New AI-Driven Tools for Enhancing Campus Well-being: A Prevention and Intervention Approach

Synthesis: This dissertation presents an integrated AI framework for campus Well-Being spanning prevention (improving feedback collection) and intervention (advancing mental health detection). It represents an important application of Large Language Models (LLMs) and Generative AI technologies to Student Experience that extends beyond academic learning to holistic student support in Higher Education. On the prevention side, TigerGPT — a personalized survey chatbot grounded in conversational design and engagement theory — achieved 75% usability and 81% satisfaction, demonstrating that AI-mediated feedback collection can improve on traditional survey instruments. AURA, an adaptive follow-up question framework using reinforcement learning, dynamically selects question types (validate, specify, reflect, probe) to deepen responses, achieving a +0.12 mean quality gain (p=0.044, d=0.66). On the intervention side, PsychoGPT provides explainable mental health assessment built on DSM-5 and PHQ-8 guidelines — a Affective Computing application that prioritizes clinical grounding and interpretability over black-box classification. The Stacked Multi-Model Reasoning (SMMR) architecture reduces hallucination risk by layering expert models: early layers handle localized subtasks while later layers reconcile findings, outperforming single-model solutions on the DAIC-WOZ Benchmark. The integrated framework — where adaptive survey insights flow into specialized mental health detection models — represents a novel Edtech Platform architecture for campus well-being. For the Administrators perspective, this work provides a concrete roadmap for universities seeking to deploy AI tools that monitor student satisfaction and detect mental health risks — areas where many institutions currently lack effective methods.

What this means for practice

• Administrators. Replace the static climate survey with a conversational channel where response rates stall: TigerGPT's pilot drew 21 feedback surveys and 17 written evaluator reports, at a 75% usability rating, 81% satisfaction, and 50% of users preferring it to a traditional questionnaire.
• Administrators. Keep the AI advisory: route quality summaries and flagged disclosures to counseling and student-affairs staff rather than acting on model output, since the proposed framework has the preventive layer hand off to human services when warranted.
• Administrators. Settle consent, data retention, and escalation rules before collecting well-being disclosures, because the design keeps adaptation within a session — reacting to the present respondent rather than pooling sensitive signals across users.
• Administrators. Specify within-session adaptation of follow-up question types when procuring or building a campus chatbot: AURA's controlled comparison (n = 20 conversations per condition) produced a +0.12 mean gain in response quality over non-adaptive baselines (p = 0.044, d = 0.66) with 63% fewer specification prompts.
• Administrators. Do not treat the clinical components as deployment-ready: Psycho Analyst and SMMR were evaluated on the public DAIC-WOZ corpus and 48 psychiatric case studies, not on a live campus population.

Limitations

• The preventive evidence is pilot-scale at a single university: 21 completed feedback surveys and 17 student evaluator reports, with the traditional questionnaire rather than a controlled alternative as the comparison.
• AURA was tested in controlled evaluations of n = 20 conversations per condition, with its priors drawn from 96 prior campus-climate conversations (467 exchanges); the dissertation notes that its exploration can be locally suboptimal in early exchanges and that the LSDE quality signal is a proxy that may reflect dataset biases.
• The mental health tools were validated on the public DAIC-WOZ dataset and 48 curated case studies, so reported performance such as Psycho Analyst's F1 of 0.929 does not establish accuracy on student disclosures in a campus service.
• The integrated prevention-to-intervention pipeline is proposed rather than deployed: no campus has run the full path from survey signals through risk detection to a service handoff.

Citation

Tang, J. (2026). New AI-Driven Tools for Enhancing Campus Well-being: A Prevention and Intervention Approach [PhD Dissertation, University of Missouri].