Facial emotional expression in response to alcohol‐promoting commercials in individuals with alcohol use disorder: A pilot study

Alcohol use disorder (AUD) is a chronic relapsing disorder in which environmental alcohol cues can trigger craving and relapse. ¹ , ² , ³ Alcohol‐promoting digital commercials are a pervasive form of alcohol‐related exposure in contemporary society, yet the psychological and emotional processes they elicit in individuals with AUD remain poorly characterized. ³ , ⁴ Facial expressions convey emotions, intentions, and evaluative reactions, ⁵ , ⁶ and automated facial expression analysis may provide a scalable complementary method for assessing rapid and implicit cue reactivity.

Fifty participants were initially recruited (AUD [n = 25] and healthy control [HC] [n = 25]). After withdrawals and technical exclusions, 16 individuals with AUD (47.6 ± 7.8 years; 3 females) and 23 HCs (42.3 ± 10.2; 3 females) were included. Current psychiatric comorbidity was assessed using the Structured Clinical Interview for DSM‐5. The study was approved by the institutional review boards of the Hizen Psychiatric Medical Center and the Institute of Science Tokyo Hospital, and all participants provided written informed consent. Please see Supplementary Methods for details.

Participants viewed four 15‐s video clips from digital alcoholic‐beverage commercials and four 15‐s clips from soft‐drink commercials as controls. Consistent with prior studies, ³ , ⁶ soft‐drink stimuli were carefully matched to the alcohol‐related commercials in design, complexity, color composition, motion, and the presence of human faces. Each clip was presented twice in pseudorandom order, with a 10–15‐s fixation interval between clips (Figure 1a). Participants also rated their desire to drink alcohol after each clip; drinking desire was calculated as the difference between the mean ratings for alcohol‐related and soft‐drink commercials. Emotional intensity (EI) of happy, sad, angry, surprised, scared, disgusted, and neutral during video viewing was quantified using FaceReader (Figure 1b), which classifies facial expressions based on the Facial Action Coding System. ⁵ , ⁶ EI values ranged from 0 to 1 were log‐transformed (log[1 + EI]) before analysis. For each emotional category, mixed‐design analyses of variance assessed group effects (HC vs. AUD), condition effects (alcohol vs. soft‐drink), and their interaction. Statistical significance was defined as p < 0.05 (two‐tailed). Please see Supplementary Methods for details.

Figure 1.

Experimental task and emotional intensity (EI) data acquisition. (a) Experimental task. Digital commercial videos promoting alcohol or soft drinks were presented in a pseudorandom order. The videos were presented on a laptop computer equipped with a 13.0‐in. display. Participants' facial expressions were continuously recorded throughout the task using the built‐in video camera positioned at the top of the laptop screen. A representative scene from the commercials is illustrated with a still comic‐style image to depict the task's structure and timing. Notably, this image was created for illustrative purposes and does not depict the actual commercial video used in the experiment. (b) EI data acquisition. Representative output from the FaceReader software illustrating the automated coding of facial emotional expressions. FaceReader estimates the EI of facial expressions (happy, sad, angry, surprised, scared, disgusted, and neutral) based on the Facial Action Coding System. A comic‐style still image depicting a human face is shown for illustrative purposes. (c) EI values during the alcohol and soft‐drink conditions. Mean EI values for happy, sad, and neutral facial expressions are shown. Mean EI values for the remaining facial expressions (angry, surprised, scared, and disgusted) are presented in Figure S1. Error bars indicate ± standard errors. AUD, alcohol use disorder; HC, healthy control.

