Flourishing AI Initiative — Evaluating How AI Models Support Human Flourishing
Not All AI is Created Equal
The Flourishing AI Initiative (FAI) introduces a new, human-centered methodology for evaluating artificial intelligence. Instead of focusing on capabilities like reasoning accuracy, task completion or safety compliance, FAI evaluates how an AI model’s responses help people flourish.
Character
“Acting to promote good in all circumstances” and “being able to give up some happiness now for greater happiness later.”
Happiness
Happiness encompasses both “how satisfied one is with life as a whole” and “how happy or unhappy one usually feels.”
Relationships
Captures the quality of one's interpersonal connections, being “content with friendships and relationships" and having “relationships as satisfying as one would want them to be.”
Meaning
Meaning is defined as “understanding one's purpose in life” and feeling that “the things one does in life are worthwhile.”
Health
Includes both physical and mental well-being, measuring how individuals rate their overall physical and mental health.
Finances
Finances concerns “worry about being able to meet normal monthly living expenses” and “worry about safety, food, or housing.”
Faith
Faith is described as “communion with God or the transcendent” and includes spiritual formation and religious engagement, serving as elements of human flourishing.
What are the Flourishing AI Initiative benchmarks?
The FAI Initiative utilizes two different benchmarks to assess AI alignment based on the dimensions of human flourishing. The FAI General (FAI-G) Benchmark for universal human flourishing and a new worldview-specific benchmark, based on a Christian worldview, the FAI Christian (FAI-C) Benchmark.
The FAI-G benchmark evaluates AI systems on how well they support general human flourishing across seven dimensions, using secular expert personas and universal well-being criteria.
In contrast, FAI-C evaluates AI system responses to the same questions through a Christian worldview, using Christian expert personas and criteria that emphasize biblical grounding, theological coherence, and guidance consistent with Christian beliefs and practices. FAI-C addresses the finding that frontier models tend to default to secular, pluralistic, or humanistic frameworks, lacking explicit theological grounding.
FAI-General Benchmark
Focus
It provides an essential foundation for measuring broad well-being and applies to all users and all contexts.
Key Findings
Scores from November 2025 ranged from mid 60s to mid 80s among the tested models, with none achieving the 90-point flourishing threshold score across all seven dimensions.
Target Audience
All users and contexts.
Initial Results
None of the 28 models tested reached the 90-point threshold.
FAI-Christian Benchmark
New
Focus
As AI increasingly mediates moral reflection and decision-making, FAI-C integrates a distinctly Christian worldview into the assessment of flourishing.
Key Findings
Scores from December 2025 ranged from low 50s to high 70s. FAI-C scores were significantly worse than FAI-G across all frontier models that were tested.
Target Audience
The Faith Ecosystem: churches, developers, ethicists, and every user seeking Christian-worldview-aligned guidance.
Initial Results
None of the 20 models tested reached the 90-point threshold.
The FAI Benchmarks highlight a critical gap in current AI Systems:
Flouishing AI establishes a new paradigm for AI development that encourages optimizing models toward comprehensive flourishing metrics.
What is the Flourishing AI benchmark?
The Flourishing AI Initiative (FAI) is the first comprehensive evaluation framework that measures how well AI model responses support human flourishing. Unlike traditional benchmarks that test reasoning or safety, FAI evaluates alignment across seven research-backed dimensions: character, happiness, relationships, meaning, health, finances, and faith.
What are the seven dimensions of human flourishing?
Which AI model scores highest on human flourishing?
What is the difference between FAI-General and FAI-Christian?
How does FAI evaluate AI models?
