MétaCan
Menu
Back to cohort

From Posts to Predictions: A User-Aware Framework for Faithful and Transparent Detection of Mental Health Risks on Social Media

2025· article· W7122474954 on OpenAlexfundno aff
Hessam Amini, Leila Kosseim

Bibliographic record

VenueInternational conference Recent advances in natural language processing · 2025
Typearticle
Language
FieldPsychology
TopicMental Health via Writing
Canadian institutionsnot available
FundersNatural Sciences and Engineering Research Council of Canada
KeywordsSocial mediaMental healthThe InternetPublic healthSocial isolationPerception

Abstract

fetched live from OpenAlex

We propose a user-aware attention-based framework for early detection of mental health risks from social media posts.Our model combines DisorBERT, a mental health-adapted transformer encoder, with a user-level attention mechanism that produces transparent postlevel explanations.To assess whether these explanations are faithful, i.e., aligned with the model's true decision process, we apply adversarial training and quantify attention faithfulness using the AtteFa metric.Experiments on four eRisk tasks (depression, anorexia, self-harm, and pathological gambling) show that our model achieves competitive latencyweighted F1 scores while relying on a sparse subset of posts per user.We also evaluate attention robustness and conduct ablations, confirming the model's reliance on high-weighted posts.Our work extends prior explainability studies by integrating faithfulness assessment in a real-world high-stakes application.We argue that systems combining predictive accuracy with faithful and transparent explanations offer a promising path toward safe and trustworthy AI for mental health support.

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 machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.004
metaresearch head score (Gemma)0.017
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: Methods
Teacher disagreement score0.011
Threshold uncertainty score0.024

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0040.017
Meta-epidemiology (narrow)0.0020.001
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0030.001
Science and technology studies0.0010.001
Scholarly communication0.0030.005
Open science0.0020.003
Research integrity0.0020.003
Insufficient payload (model declined to judge)0.0040.002

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.

Opus teacher head0.065
GPT teacher head0.465
Teacher spread0.401 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreMethods

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".

Quick stats

Citations0
Published2025
Admission routes1
Has abstractno

Explore more

Same venueInternational conference Recent advances in natural language processingSame topicMental Health via WritingFrench-language works237,207