Age and sex did not differ significantly between groups, although estimated full‐scale intelligence quotient (IQ) was slightly lower in the AUD group (Table S1). Alcohol Use Disorders Identification Test (AUDIT) scores were significantly higher in AUD than HCs. Drinking desire scores did not differ significantly between groups (HC 2.7 ± 1.8; AUD 1.8 ± 1.9, p = 0.14). Figures 1c and S1 depict mean facial EI values across conditions. For happiness EI, the condition effect was not significant (F = 1.48, p = 0.23), whereas significant group (F = 5.97, p = 0.02) and group × condition effects (F = 4.24, p = 0.046) were observed. Post hoc analysis showed higher happiness EI in AUD than HCs during alcohol‐related commercials (p = 0.04), although this did not survive Bonferroni correction. For sadness EI, significant group (F = 11.70, p < 0.01), condition (F = 5.04, p = 0.03), and group × condition effects (F = 5.84, p = 0.02) were found. Sadness EI increased during alcohol‐related versus soft‐drink commercials only in AUD (p < 0.01) and was higher in AUD than HCs during alcohol‐related commercials (p < 0.01). For neutral EI, only the group effect was significant (F = 4.70, p = 0.04), with no condition effect (F = 3.39, p = 0.07) or group × condition interaction (F = 1.98, p = 0.17). No significant main effects or interactions were detected for the remaining facial expressions (Table S2). As a sensitivity analysis, mixed‐design analyses were performed with estimated full‐scale IQ included as a covariate. All effects for sadness remained significant after adjustment for IQ (group: F = 8.79, p < 0.01; condition: F = 5.18, p = 0.03; and interaction: F = 6.19, p = 0.02), whereas the group and interaction effects for happiness (F = 3.04, p = 0.09; F = 3.67, p = 0.06, respectively) and the group effect for neutral EI (F = 2.93, p = 0.095) were no longer significant.

These preliminary findings suggest that alcohol‐promoting commercials elicit altered facial emotional responses in AUD. The concurrent modulation of happiness‐ and sadness‐related expressions may reflect emotional ambivalence toward alcohol cues: alcohol‐related advertising may activate appetitive reward expectancies while also evoking negative self‐referential affects. ⁷ , ⁸ The robust, condition‐specific sadness response suggests that negative emotion may be a particularly important component of alcohol‐cue reactivity in AUD. Conversely, the effects observed for happiness and neutral EI were attenuated after adjustment for IQ, requiring a cautious interpretation.

Alcohol marketing may increase immediate alcohol consumption. ⁹ However, the generalizability of these findings requires further investigation. Animated characters and computer‐generated imagery are also commonly used in digital advertising. Given evidence that cartoon and real faces may be processed differently at the neural level, ¹⁰ future studies should examine whether alcohol‐related commercials featuring animated content elicit emotional responses comparable to those evoked by real‐world commercials. Further investigation of emotional reactivity to digital alcohol advertising in individuals with AUD may contribute to public health discussions regarding strategies to reduce potentially harmful advertising exposure among vulnerable populations.

This study has several limitations, including its modest sample size, psychiatric comorbidities, heterogeneous psychotropic medication use, and cross‐sectional design. Although soft‐drink stimuli were carefully matched to alcohol‐related stimuli, complete equivalence across naturalistic commercial features, such as foreground–background ratio and zooming speed, was difficult to achieve. Nevertheless, the present findings suggest the potential utility of automated facial expression analysis as a complementary tool for assessing emotional cue reactivity in AUD, while underscoring the need for further validation in larger longitudinal samples.

AUTHOR CONTRIBUTIONS

Takefumi Ueno: Conceptualization, data curation; funding acquisition; investigation; methodology; project administration; writing–original draft; writing–review & editing. Naho Nakayama: Conceptualization; data curation, investigation; methodology; writing–review & editing. Shou Fukushima: Conceptualization; data curation; investigation; methodology; writing–review & editing. Kazunori Matsuguchi: Conceptualization, data curation; investigation; methodology; writing–review & editing. Kazuhiro Tojiki: Conceptualization; data curation, investigation; methodology; writing–review & editing. Chenyu Qian: Data curation; investigation, methodology; validation; writing–review & editing. Shisei Tei: Conceptualization; methodology, supervision; validation; writing–review & editing. Hidehiko Takahashi: Conceptualization; investigation; methodology; project administration; supervision; writing–review & editing. Junya Fujino: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration, writing–original draft; writing–review & editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS APPROVAL STATEMENT

The study was approved by the institutional review boards of the National Hospital Organization Hizen Psychiatric Medical Center and the Institute of Science Tokyo Hospital, Japan.

PATIENT CONSENT STATEMENT

All the participants were informed of the purpose of the study and that their participation was anonymous and voluntary. They provided their written informed consent to participate in this study.

CLINICAL TRIAL REGISTRATION

N/A.

DECLARATION OF GENERATIVE AI AND AI‐ASSISTED TECHNOLOGIES IN THE MANUSCRIPT PREPARATION PROCESS

During the preparation of this work, the authors used (ChatGPT, OpenAI, GPT‐5, 2025–2026) in order to translate the manuscript from Japanese to English. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article. No AI tools were used for data analysis, study design, or interpretation of results.

Supporting information

Supporting File 1

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.