PSYCHED Adaptive Dataset and Supplementary Materials

This repository contains all datasets and supplementary materials used in the development and evaluation of the PSYCHED Adaptive diagnostic system. The Question Selection Survey (PDF + responses) captures the initial 88-item instrument and the 70 student responses used to refine the final 56 diagnostic questions. The training dataset includes both synthetic data generated using LLaMA Instruct and LLaMA Nemotron reward models, as well as cleaned real data from the DAIC-WOZ corpus. Validation materials include the PHQ-9 + MDI comparative survey (PDF), response dataset, and pilot-test performance results. A complete user manual for the PSYCHED Adaptive system is also included. Question selection survey.pdf: Contains the original 88-item questionnaire administered to university students for initial item selection. Question selection survey responses – Student Mental Health Assessment.xlsx: Includes 70 student responses to the 88-item questionnaire, used to statistically select the final 56 items. Synthetic Dataset.csv: Contains the synthetic Likert-scale dataset generated using LLaMA Instruct + LLaMA Nemotron reward model for training the classifier. DAIC WOZ cleaned.csv: Provides the cleaned subset of the DAIC-WOZ dataset used as real training data for the adaptive engine and Likert classifier. Validation survey.pdf: The PHQ-9 + MDI + PSYCHED Adaptive comparative evaluation survey administered to real users during system validation. Validation survey responses.xlsx: Contains participant responses comparing PSYCHED Adaptive with benchmark tools (PHQ-9 and MDI). Validation – psyched adaptive results.csv: Reports outcome metrics from the pilot study where users interacted with the PSYCHED Adaptive system in real conditions. PSYCHED Adaptive – usermanual.docx: Provides the end-user manual explaining how to operate the PSYCHED Adaptive system. PSYCHED Adaptive- CODE.ipynb: Code to validate the questionnaire (item reduction from 78->56) and to train a likert classifier. And to validate the the PSYCHED Adaptive comparison with PHQ-9 and MDI.