AI for Analyzing Mental Health Disorders Among Social Media Users: Quarter-Century Narrative Review of Progress and Challenges
Bibliographic record
Abstract
BACKGROUND: Mental health disorders are currently the main contributor to poor quality of life and years lived with disability. Symptoms common to many mental health disorders lead to impairments or changes in the use of language, which are observable in the routine use of social media. Detection of these linguistic cues has been explored throughout the last quarter century, but interest and methodological development have burgeoned following the COVID-19 pandemic. The next decade may see the development of reliable methods for predicting mental health status using social media data. This might have implications for clinical practice and public health policy, particularly in the context of early intervention in mental health care. OBJECTIVE: This study aims to examine the state of the art in methods for predicting mental health statuses of social media users. Our focus is the development of artificial intelligence-driven methods, particularly natural language processing, for analyzing large volumes of written text. This study details constraints affecting research in this area. These include the dearth of high-quality public datasets for methodological benchmarking and the need to adopt ethical and privacy frameworks acknowledging the stigma experienced by those with a mental illness. METHODS: A Google Scholar search yielded peer-reviewed articles dated between 1999 and 2024. We manually grouped the articles by 4 primary areas of interest: datasets on social media and mental health, methods for predicting mental health status, longitudinal analyses of mental health, and ethical aspects of the data and analysis of mental health. Selected articles from these groups formed our narrative review. RESULTS: Larger datasets with precise dates of participants' diagnoses are needed to support the development of methods for predicting mental health status, particularly in severe disorders such as schizophrenia. Inviting users to donate their social media data for research purposes could help overcome widespread ethical and privacy concerns. In any event, multimodal methods for predicting mental health status appear likely to provide advancements that may not be achievable using natural language processing alone. CONCLUSIONS: Multimodal methods for predicting mental health status from voice, image, and video-based social media data need to be further developed before they may be considered for adoption in health care, medical support, or as consumer-facing products. Such methods are likely to garner greater public confidence in their efficacy than those that rely on text alone. To achieve this, more high-quality social media datasets need to be made available and privacy concerns regarding the use of these data must be formally addressed. A social media platform feature that invites users to share their data upon publication is a possible solution. Finally, a review of literature studying the effects of social media use on a user's depression and anxiety is merited.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.020 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.003 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.004 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".