., 2017)
This task aimed to assess the ability to identify complex emotions (by presenting pictures with a context; the emotions proposed included basic emotions and more complex emotions) and the degree of emotion sharing (a component of empathy). It included forty photographs (20 positive and 20 negative), each of which was viewed twice by the participant. In one of the presentations, the participant was required to choose the correct emotion displayed by the person in the photograph from a set of four choices. During the other presentation, participants rated the level of emotion displayed by the person in the photograph from 1 (none at all) to 9 (very strong). Regarding performance indicators, the software recorded the number of errors and reaction times for the identification questions and the rating and reaction times for the sharing component. A French team translated this task from German into French (Edele et al., 2013
; Grynberg et al., 2017
), which led to altered emotion identification abilities but preserved emotion sharing in a population of patients with substance use disorder (Grynberg et al., 2017
). This task was used to complement the ER-40 because it permitted the assessment of facial emotion identification abilities in response to more complex stimuli and the subsequent assessment of this SC component at both levels (simple and complex facial emotion processing).
Social Decision-Making: The Chicken Game (CG) Paradigm
This task was an economic game assessing the level of cooperation. In each trial, participants were required to choose between two possible choices (i.e., a cooperative choice or a noncooperative choice). The outcome of the game depended on the participant’s own choice as well as on the choice of a partner in the game. In our study, the game partner was always the computer, but it was presented to the participant as either a computer (the “computer” condition) or another participant in the study (the “coplayer” condition). If both participants chose to cooperate, they both won ten points. If both participants defected, they each lost 30 points. If one cooperated and the other did not, the co-operator lost 10 points, and the noncooperator won 30 points. Matrices of payment were available to the participants throughout the task. Four trials were conducted: 2 for the “computer” condition and 2 for the “coplayer” condition (to preserve the credibility of a game played with another player, a waiting time of a few seconds was provided to simulate waiting for a connection from a coplayer). Choices (concerning whether to cooperate) made by the computer (either in the “computer” or the “coplayer” conditions) and the first condition to appear (“computer” or “coplayer”) were randomized. The software recorded choices and reaction times. In this task, the risk of defecting is more clearly identifiable than in the prisoner dilemma (another task that is often used to assess cooperation abilities). Not cooperating is the riskiest decision but also the one in which the gains are higher (de Heus et al., 2010
). This task permitted to show diminished cooperation abilities in a population of participants with GmD (Su et al., 2018
We assessed social metacognition concurrently with the other cognitive tasks: once before the completion of the task but after the instructions were provided and again after completion of the task. On a scale from 0 to 100%, participants rated their performance on the task either prospectively (before the task) or retrospectively (after the task) (Nelson & Narens, 1994
). This method did not add cognitive load to the completion of the task and was intended to assess appraisal and monitoring abilities (Quiles et al., 2014
Interpersonal Reactivity Index (IRI) (Davis, 1980)
This questionnaire assessed empathy abilities using four scales. The Personal Distress scale (PD) reflected the ability to feel shaken or moved by someone else’s distress. The Empathic Concern scale (EC) evaluated the ability to feel compassion or worry for someone. The Fantasy scale (FS) assessed the tendency to identify imaginatively with the thoughts and feelings of fictional characters. Finally, the Perspective Taking scale (PT) reflected the ability to adopt the perspective of someone else. The French version of this questionnaire was validated and demonstrated appropriate psychometrical properties (Gilet et al., 2013
). In the present study, a seven-point scale with responses ranging from 1 (does not describe me well) to 7 (describes me very well) was used, as proposed in the French version of the questionnaire (Gilet et al., 2013
Toronto Alexithymia Scale (TAS-20) (Loas et al., 1996)
This questionnaire assessed the level of alexithymia and included 20 questions that were answered on a scale ranging from 1 (complete disagreement) to 5 (complete agreement). It permitted us to calculate a global score and three subscores: difficulty identifying emotions (DIF), difficulty describing one’s own emotions (DDE), and thoughts aimed at the exterior instead of internal sensations (externally oriented thinking, EOT).
Participants were asked to answer several questions on a scale ranging from 1 (totally agree) to 5 (totally disagree) regarding the acceptability of the study. Participants were required to indicate (1) the clearness of the instructions, (2) whether they would have visited the lab if the experiment had not been conducted online, and (3) whether they would have preferred to meet the experimenter.
Analyses were performed using R software (version 1.25001).
