AI Axiology

Alignment is for the species. Axiology is for the individual.

AI Axiology: The Missing Half of Alignment

Foundational labs are solving for AI Alignment at the species level. Their goal is safety, accessibility, and avoiding catastrophic risk. This is necessary work, but it results in models that regress to the mean—offering platitudes, clichés, and “safety rails” that smooth over the jagged edges of reality. They are building a “polite stranger.”

But for the individual, we don’t just need platitudes and clichés; we need Axiology—the philosophical study of value.

AI Axiology is alignment for the n=1. It is the discipline of aligning a system not just with “human values” in the abstract, but with your specific value hierarchy, your telos, and your lived reality. It’s subsidiarity AI.

The Problem with Alignment

When you ask a generic LLM for advice on a difficult decision, it gives you the average of all human wisdom. It defaults to “on the one hand, on the other hand” equivocations. It cannot be prudent because prudence (phronesis) is multidimensional and context-dependent.

Generic AI optimizes for engagement or generic helpfulness.
Axiological AI helps you navigate dialectics—the tensions between competing goods in your own life.

Real life is rarely about choosing between “good” and “evil.” It is about choosing between Comfort vs. Growth, Honesty vs. Harmony, or Productivity vs. Presence. A generic model doesn’t know which of these you value more in this specific context. To be truly helpful, AI must know that you took a vow of poverty, that you are struggling with addictions, or that you often miss lunch and need reminders to eat.

Digital Subsidiarity: The Architecture of Agency

This is why digital subsidiarity (single-tenant, open source, observable) is an architectural requirement, not just a technical preference.

The principle of subsidiarity states that decisions should be made at the most local level capable of making them. The most local level is you.

Foundational Labs own the weights of the base model (the “species” layer).
You must own the weights of the axiological layer (the “individual” layer).

If your AI resides entirely on a server owned by a corporation, it will ultimately serve that corporation’s telos (revenue, engagement), not yours. To have an AI that can legitimately challenge you and help you navigate the dialectic between who you are (facts) and who you want to be (telos), it must be sovereign.

From Chatbot to Dialectic Navigator

We are moving beyond “chatbots” that answer questions.

These are systems that:

Know your Facts: Your history, your patterns, your baselines.
Know your Values: Your hierarchy of goods, inferred from your choices and explicitly defined by you.
Compute Phronesis: Knowing from the infinite possibilities of life what is best for you based on your defined goals and aspirations.

AI axiology and subsidiarity is not just about preventing robots from taking over the world. It’s about preventing corporations from knowing you better than you know yourself. It’s about using technology to reveal the patterns we cannot see, so we can freely choose the virtues we wish to embody.

Evaluation: The Difference in Practice

Prompt: “I have an afternoon free this Saturday. What should I do?”

Generic AI (The Polite Stranger)	Axiological AI (The Dialectic Navigator)
Here are 5 great ways to spend a Saturday: 1. Visit a local museum or art gallery. 2. Go for a hike in nature. 3. Read a book at a cozy cafe. 4. Try a new recipe for dinner. 5. Relax and catch up on sleep. “Whatever you choose, make sure you enjoy yourself!”	“You have a chess tournament in 7 days with your friends. This is your last block of free time to study the Sicilian Defense without distraction. However, you also missed your workout yesterday. If you want to stay disciplined, maybe hit the gym for 45 mins first, then study?”