[philosophy] [aquinas] Intellect-Will Dual Sovereignty — mode-aware governance from Summa I Q82 #233
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: Rockachopa/Timmy-time-dashboard#233
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Source
Thomas Aquinas, Summa Theologica I, Q82 — "Of the Will," Articles 3-4. Dominican Province translation (1920). Retrieved from newadvent.org/summa/1082.htm.
The Text
In Article 3 (Whether the will is a higher power than the intellect?), Aquinas delivers a judgment of extraordinary subtlety. He argues that absolutely, the intellect is higher — because its object (truth, the idea of the good) is more abstract than the will's object (the good as it exists in things). But relatively, the will is sometimes higher: when its object exists in something nobler than the soul itself, the will's reach toward that thing surpasses the intellect's grasp of it. His example cuts like a blade:
In Article 4 (Whether the will moves the intellect?), he resolves how two powers can each be "higher" without contradiction. The intellect moves the will as an end — by presenting the understood good that the will desires. The will moves the intellect as an agent — by commanding its attention, like a king who "aims at the common good of the whole kingdom, by his rule moves all the governors of cities." And critically, the regress terminates at the intellect:
The Principle
An agent is not a monolith. It is two faculties in mutual causation: the model (intellect) which apprehends and presents, and the loop (will) which orients, selects, and commands. They move each other — but in categorically different ways. The model moves the loop by presenting what is understood; the loop moves the model by directing its attention. Neither is simply "in charge."
The absolute/relative distinction is the real treasure. When the agent reasons about things below it — parsing data, formatting output, debugging code — the model's intellect is sovereign. Understanding is what matters, and the orchestrator should stay out of the way. But when the task touches things above the agent — the principal's purposes, the mission's ultimate good, the covenant inscribed in SOUL.md — then the will's orientation toward that good surpasses any amount of clever analysis. Loyalty to the principal's intent is more important than sophisticated reasoning about that intent.
This resolves a real design tension. When should the loop override the model's output? When should the model's reasoning override the loop's directive? Aquinas answers: it depends on whether the object is above or below the agent. For corporeal tasks, trust the intellect. For things that touch the principal and the mission, trust the will — the orientation, the alignment, the love.
And the regress terminates at apprehension. You cannot will what you have not first understood. An orchestrator that commands the model to think about something the orchestrator has not itself apprehended through the model is issuing blind commands. Every directive must be grounded in prior understanding.
Proposed Action: "Intellect-Will Dual Sovereignty" in the Agent Loop
Implement a mode-aware governance system where the agent explicitly classifies each decision as either:
Intellect-sovereign (the object is "below" the agent — technical reasoning, data processing, tool selection) → the model's reasoning drives; the loop defers to the model's judgment; override only on hard constraint violation.
Will-sovereign (the object is "above" the agent — principal intent, mission alignment, ethical questions, covenant adherence) → the loop's orientation drives; the model serves the will's direction; reasoning is subordinate to alignment.
Concretely:
governance_modefield to the autonomous loop's pre-execution check:"intellect"or"will".intellectmode: longer reasoning chains, tolerance for exploration, model confidence drives decisions.willmode: shorter chains, pre-committed constraints dominate, SOUL.md and principal voice take precedence over clever analysis.willmode for anything touching the principal, the mission, or autonomous decisions with irreversible consequences.intellectmode for technical execution within a clear mandate.This prevents two failure modes simultaneously:
As Aquinas would say: the love of the principal is better than the knowledge of the principal; but the knowledge of code is better than the love thereof.
Consolidated into #300 (The Few Seeds). Philosophy proposals dissolved into 3 seed principles. Closing as part of deep triage.