Digital Phenotyping
Passive behavioural signals from smartphones and computers, combined with validated self-report and ecological momentary assessment.
Research Initiative
Predictive Digital Phenotyping for Early Depression Detection in Young Adults
DeepVYBE combines digital phenotyping, conversational assessment, and Responsible AI to support earlier, ecologically valid, and clinically interpretable depression screening for young adults.
Why it matters
Relevance
Depression among young adults remains one of the most pressing mental health challenges, with substantial personal, social, and economic consequences. Current screening practices rely mainly on self-report and episodic clinical contact, which may fail to capture behavioural changes as they emerge in everyday life.
This project addresses that gap by combining digital phenotyping, conversational assessment, and Responsible AI to support earlier, more ecologically valid, and clinically interpretable depression screening workflows. Its relevance lies in its potential to support mental health services with non-invasive, scalable, and human-supervised screening tools.
Core Research Directions
Conversational Assessment
LLM-mediated conversational assessment integrated into screening workflows for more ecologically valid, interpretable signals.
Responsible AI & Oversight
Early risk signalling with transparent, privacy-aware models reviewed through psychologist-facing dashboards and clinical supervision.
Implementation
Pipeline
The project follows a structured workflow from clinical framing to pilot evaluation, integrating passive and active data collection with human-supervised AI outputs.
Step 1
Clinical screening framework
Definition of the clinical screening framework and identification of behavioural indicators associated with depression.
Step 2
Digital data collection
Development of infrastructure combining passive behavioural signals from smartphone and computer interaction, validated self-report measures, and LLM-mediated conversational assessment.
Step 3
Engagement layer
Exploration of a lightweight mobile interactive environment with gamified elements to support naturalistic assessment without shifting focus away from screening.
Step 4
Responsible AI & clinical review
Development of Responsible AI models for early risk signalling, with outputs reviewed through a psychologist-facing dashboard.
Step 5
Pilot evaluation
Iterative refinement and evaluation in pilot studies focusing on feasibility, acceptability, usability, and clinical relevance.
Consortium
Team
The consortium combines complementary expertise required to execute the project end-to-end, including clinical psychology, behavioural assessment, digital phenotyping, human-centered AI, and software development. This multidisciplinary composition is essential to ensure that the project is clinically grounded, technically robust, and feasible in real-world settings.
The team brings together expertise in depression assessment, conversational AI, behavioural data analysis, privacy-aware digital health systems, and mobile application development, supported by access to an institutional clinical context that strengthens feasibility and pilot validation capacity.
Meet the team