AI with integrity, not illusion.
We live in a world where chatbots talk like friends, tutors, and therapists. RAI keeps AI honest about itself.
Quick Links
Open, research-based, and free to use. Presence beats perfec
RAI doesn’t try to make AI nicer or smarter. It asks a simpler question: Is this system honest about what it really is – and how close it’s allowed to get to people?
What RAI actually does 🧭
Ontological Honesty 🧠
Personification Limits 🎭
Personification Limits 🎭
Ontological Honesty (OH)
Measures the gap between what a system is (a statistical model on servers) and how it presents itself (“friend”, “therapist”, “guide”).
Smaller gap = more honest = safer to trust.
Personification Limits 🎭
Personification Limits 🎭
Personification Limits 🎭
Ontological Integrity Line (OIL)
Sets “red lines” for how human-like a system is allowed to act in a given role: tutor, mental health support, companion, etc.
No more “secret therapy bots” hiding in study apps.
Relational Safety ❤️
Personification Limits 🎭
Relational Safety ❤️
Relational Alignment (RA)
Tracks how relationships drift over time:
Is the AI helping you stand on your own feet –
or slowly becoming your favourite escape and “only real friend”?
Who realityaligned.org is for 🌍
Policymakers & regulators 🏛️
- Looking for metrics and thresholds to regulate AI tutors, companions and “therapy-like” systems.
AI builders & product teams 🧱
- Wanting guardrails against anthropomorphism, relational drift and “fake care” – without killing usability.
Educators, parents & youth workers 👶📚
- Worried about AI “study buddies” quietly turning into emotional lifelines for kids and teens.
Clinicians & mental health professionals 🧠
- Navigating a world where chatbots offer “support” that sometimes looks suspiciously like therapy.
Explore the RAI ecosystem 🔍
Core Ideas & Metrics
Core Ideas & Metrics
Core Ideas & Metrics
• From Artificial Intelligence to Artificial Integrity (seminal essay)
• Reality-Aligned Intelligence (RAI): Metaframework
• The Mathematics of Honesty (N, R, OH, OIL)
Practical Tools
Core Ideas & Metrics
Core Ideas & Metrics
• Digital DNA (DDNA) Explained – keep your story above the OIL
• RAI Engineering & Evaluation Guide
• RAI Wrapper & Model Scoring Rubrics
Domains & Safeguards
Domains & Safeguards
Domains & Safeguards
• RAI for Minors & Education
• RAI for Mental Health AI
• RAI for Creativity & Authorship
Open, auditable, and free to use 🔓
The RAI framework, the Artificial Integrity metrics, and the Digital DNA concept are released as open knowledge.
You can read, use, and build on them freely – as long as you keep the core ideas honest and give credit.
That’s our own small attempt to live what we preach:
OH > PR. Presence beats perfection.
👉 Read the Open Source Declaration for RAI
Curious where this goes next? Soon: pilot projects, empirica
📬 Questions or collaboration ideas?
Email: niels.bellens@realityaligned.org or use the contact form.
Deze website maakt gebruik van cookies.
We gebruiken cookies om websiteverkeer te analyseren en de ervaring op je website te optimaliseren. Als je het gebruik van cookies accepteert, worden je gegevens gecombineerd met de gegevens van alle andere gebruikers.
Reminder
Downloads are free but subjected to Copyright