The original aim of the analyses was to compare performance on SC tasks and scores on self-report questionnaires between HGB and LGB on the one hand and between HGM and LGM on the other hand to assess the relationship between the frequency of the behaviour and SC profile. Subsequently, we intended to assess the link between addiction levels and SC performance. Nevertheless, incoherent data appeared with respect to the relation between the variable assessing the mean number of hours played on average per month and the variable in the inclusion criteria that asked about the mean frequency and duration of gaming or gambling during the last three months. As questions used for inclusion were binary questions (i.e., they were answered yes or no) and because participants did not know at that time whether they would continue participating in the study, thus indicating that their levels of attention and motivation were not high at this time, we considered this variable to be less relevant. We chose to use only the mean number of hours played per month on average as the most relevant variable for defining the frequency of the behaviour. This question was asked later, at a time when participants were more involved in the study in terms of motivation and attention. Moreover, participants were required to give a numeric answer, which requires more thinking and is thus supposed to be more reliable. Thus, we chose to keep only the analysis of the link between level of addiction and frequency of behaviour and SC measures.
In all cognitive tasks assessing reaction times (RTs), we excluded all RTs less than 250 ms because they were considered to be anticipative answers.
All correlations between variables assessing SC (self-reports and cognitive tasks) on the one hand and the mean number of hours played or addiction severity scores on the other hand were tested via Spearman correlations because these variables were not normally distributed (as determined by a Kruskal‒Wallis test). For all significant correlations, several independent linear regressions with performance on the corresponding SC task as the dependent variable and the mean number of hours played or addiction severity score as the independent variable were conducted. For each regression, two models were generated, one of which included only the significant correlation between SC measures, while the other added the alexithymia score level. Both models were compared by using a likelihood ratio test, the Akaike Information Criterion (AIC), and the Bayesian Information Criterion (BIC) to select the model that best fit our data. We used this strategy to explore whether the significant correlations observed were better explained by the level of addiction or the frequency of the behaviour.
Additionally, as our objective was to separate the effect of frequency on SC performance from that of addiction, we chose to assess the effects of both variables statistically for every variable that appeared to be correlated with level of addiction and frequency of behaviour.
When this situation appeared, we conducted a linear regression using the variable of interest as the dependent variable and the addiction severity score as the independent variable. We then added the mean number of hours played to a new model. Both models were compared by using a likelihood ratio test, the AIC, and the BIC to select the best model fitting our data.
Finally, participants’ metacognitive scores were normalized using the means and standard deviations scores of all answers. Correlations between these z scores and performance on SC tasks, behaviour frequency and addiction levels were subsequently estimated.
The eSMILE study aimed to characterize SC abilities in accordance with gaming and gambling frequency using cognitive tasks and self-reports. We chose the online modality to (1) prevent people from moving during the pandemic and (2) assess the possibility of evaluating SC online. We included more than 100 participants in this study.
Regarding self-reports, the analysis showed that the more severe a person’s addiction was, the higher that person’s levels of anxiety and depression. Additionally, self-reported measures of empathy and alexithymia were not linked with the severity of addiction or the frequency of the behaviour.
Using cognitive tasks, we found a dissociation between simple and complex facial emotion identification abilities for gamers. For gamblers, this study found a negative link between the ability to identify a facial emotion and the level of addiction severity. Regarding social metacognition, particularities emerged for both groups regarding their assessments of facial identification abilities and levels of cooperation.
Gamers’ Ability to Identify Facial Emotions and Metacognitive Abilities
As discussed above, we showed that the more frequently gamers played, the better they were at identifying sadness; the more severe their addiction was, the more accurate they were in identifying simple and highly intense facial emotions. These results are contrary to our expectations, even if research has shown that GmD subjects engage in a specific form of unconscious processing of emotional faces in the context of a population of patients with GmD (He et al., 2019
; Peng et al., 2017
). These studies did not assess the ability to correctly identify a facial emotion. Our results may also be linked with the fact that playing VGs may enhance cognitive functioning (Bediou et al., 2018
). Indeed, playing VGs may improve the ability to efficiently process visual cues and may have thus helped our participants answer more accurately when identifying basic facial emotions as a result of this training.
