01 / Human behaviour
People are already bringing deeply personal concerns into AI conversations.
These interactions create a new context for reflection, emotional support, health information and help-seeking.
GCAH · Independent · Interdisciplinary · Public interest
Human evidence for an AI-shaped mental-health future.
The Global Centre for AI-Mediated Mental Health Research & Safety is a proposed research institution connecting clinical knowledge, privacy-preserving science, cultural intelligence and responsible AI development.
One centre. Multiple disciplines. A shared evidence standard.
GCAH / Founding interface / 01
01 Project goal
Build the evidence base that AI-mediated mental health now requires.
The Centre brings research, clinical practice, technology and public-interest governance into one coherent institution, designed to understand how people use conversational AI in moments of reflection, loneliness, uncertainty and psychological distress.
The goal is to make AI-assisted psychological support more scientifically grounded, clinically informed, culturally responsive, privacy-preserving and publicly accountable.
02 Why it matters
A new human environment is forming faster than its research standards.
02 / Cultural meaning
Distress is expressed through language, culture, identity and lived context.
Research must recognise indirect, multilingual and culturally specific expressions that generic systems can overlook.
03 / Public infrastructure
Clinical insight and technical capability need a trusted place to meet.
A dedicated centre can establish shared methods, evaluation standards and governance that serve people across institutions.
The founding research question
How can privacy-preserving analysis of human–AI interactions, combined with consent-based research, improve the safety, accessibility and evidence base of AI-assisted mental-health support?
03 Centre architecture
Six connected capabilities form the institutional core.
Each capability answers a distinct need. Together they create a repeatable research system from observation to public benefit.
Measurement science
Define reliable constructs for distress, emotional reliance, supportive responses, harmful responses, help-seeking and changes over time.
Cultural intelligence
Develop multilingual taxonomies and culturally informed ways to recognise sensitive psychological language.
Clinical evaluation
Bring psychologists, psychiatrists and specialist clinicians into the design of evidence, evaluation and research translation.
Privacy infrastructure
Create secure analytical environments, aggregate research outputs, synthetic resources and consent-based study pathways.
AI safety research
Study patterns associated with helpful, unhelpful and harmful AI behaviour, including emotional reliance and response quality.
Public-interest governance
Align ethics, data protection, lived experience, community representation and scientific independence.
04 Research programme
A structured programme from institutional foundation to public translation.
Governance & research design
Establish the ethical framework, legal analysis, data-protection architecture, research protocols and interdisciplinary decision structure.
Measurement & taxonomy
Create measurable constructs, multilingual categories, evaluation rubrics and clinically reviewed research definitions.
Privacy-preserving analysis
Study aggregate patterns in AI-mediated mental-health interactions through controlled, pre-registered analytical methods.
Longitudinal human research
Examine how different forms of AI interaction relate to wellbeing, loneliness, emotional reliance, help-seeking and service engagement over time.
Intervention evaluation
Design and assess focused forms of support such as psychoeducation, reflective exercises, service navigation and clinician-supervised tools.
Public-interest translation
Produce research papers, public reports, evaluation standards, multilingual resources, benchmark materials and policy recommendations.
06 Partnership model
The Centre is a shared table for disciplines that rarely sit together.
Its strength comes from structured collaboration across science, care, technology, governance and lived experience.
People with lived experience
Clinical psychologists
Psychiatrists
Suicide-prevention specialists
Research ethicists
Privacy experts
AI-safety researchers
Multilingual specialists
Child safeguarding experts
Public-interest organisations
Academic institutions
Technology partners
07 Public value
Research outputs designed to strengthen an entire field.
Peer-reviewed research and public reports
Clinically informed evaluation methods
Multilingual and cross-cultural taxonomies
Privacy-preserving research resources
Safer pathways between AI and human support
Governance frameworks for responsible research
Founder / Concept originator / 2026
08 Founder
Pezhman Farhangi also known as Peter Far
A London-based independent researcher, creator and humanitarian practitioner. His background includes psychology study, safeguarding training, crisis and helpline coordination, refugee advocacy, GDPR-aware casework, and the development of public-interest research and technology frameworks. GCAH brings those strands into a focused institutional proposal for responsible AI-mediated mental-health research and safety.
Psychology BSc · in progress
Safeguarding
Humanitarian support
GDPR-aware casework
Public-interest research
09 Research contact
Discuss research participation, governance, partnership or funding.
GCAH is currently a founding proposal, not a clinical service. Enquiries from research institutions, clinicians, ethics and privacy specialists, public-interest organisations, funders and responsible technology partners are welcome.
Formal and research enquiries
info@iusmusic.com
Project page
iusmusic.com/GCAH.html
Website operator
I/US Music Ltd · Company no. 17318154
Registered office
8, 88 London, SE6 4HW, United Kingdom
Authorship & design provenance
Original concept presentation, information architecture, visual system and written expression © 2026 Pezhman Farhangi (AKA Peter Far). All rights reserved.
GCAH / PF / C2026 / V1.2