Using AI to Improve Natural Intelligence — The Assent Inventory Applied v1.0

 

Using AI to Improve Natural Intelligence — The Assent Inventory Applied v1.0

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.

---

The Beginning

The beginning is still Th1: everyone wants happiness. By the corpus’s chain, happiness lies wholly in correct assent. Natural intelligence, in this sense, is the rational faculty that assents. Improving it means improving that faculty’s condition, and Excerpt 7a names the only mechanism for doing so: each refused false impression makes that type of impression weaker and rarer, and each assent makes it stronger.

AI Output Is an Impression

AI output is an impression. It arrives from outside, it is propositional, and it claims the world is a certain way. So every AI reply is one more occasion for assent, which puts it inside the inventory’s grid. The Corrective Layer adds the key fact: the model’s dangerous failure is the coherent falsehood, and only an agent standing outside the mechanism can catch it. That agent is you, and catching errors is the exercise that strengthens the faculty.

The Grid Applied to AI Use

• Before. Formulate standing propositions in advance (clauses c and d). Fluency is not truth (C5). I am the checker, not one more output stream (C1, C2). I must know the foundations myself, because tracing a claim to its warrant needs a floor to reach (C4). The AI can’t be that floor for you.
• At. Treat each factual claim as withheld until it is traced to its source. Excerpt 7 says any impression can be questioned; the habit is to actually question it. Watch the value impressions too: “this is excellent” is a pull toward assent, and irritation at an error is clause (a) failing.
• After. When an error surfaces after you’ve assented, recovery is simply the next choice: correct it and move on, with no carryover. When you have checked correctly, clause (e) applies: assent to that.
• Across a life. Every coherent falsehood you refuse trains discrimination. Every “sounds right, so it is” that you accept trains the opposite. This is the grid’s open cell, fact across a life. The corpus doesn’t state the effect for factual impressions, so here it stands as the practice’s own test, not as doctrine.

Use AI to Exercise the Faculty, Not to Replace It

• Ask for arguments and judge each line. Don’t ask for conclusions to adopt.
• Ask the AI to find the weakest link in your own chain, then assess whether it really is weak.
• Require sources for every claim, and open them. Your URL for the Nine Excerpts is what exposed that line 1 was Sterling’s own wording.
• Keep the ratification sequence. “Nothing enacted without explicit ratification” is the discipline of assent turned into a working procedure. Today’s corrections were that procedure working: the misplaced id, the labels on derived lines, the Excerpt 10 attribution.

The Rule Itself Applies Here Too

The rule itself applies here too. Desire only correct assent. That means not desiring a polished document or a productive session, both of which are externals. The document is an object of aim, pursued with reservation. The good lies in your judgment of it.

---

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.