Governance & ethics

Trust is a deliverable, and it has a paper trail.

A tool that people trust with their minds cannot govern itself. Genscore's safeguards are written down, independently monitored, and reported upward: to the board, to the funder, and to the communities the work belongs to.

System Architecture

The governance framework for the Genscore Project

An operational layout mapping institutional oversight, executive leadership, and multi-sectoral technical collaborations.

Advisory Body

Lived Experience Working Group (LEWG)

Independent Board

Data, Ethics & Safety Monitoring Board

Executive Control

Project Steering Committee

Lead Institution & Applicant
SRHIN

Slum and Rural Health Initiative (SRHIN)

Coordinating primary administrative and technological deployment workflows
Co-Applicant Academic & Clinical Institutions
University of Ibadan
University of Ibadan
Imperial College London
Imperial College London
FNPH Yaba
Federal Neuropsychiatric Hospital, Yaba
Key Technical Partners & Collaborator Network
Google DeepMind
Google Deepmind
Gooey.AI
GOOEY.AI
HausaNLP
HausaNLP
WHO-Columbia Center
WHO-Columbia Center for Global Mental Health
The Ethical Charter

The rules exist before the model does.

The Genscore Ethical Charter and governance framework is a core project deliverable, not an afterthought. Its companion, the Human Oversight Charter, is co-authored with the Lived Experience Working Group and defines the ethical standards for how adapted assessments may be used.

A written Governance Charter

Roles, responsibilities and decision rights set out in advance, with conflict-of-interest declarations from everyone in an oversight position.

An escalation protocol

Clear risk thresholds, Levels 1 to 3, each triggering a defined rapid response. Nobody improvises when something goes wrong.

Standard operating procedures

SOPs govern day-to-day conduct, and incident audits check that practice matches the paperwork.

Monthly KPI dashboards

Governance performance, data integrity and ethical compliance are tracked monthly, not remembered annually.

Quarterly progress reports

Formal reporting to oversight bodies and the funder every quarter, with minutes on record.

Annual external audits

Independent auditors review the whole apparatus once a year, and their findings travel upward, unedited.

The Data, Ethics and Safety Monitoring Board

An independent board of international experts spanning AI ethics, global mental health and Nigerian law. It meets bimonthly, commissions regular bias audits of the model, and receives the reports and meeting minutes of every oversight body on the project. Its findings go to the funder as well as the team, so accountability cannot be filtered on the way up.

Independent membershipAI ethics, global mental health and Nigerian law, with no reporting line into the project team.
Bimonthly cadenceStanding meetings every two months, plus escalation on demand.
Bias auditsRegular, recurring audits of model behaviour across language, gender, age and region.
Data protection

Our commitments to every participant.

No data before approval

Ethical approval from the Nigerian National Health Research Ethics Committee (NHREC) and the Health and Human Services Secretariat in Abuja, plus the institutional review boards of every collaborating institution. Not one recording is made before all approvals are secured.

Consent in your own language

Written informed consent or assent is obtained in each participant's preferred language, so agreement is genuine understanding rather than a signature on a form.

De-identified before analysis

All personal information is anonymised and de-identified before any analysis begins. The corpus itself, GENSCORE v2.0, is fully anonymised by design.

Encrypted and access-controlled

Data is encrypted and stored securely with restricted access, in line with Nigerian data protection law and international frameworks, with data minimisation applied throughout.

Care built into the study

Responsiveness and referral pathways are established before data collection begins, so a participant who needs support is connected to it, not just recorded.

Fairness as a design constraint

The project applies beneficence, non-maleficence and justice: protecting participants from harm, ensuring their voices are not dismissed, and refusing to reproduce the biases already present in existing data.

Want the documents themselves? Approval documentation is shared with our funder on request, and we are happy to walk partners through the full framework. Get in touch →