Interestingly, we also showed that the more frequently gamers played, the lower their accuracy in identifying positive emotions in complex pictures. This finding may seem to contradict the results mentioned above. Nevertheless, those results were obtained in the context of a more complex task and may suggest that emotion processing is facilitated by playing VGs in the context of responding to simple stimuli but not complex stimuli. This dissociation is possible and has been demonstrated in individuals on the autism spectrum without an intellectual disability with respect to simple and complex prosody identification (Icht et al., 2021
However, our study remains exploratory and does not permit us to make specific conclusions regarding the dissociation between simple and complex facial emotion identification. To answer this question, we would need to compare participants with a high frequency and a low frequency of gaming and control for addiction severity on a task such as the Reading the Mind in the Eyes Test. Indeed, this task comprises pictures that include only eyes, to which participants must attribute a complex emotional state. One study showed that the level of performance on this task with respect to negative emotions in among students might predict the severity of GmD (Aydın et al., 2020
Contrary to our hypothesis, we did not find any link between the level of addiction and accuracy in identifying simple or complex facial emotions in gamers. This finding could be linked to our sample’s low level of addiction and the fact that they were not seeking treatment. Our sample did not represent patients in a clinical setting who faced with social difficulties. It would be interesting to use the ER-40 and the MET-CORE to investigate treatment-seeking patients to confirm the absence of any alteration in the ability to identify a facial emotion in the context of GmD while continuing to use self-reports to control for alexithymia because this ability impacts facial emotion processing (Grynberg et al., 2012
Interestingly, the analysis of the metacognitive scores after the ER-40 task (assessing basic emotion recognition) showed that the higher gamers’ levels of addiction were, the lower their estimations of their performance. Our results suggest that pathological gamers tend to underestimate their ability to identify social stimuli. Metacognitive skills were proven to be deficient in a group of patients with cocaine use disorder (Balconi et al., 2014
) and patients with Parkinson’s disease and GbD compared to patients with Parkinson’s disease but without GbD (Angioletti et al., 2020
). In both studies, participants completed the Iowa Gambling Task (a decision-making task under conditions of uncertainty (Bechara, 2005)) and were required to estimate their performance on this task. These results suggest that addiction may be linked to a global deficit in metacognition. Indeed, our results showed that in cases of GmD, social metacognition might be altered. Nevertheless, further studies are needed to determine whether this tendency to underestimate one’s performance exists in the context of other tasks. If a deficit in metacognition appears in regard to several cognitive functions, this deficiency would be a common trait among cases of GbD, cocaine use disorder, and GmD. In this situation, metacognition deficits would be linked to addictive processes rather than the ingestion of psychoactive substances.
Gambler’s Ability to Identify Facial Emotions
As discussed above, we showed that, for gamblers, the more frequently they played or the higher their level of addiction, the lower their accuracy in identifying complex negative emotions. These results seem to indicate that the errors made by gamblers may be linked to addiction rather than the frequency of the behaviour. Additionally, we controlled for alexithymia level and showed that these difficulties were not linked to this factor.
Using this framework, one study showed an alteration of emotion identification abilities in GbD patients with respect to both faces and voices (Kornreich et al., 2016
). Thus, the research of Kornreich et al. and our study both suggest that SC, specifically the ability to identify emotions, is impaired in this context. Further studies comparing high frequency, low frequency, and GbD patients must be conducted to confirm this hypothesis. Indeed, we must investigate whether these difficulties may be linked more closely with a certain type of emotion (negative, positive ones) and whether other abilities are impaired. Actually, if facial emotion recognition is altered in cases of GbD, this situation could lead to other deficits because this type of information is used when inferring mental states (theory of mind, Achim et al., 2020
; Premack & Woodruff, 1978
). This type of difficulty would explain some of the difficulties that have reported by patients regarding familial difficulties, for example (Black et al., 2012
Additionally, this component could be a target in therapies. Indeed, the use of cognitive remediation may help patients deal with social difficulties and enhance their social functioning (Kurtz et al., 2016
; Kurtz & Richardson, 2012
). This approach would lead to improving the social sphere, which is important in cases of addiction (Kim & Hodgins, 2018
No significant relationships were found when assessing levels of cooperation in the social decision-making task. Some people indicated that they were familiar with the dilemma (even if they identified it as the prisoner dilemma instead of the CG). We decided to remove the scores of participants who were familiar with the prisoner’s dilemma (n = 4) to determine whether those participants may have biased our results. Nevertheless, the correlations among the number of cooperative choices, behaviour frequency, and levels of addiction remained nonsignificant.
