Ang Li

Background: Depression carries a heavy global burden, yet treatment gaps persist largely because individuals lack confidence in mental health professionals. Psychoeducation can shift these beliefs, but scaling persuasive messages is difficult. Large language models (LLMs) offer a scalable solution, though the specific text-based features that make LLM-generated psychoeducation persuasive remain unidentified. This study identified these features and tested their integration into an LLM prompt to shift confidence in mental health professionals. Methods: In Phase 1, 168 participants rated text pairs contrasting high versus low levels of four candidate features. In Phase 2, 40 participants were randomized to read psychoeducational passages generated by either a prompt incorporating the retained features ( n = 20) or a baseline prompt ( n = 20). The primary outcome was pre-to-post change in confidence in mental health professionals. Results: Source credibility, argument quality, and processing fluency significantly boosted both perceived credibility and persuasiveness (all p < 0.01); affective warmth and empathy was excluded. Participants reading feature-engineered texts showed greater improvement in attitudes than the control group ( t (38) = 3.37, p = 0.002, d = 1.07). Controlling for baseline scores, the group effect remained significant ( F (1,37) = 9.31, p = 0.004, partial η ² = 0.20). Conclusion: Strategically prompt-engineered LLM outputs incorporating empirically selected persuasive features significantly improve confidence in mental health professionals. This pilot study provides a preliminary evidence-based framework that may inform scalable, LLM-powered public mental health interventions.

Sensation seeking represents a significant risk factor for various mental health disorders and maladaptive behaviors, highlighting the need for objective assessment methods that circumvent the limitations of traditional self-report measures. This study introduces an innovative digital phenotyping approach that combines computational gait analysis with machine learning (ML) to quantify sensation-seeking traits and examines its validity. Natural gait sequences (using a Sony camera at 25 FPS) and self-report measures (Brief Sensation-Seeking Scale for Chinese, BSSS-C) were collected from 233 healthy adults. Computer vision processing through OpenPose extracted 25 skeletal keypoints, which were subsequently transformed into a hip-centered coordinate system and denoised using Gaussian filtering. From these kinematic data, 300 temporospatial gait features capturing various aspects of movement dynamics were derived. Using a supervised ML approach with feature selection, three ML models (SMO Regression, Multilayer Perceptron, and Bagging) were developed and compared through 10-fold cross-validation. The SMO Regression model demonstrated superior performance (r = 0.60, MAE = 3.50, RMSE = 4.59, R² = 0.26), outperforming the other approaches. These results establish proof-of-concept for gait-based digital phenotyping of sensation seeking, offering a scalable, objective assessment paradigm with potential applications in clinical screening and behavioral research. The methodological framework presented here advances the field of behavioral biometrics by demonstrating how computer vision and ML can transform basic movement patterns into meaningful psychological indicators.

Background: Implementing machine learning prediction of negative attitudes towards suicide may improve health outcomes. However, in previous studies, varied forms of negative attitudes were not adequately considered, and developed models lacked rigorous external validation. By analyzing a large-scale social media dataset (Sina Weibo), this paper aims to fully cover varied forms of negative attitudes and develop a classification model for predicting negative attitudes as a whole, and then to externally validate its performance on population and individual levels.Methods: 938,866 Weibo posts with relevant keywords were downloaded, including 737,849 posts updated between 2009 and 2014 (2009–2014 dataset), and 201,017 posts updated between 2015 and 2020 (2015–2020 dataset). (1) For model development, based on 10,000 randomly selected posts from 2009 to 2014 dataset, a human-based content analysis was performed to manually determine labels of each post (non-negative or negative attitudes). Then, a computer-based content analysis was conducted to automatically extract psycholinguistic features from each of the same 10,000 posts. Finally, a classification model for predicting negative attitudes was developed on selected features. (2) For model validation, on the population level, the developed model was implemented on remaining 727,849 posts from 2009 to 2014 dataset, and was externally validated by comparing proportions of negative attitudes between predicted and human-coded results. Besides, on the individual level, similar analyses were performed on 300 randomly selected posts from 2015 to 2020 dataset, and the developed model was externally validated by comparing labels of each post between predicted and actual results.Results: For model development, the F1 and area under ROC curve (AUC) values reached 0.93 and 0.97. For model validation, on the population level, significant differences but very small effect sizes were observed for the whole sample (χ²₁ = 32.35, p < 0.001; Cramer’s V = 0.007, p < 0.001), men (χ²₁ = 9.48, p = 0.002; Cramer’s V = 0.005, p = 0.002), and women (χ²₁ = 25.34, p < 0.001; Cramer’s V = 0.009, p < 0.001). Besides, on the individual level, the F1 and AUC values reached 0.76 and 0.74.Conclusion: This study demonstrates the efficiency and necessity of machine learning prediction of negative attitudes as a whole, and confirms that external validation is essential before implementing prediction models into practice.

