Sector
Healthcare
Problem to solve
The client was developing an AI-based platform that generates feedback to support psychotherapeutic work. Since it involves sensitive health data, people in potentially vulnerable conditions and content that enters the therapeutic relationship, a structured assessment of the ethical impact and risk profiles was needed already at an early stage of development, with time and costs proportionate to this phase.
Goal
Provide the client with a recognizable and immediately usable ethical assessment, built on the six dimensions of Immanence’s VERI framework (Governance & Accountability, Data Quality & Protection, Technical Robustness & Safety, Fairness/Non-discrimination & Inclusion, Transparency, Human Rights/Social & Environmental Sustainability), with a risk score for each dimension and a set of operational recommendations.
Methodologies and frameworks
- Immanence’s proprietary VERI framework, six ethical dimensions
- Guided assessment on ARIA, with constant support from the Immanence team
- Risk score per dimension
- Analysis and final report with operational recommendations by Immanence
Why it matters
- In a clinical setting, AI errors affect patients and the therapeutic relationship
- Health and mental health data are among the most sensitive categories to protect
- Assessing risks at an early stage makes it possible to integrate their mitigation into the design, at lower cost
- Prepares the product for the requirements of the AI Act and healthcare regulations
Methodologies and frameworks adopted
- Participatory process: up to 8 semi-structured interviews (30 min) with selected stakeholders
- Analysis of internal documentation
- Alignment workshop (1 hour, max 10 participants) on orientations, needs and expectations
- Formalization and validation of the Ethical Code (document + summary sheet)
- References: international AI Ethics principles, AgID Guidelines, AI Act
Takeaways
- Assessing ethics at an early stage costs less than fixing it once the product is finished;
- A guided platform like ARIA, combined with human expertise, makes assessment scalable without losing rigour;
- A score per dimension makes risks readable and comparable over time;
- In a clinical setting, the professional’s supervision remains the central safeguard.
Client’s feedback
“Once my app had to go out into the world, the questions began: is the data safe? Is it compatible with my users’ professional code of ethics? Am I doing enough on ethics and governance? AI couldn’t answer those questions, so I turned to Immanence. What began as an assessment became a real revolution in how I think about artificial intelligence and its impact. Having them as a sparring partner made the difference between a product thrown together at random and an app that truly cares about its users, safe and compliant with GDPR and the AI Act.”