Regarding the link with social metacognition, the higher gamblers’ levels of addiction severity were, the more confident they were before the CG. Nevertheless, level of addiction severity was not correlated with more cooperative choices. Additionally, the metacognitive scores of gamers tended to correlate with their behaviour towards the computer, while the metacognitive scores of gamblers tended to correlate with their levels of cooperation with the other player. A study of GmD subjects using the CG showed that participants with GmD tended to vary in terms of their level of cooperation depending on the partner with whom they were playing (Su et al., 2018
). Even if we did not find that more cooperative choices were linked with addiction in one condition or the other, we can assume that gamers attributed their success to the fact that they cooperated with the computer, while gamblers anticipated that cooperation with the other player would be a good performance. Further research is needed to investigate the profiles of cooperation using different methods (more trials, different stories), but preserving both conditions seems important.
Empathy Abilities in BA
No significant relationships were found between behaviour frequency or addiction and empathy measures using self-report and cognitive tasks. These results suggest the preservation of this ability, as has been found in the context of alcohol use disorder using the MET-CORE (Grynberg et al., 2017
) and suggested by tasks completed by participants without GmD (Ferguson & Colwell, 2020
; Gao et al., 2017
; Kühn et al., 2018
; Miedzobrodzka et al., 2021
; Szycik et al., 2017
). Nevertheless, lower empathic abilities have been reported with IRI in a GbD population (Tomei et al., 2017
). Thus, further studies are needed to confirm the presence of empathy alterations in GbD by assessing a population of treatment-seeking patients. It is also necessary to assess the attribution component of empathy using another task, such as the Levels of Emotional Awareness Scale [LEAS (Lane et al., 1990
)]. This scale demonstrated a lower level of emotional awareness in a population of substance use disorder patients (Carton et al., 2010
). Using this task in the context of GbD and GmD patients may be a way of highlighting the dissociation between the BAs with respect to this SC component. This approach would permit the adaptation of treatment depending on the features of each BA.
Difference Between Gamers and Gamblers
As showed in results above, this study did not highlight the same difficulties in gamers and gamblers. Indeed, gamblers’ ability to identify facial emotions may be affected while it seems to be the gamers’ ability to assess their own social abilities that seems altered. Despite the fact that in the CIM 11 (World Health Organization, 2018
), criteria seem very close, characteristics of each practice may impact differently the cognition in general.
Indeed, gaming literature analysis showed the positive impact of gaming on cognitive function such as attention. Authors pointed out that difference only appear for the gamers that play the most, and that preferred game type may also impact attention and flexibility ability (Nuyens et al., 2019
). Regarding social cognition, a systematic literature review showed that high frequency of gaming may be linked with improvement of moral competencies and changes in social decision making (Hurel et al., 2023b
). Type of game could also impact this type of competencies because multiplayer games may need more theory of mind abilities. Motives to play could also impact the social investment of gamers and train (or not) social abilities.
For gambling, there is a continuum between luck and strategy. Indeed, gamble type vary the possibility to train certain cognitive functions. The most strategic and social gambling type may be poker (Palomäki et al., 2020
). One study showed specificities in the allocation of attention of social cues in poker players compared to control (Hurel et al., 2023a
Thus, it seems that both gaming and gambling may impact social cognition but for different reasons. Both practices may train certain social cognitive functions, depending on the type of game or gamble used. It may be interesting to compare social cognition performances between two groups of gamblers or gamers with different preferred game or gamble.
The Online Experimentation
We wanted to assess the possibility of conducting such studies online. Our conclusions regarding this experiment are mixed. Indeed, our statistical plan was affected by instances in which inclusion criteria were not respected. This situation would not have happened if we had checked these criteria by phone, for example. Additionally, we encountered some difficulties regarding the number of inclusions. Indeed, while we planned to recruit 200 participants, we ultimately only recruited 105 over nine months. Throughout the duration of this study, we understood that it would have been easier to conduct this study via smartphone because many people do not have access to a computer, and diffusion via social media would have been facilitated. Moreover, including a reward for participation (such as a prize or gift certificate) could have increased participation and motivation.
However, the acceptability questions showed that participants might prefer not to meet the experimenter, thus suggesting that online participation may be easier. Our conclusion is that our methodology can be improved, specifically with respect to the construction of the task (accessibility on smartphones, prior confirmation of inclusion criterion).
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.