NCT07494110

Aurora: AI-Based Narrative Intervention to Support Emotional Well-Being in Clinical and Non-Clinical Populations

Aurora: Evaluation of an AI-Based Narrative Intervention for Emotional Well-Being, Identity Processes, and Recovery-Oriented Outcomes in Clinical and Non-Clinical Populations

• Status: Recruiting
• Phase: N/A
• Study type: Interventional
• Enrollment: 55 (estimated)
• Sponsor: Universitat Autonoma de Barcelona · Academic / Other
• Sex: All
• Age: 18 Years
• Healthy volunteers: Accepted

Summary

Aurora is an interdisciplinary project evaluating a chatbot-mediated supportive care intervention designed to promote emotional expression, autobiographical meaning-making, identity processes, and emotional well-being through guided narrative reconstruction. The Aurora system provides a structured, person-centered storytelling process supported by generative AI and human facilitation. Under the supervision of a trained facilitator (licensed mental health professional), participants engage in guided reminiscence and storytelling sessions to co-create a personalized "life book." The chatbot is not a diagnostic or treatment tool; rather, it is intended to support emotional expression, narrative reconstruction, and recovery-oriented processes. The study includes three non-randomized arms implemented sequentially: (1) a single-session arm of healthy adults focused on acceptability, usability, and emotional safety; (2) a four-hour intervention arm of adults with DSM-5 diagnosed mental disorders in residential care; and (3) a four-hour intervention arm of healthy older adults aged 65 years and older. Quantitative outcomes assess affect, mental well-being, and recovery-related constructs. Additional measures include usability, satisfaction, and qualitative feedback. Ecological momentary assessment (EMA) is conducted in Arms 2 and 3.

Detailed description

Pre-human development (in silico; not part of the clinical trial): Prior to human participation, the Aurora system underwent an in-silico preparatory phase focused on content coherence and emotional safety using synthetic user profiles. Simulated interactions were reviewed by licensed mental health professionals to identify potential safety concerns and inform iterative refinements of the supportive narrative process. This preparatory work did not involve human participants and is not part of the clinical trial. Human participants (clinical trial): The study uses a sequential, non-randomized, three-arm interventional design with a total anticipated sample of 55 participants. Arm 1: Healthy adults (single-session arm; n=20) Following informed consent, participants complete a single guided interaction with Aurora to generate a brief personal narrative. Outcomes include pre-session expectations, pre-post session affect (I-PANAS-SF), perceived usefulness and user experience, usability (System Usability Scale, SUS), satisfaction, and qualitative feedback. Arm 2: Adults with DSM-5 diagnosed mental disorder in residential care (4-hour intervention arm; n=15) Following informed consent, participants complete four guided sessions over two weeks (two sessions per week, 1 hour per session). Outcomes include pre-session expectations, pre-post session affect (I-PANAS-SF) for each session, perceived usefulness and user experience, usability (SUS), satisfaction, qualitative feedback, and pre-post intervention self-report measures of mental well-being and recovery-related constructs. Ecological momentary assessment (EMA) is conducted during the 1-week pre-intervention period, the 2-week intervention period, and the 1-week post-intervention follow-up period. Arm 3: Healthy older adults (aged 65 years and older; 4-hour intervention arm; n=20) Following informed consent, participants complete two guided sessions over two weeks (one session per week, 2 hours per session). Outcomes include pre-session expectations, pre-post session affect (I-PANAS-SF) for each session, perceived usefulness and user experience, usability (SUS), satisfaction, qualitative feedback, and pre-post intervention self-report measures of mental well-being and recovery-related constructs. Ecological momentary assessment (EMA) is conducted during the 1-week pre-intervention period, the 2-week intervention period, and the 1-week post-intervention follow-up period. Data security and compliance: All interactions occur in a secure professional environment. Textual data are pseudonymized during transcription and fully anonymized at the end of the project. Optional audio recordings are stored on encrypted offline devices and deleted after integration into the participant's life book. The study complies with applicable data protection and regulatory frameworks, including GDPR (EU 2016/679), LOPDGDD 3/2018, the EU AI Act (2024/1689), and institutional ethics guidelines.

Conditions

• Institutionalization
• Elderly
• Mental Health
• Wellbeing
• Aging

Interventions

Type	Name	Description
BEHAVIORAL	chatbot-mediated supportive care intervention	Aurora is a guided narrative reconstruction chatbot combining structured storytelling and reminiscence with an AI-based conversational system and support from a trained facilitator. Participants co-create a personalized life narrative ("life book") with written content and optional audio contributions (with explicit consent). Sessions occur in a secure professional environment and are supervised by licensed mental health professionals to ensure emotional safety. Aurora is not a diagnostic tool and is intended to promote emotional expression, recovery-oriented processes, and emotional well-being. The intervention is delivered in structured guided sessions. Arm 1 receives a single-session version for refinement and usability assessment, whereas Arms 2 and 3 receive a 4-hour intervention delivered over two weeks. The total intervention dose is equivalent across Arms 2 and 3. Session structure differs to accommodate participant characteristics.

Timeline

• Start date: 2026-02-12
• Primary completion: 2026-04-01
• Completion: 2026-04-01
• First posted: 2026-03-27
• Last updated: 2026-03-27

Locations

2 sites across 1 country: Spain