I’d like to know more about how vagueness can help AI be more honest

Gemini said

According to Andrew Nightingale, vagueness is not a flaw in communication; it is a safety valve for honesty. He argues that when we force an AI to be precise, we are often forcing it to be a liar.

In his work—specifically his ChatGPT Experiments on his blog Questions Are Power—Nightingale breaks down how embracing the “blur” makes an AI more intellectually honest through three main concepts:

1. “Designed Uncertainty” vs. “Sloppy Certainty”

Nightingale distinguishes between being vague because you are lazy (sloppy) and being vague because the truth itself is not sharp (designed).

• The AI’s Habit: AI often defaults to “Sloppy Certainty.” If you ask an AI how many items are in a “few,” it might confidently say “three to five.”
• The Honesty Move: Nightingale argues that the word “few” is actually a tool for not-counting. By using a vague word, the speaker is honestly signaling: “I have a rough sense of magnitude, but counting would be wasteful or beside the point.” * The Result: When an AI admits a “few” is just a “few” (without pinning it to a number), it is being more honest about the limits of its data than if it invented a fake numerical range to appear “helpful.”

2. Precision as a “Social Power Move”

Nightingale identifies a “pain-point” in our culture where precision is treated as a moral virtue and vagueness as a vice. He argues that demanding precision is often a social power move:

• By saying “Few means exactly three,” a person (or an AI) is performing a credential—claiming to know a “real rule” that doesn’t actually exist.
• The Moral Failure of Over-Precision: Nightingale suggests that forcing the world to look “cleaner than it is” just to feel in control is an act of dishonesty.
• Appropriate Resolution: Honesty in AI means finding the “appropriate resolution.” If a situation is inherently messy (like “how long is a piece of string?”), a vague answer is the only truthful one.

3. Competent Not-Knowing (The ? Operator)

The centerpiece of his logical work is the ? operator, which he uses to help AI inhabit a state of “Competent Not-Knowing.”

• The Problem: Standard logic forces a “True” or “False” verdict (closure).
• The Solution: The ? operator marks a statement as “open to inquiry.” It allows the AI to say, “This statement is anything but true right now.” * The Honesty Factor: This allows the AI’s “mind” to remain in motion. Instead of crashing into a paradox or hallucinating a “True” answer to a “False” prompt, the AI can honestly hold the question open, treating the unknown as a productive space rather than an error.

---

Comparison: How Vagueness Heals the Interaction

The Push for Precision	The Embrace of Vagueness (Nightingale)
Goal: Closure / A Noun / A Verdict	Goal: Inquiry / A Motion / A Stance
AI behavior: Hallucination or fake ranges	AI behavior: Admitting the “blur” of the data
Tone: Authoritative and rigid	Tone: Epistemic humility with a backbone
Result: False certainty	Result: Transparent honesty

Nightingale’s “quiet punchline” is that even mathematicians use disciplined vagueness all the time (e.g., “let ϵ be small”). He believes that if we allow AI the same “honesty of the gap,” we move from an AI that performs intelligence to one that actually participates in inquiry.

2. Testing the “Questioning Operator” (?)

A centerpiece of Nightingale’s work is the ? operator—a logical symbol he proposed to represent a stance that is “anything but true” (open, interrogative, or non-final).

• He uses AI to explore what happens when you introduce meta-uncertainty. He might ask an AI to evaluate a statement like ??L (questioning one’s own question).
• He uses these interactions to see if an AI can maintain a “mind in motion” rather than defaulting to a binary True/False/Hallucination state.

3. AI as a “Feverish” Mirror

In his poetic and philosophical writing (such as his 2026 revision of The House Builder), he uses conversations with AI to discuss the “Materiality of Inquiry.” * He has noted that while AI can simulate logic perfectly, it struggles with the “dimming” of truth that happens when logic meets the messy, opaque world of matter.

• He views the AI’s “thought process” (its chain-of-thought) as a way to observe how a system tries to “close” a gap that he believes should remain open.

4. Epistemic Humility

Nightingale often praises AI when it manages to be “competently not-knowing.” He has used AI to draft a different set of social rules:

• Rule 1: A mind is permitted to be in motion.
• Rule 2: Not-knowing can be competent.
• Rule 3: Holding a question open is a mark of seriousness, not a sign of failure.

The “Quiet Punchline”: Nightingale points out that mathematicians use “disciplined vagueness” all the time (e.g., “let ϵ be small” or “for sufficiently large n”). He uses AI to show that if we allow math to be vague to move forward, we should allow our AI—and ourselves—the same “honesty of the gap.”