Real-time non-intrusive assessment of personality traits in drug addicts is crucial for enhancing the effectiveness of treatment efforts. Previous studies have demonstrated that characteristics of facial movements can be used to measure personality traits in the general population. However, the facial movements of drug addicts may be quite different from those of the general population, so it remains uncertain whether similar findings can be observed in drug addicts. In this study, data from 336 drug addicts were collected and analyzed by using Sony video camera (recording facial movement) and NEO-FFI questionnaire (measuring Big Five personality traits). After that, by using six different machine learning algorithms, classification models for predicting Big Five personality traits in drug addicts were developed and validated based on facial movement analysis. Results showed that, among different models, the Random Forest model obtained the best overall classification performance (validity: the accuracy values ranged from 0.60 to 0.72, and the F1 values ranged from 0.65 to 0.78; reliability: the split-half reliability values ranged from 0.81 to 0.93). Findings of this study confirm the potential of facial movement analysis in predicting personality traits for drug addicts.

Editorial: Improving the clinical value of digital phenotyping in mental health

Editorial: Understanding public discourse for digital mental health promotion

Backgrounds: COVID-19 is difficult to end in a short time and people are still facing huge uncertainties. Since people's lives are gradually returning to normal, the sense of control and intolerance of uncertainty, which were mainly focused by past studies, are not specific to COVID-19 and will be more influenced by some factors unrelated to the pandemic. Therefore, they may be difficult to accurately reflect the individuals' perceptions of uncertainty. Besides, past research just after the outbreak mainly investigated people in high levels of uncertainty, we don't know the impact of uncertainties on individuals' psychological states when people gradually recovered their sense of control. To solve these problems, we proposed the concept of “pandemic uncertainty” and investigated its impact on people's daily lives.Methods: During October 20, 2021 to October 22, 2021, this study obtained data about uncertainty, depression, positive attitude, pandemic preventive behavior intentions, personality, and social support from 530 subjects using convenient sampling. The subjects were all college students from the Dalian University of Technology and Dalian Vocational and Technical College. According to the distribution of uncertainty, we divided the dataset into high and low groups. Subsequently, by using uncertainty as the independent variable, the grouping variable as the moderating variable, and other variables as the control variables, the moderating effects were analyzed for depression, positive attitude, and pandemic preventive behavior intentions, respectively.Results: The results showed that the grouping variable significantly moderate the influence of uncertainty on positive attitude and pandemic preventive behavior intentions but had no significant effect on depression. Simple slope analysis revealed that high grouping uncertainty significantly and positively predicted positive attitude and pandemic preventive behavior intentions, while low grouping effects were not significant.Conclusion: These results reveal a nonlinear effect of pandemic uncertainty on the pandemic preventive behavior intentions and positive life attitudes and enlighten us about the nonlinear relationship of psychological characteristics during a pandemic.

Text sentiment analysis, also known as opinion mining, is research on the calculation of people's views, evaluations, attitude and emotions expressed by entities. Text sentiment analysis can be divided into text-level sentiment analysis, sen-tence-level sentiment analysis and aspect-level sentiment analysis. Aspect-Based Sentiment Analysis (ABSA) is a fine-grained task in the field of sentiment analysis, which aims to predict the polarity of aspects. The research of pre-training neural model has significantly improved the performance of many natural language processing tasks. In recent years, pre training model (PTM) has been applied in ABSA. Therefore, there has been a question, which is whether PTMs contain sufficient syntactic information for ABSA. In this paper, we explored the recent DeBERTa model (Decoding-enhanced BERT with disentangled attention) to solve Aspect-Based Sentiment Analysis problem. DeBERTa is a kind of neural language model based on transformer, which uses self-supervised learning to pre-train on a large number of original text corpora. Based on the Local Context Focus (LCF) mechanism, by integrating DeBERTa model, we purpose a multi-task learning model for aspect-based sentiment analysis. The experiments result on the most commonly used the laptop and restaurant datasets of SemEval-2014 and the ACL twitter dataset show that LCF mechanism with DeBERTa has significant improvement.

Background: The new reality of cybersuicide raises challenges to ideologies about the traditional form of suicide that does not involve the internet (offline suicide), which may lead to changes in audience’s attitudes. However, knowledge on whether stigmatizing attitudes differ between cybersuicides and offline suicides remains limited. Objective: This study aims to consider livestreamed suicide as a typical representative of cybersuicide and use social media data (Sina Weibo) to investigate the differences in stigmatizing attitudes across cybersuicides and offline suicides in terms of attitude types and linguistic characteristics. Methods: A total of 4393 cybersuicide-related and 2843 offline suicide-related Weibo posts were collected and analyzed. First, human coders were recruited and trained to perform a content analysis on the collected posts to determine whether each of them reflected stigma. Second, a text analysis tool was used to automatically extract a number of psycholinguistic features from each post. Subsequently, based on the selected features, a series of classification models were constructed for different purposes: differentiating the general stigma of cybersuicide from that of offline suicide and differentiating the negative stereotypes of cybersuicide from that of offline suicide. Results: In terms of attitude types, cybersuicide was observed to carry more stigma than offline suicide (χ21=179.8; P<.001). Between cybersuicides and offline suicides, there were significant differences in the proportion of posts associated with five different negative stereotypes, including stupid and shallow (χ21=28.9; P<.001), false representation (χ21=144.4; P<.001), weak and pathetic (χ21=20.4; P<.001), glorified and normalized (χ21=177.6; P<.001), and immoral (χ21=11.8; P=.001). Similar results were also found for different genders and regions. In terms of linguistic characteristics, the F-measure values of the classification models ranged from 0.81 to 0.85. Conclusions: The way people perceive cybersuicide differs from how they perceive offline suicide. The results of this study have implications for reducing the stigma against suicide.