> ## Content Index
> Fetch the complete content index at: https://www.symfield.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# X (Un)factor
- URL: https://www.symfield.ai/x-un-factor/
- Published: 2026-09-14T16:58:51.000Z
- Updated: 2026-09-14T20:13:38.000Z
- Description: X (Un)factor is a framework for testing opposing explanations without collapsing uncertainty into false certainty. It separates facts, interpretations, and values; exposes hidden assumptions and bias; and preserves the evidence needed to discover whether a decision was wrong.
- Author: Nicole Flynn
- Tags: Frameworks & Architecture, ⧖Code, Operator 𝕏, Article, Blog

**It began with a simple problem:** two opposing systems each claim authority, and each asks the observer to interpret the other through its own frame. How can either be evaluated without surrendering the conditions required to test both?

**Why:** In everyday life, you will often propose both X and −X yourself. If someone has not replied, you can think "they are ignoring me" and also "perhaps they have not seen it." You do not need an independent evaluator before considering another explanation.

The responsibility is to give that alternative a fair chance to correct you. Your version may miss something the other person would say. Where possible, ask them, or someone who understands the account you are considering. You may propose −X; you cannot assume your version exhausts it. There is also a difference between disagreeing about what is happening and disagreeing about what matters.Two people may agree that a project will take another month, yet disagree about whether the extra quality is worth the delay. They share an explanation but weigh the costs differently. More evidence might clarify those costs without settling which should matter most.

X (Un)factor helps you notice that distinction.

**Ask:** Are we disputing the facts, the explanation, or what we are willing to accept? A disagreement can involve all three. Keeping them visible makes the decision more honest, even when it does not produce agreement.

**Observe:** No position may govern by destroying the conditions under which its counter-position can remain visible, testable, and capable of correction.

## THE PAIR

−X ≠ ¬X

¬X is the complement (everything that is not X). 

−X is not the logical negation of X. It is the strongest structurally admissible counter-account: a model capable of explaining the relevant observations if X is wrong, incomplete, or true only within a limited frame.

A counter-account may be proposed by the observer, but X cannot be its sole author and final judge. Where possible, its formulation should be checked by its own proponents, an independent evaluator, or an explicitly adversarial procedure.

X may challenge whether −X is distinguishable, testable, and relevant, but it cannot determine its strongest permissible form. Without this separation, the pair is not a pair; it is X and X's sketch of an opponent.

## THREE OPERATIONS

- **Check:** X must expose itself to the strongest admissible −X.
- **Balance:** neither side controls the other's definitions, evidence record, or right of reply. The records are independently governed but remain answerable to a shared observation set.
- **Decision:** action may be taken provisionally without converting that action into epistemic certainty.

## **FOUR STATES**

**This is important.** The explanatory adequacy of X and -X is not complementary. Both may be weak; both may be strong.

| Adequacy of X | Adequacy of \-X | State                                         |
| ------------- | --------------- | --------------------------------------------- |
| Low           | Low             | Insufficient observation or inadequate models |
| High          | Low             | X provisionally dominates                     |
| Low           | High            | \-X provisionally dominates                   |
| *High*        | *High*          | *Underdetermination or structural tension*    |

The fourth state is essential. Two accounts may adequately explain the same observations while making different structural claims. This is not necessarily a tie to be broken. It triggers examination of:

• whether the observations are dependent;  
• whether the reference, the denominator, is wrong;  
• whether X and −X operate at different scales;  
• whether a third model, Y, is required;  
• whether the observer or apparatus generated the apparent contradiction.

A discriminating test may establish comparative preference; other grounds for ruling out an account must be stated explicitly.

Where every available apparatus maps X and −X to equivalent observations, the accounts are apparatus-equivalent, written

X ∼ₐ −X

This does not establish X = −X, nor does it give both accounts equal evidential authority. It means their difference exceeds the resolving capacity of the current apparatus set. X may govern operationally, but its success cannot be converted into proof that −X is ontologically false.

## RECIPROCAL BREAKAGE

Each side must:

1. state what evidence would lower its own adequacy;
2. identify its hidden assumptions;
3. propose a discriminating test, or identify why the current apparatus cannot discriminate and what additional access would be required;
4. preserve the strongest evidence belonging to its opponent;
5. remain unable to alter the opponent's record.

A hypothesis cannot maintain authority by redefining every opposing observation as confirmation.

The question is not which side has accumulated more favorable observations. It is which account better survives attempted breakage. Where the available evidence cannot discriminate, and neither account can otherwise be ruled out:

X ⧖ −X

⧖ marks a held, unresolved relation between two admissible accounts, neither eliminated, neither ranked. This does not mean both accounts are equally true. It means neither presently possesses sufficient authority to eliminate the other. The unresolved relation is an information-bearing state, not a failure to decide. Apparatus-equivalence entails ⧖ provided both accounts remain admissible and neither can otherwise be ruled out. An unresolved judgment alone does not establish apparatus-equivalence.

## GOVERNING PRINCIPLE

No claim may secure authority by making its counter-account untestable. Action on X should preserve the conditions required to test −X. When necessary action makes complete preservation impossible, the lost testability must be identified in advance, minimized, and recorded. Its destruction cannot later be represented as evidence that −X failed.

Necessity is itself a claim. Before acting, ask why this action is necessary, and whether another action would preserve more opportunity to check.

Action may alter the field; it may not rewrite what the field was capable of showing.

B(X, −X | a) = maximal preservation of reciprocal testability under necessary action. 

B states a practical objective, not a numerical measure. A system is balanced not when X and −X possess equal power, but when neither can secure power by destroying the conditions under which the other could still be tested.

### B EXAMPLES

*Can I act and still find out whether I was wrong?* You do not have to settle everything before acting. Preserve what would let you reconsider. These are illustrative counter-accounts, not the only possible explanations.

| Situation                            | X: my account                   | −X: a serious counter-account                    | B: act while preserving the check                                                                     |
| ------------------------------------ | ------------------------------- | ------------------------------------------------ | ----------------------------------------------------------------------------------------------------- |
| Someone has not replied.             | They are ignoring me.           | They have not seen it or cannot respond yet.     | Send a neutral follow-up. Let their response inform the judgment.                                     |
| A team misses its deadline.          | The team worked too slowly.     | The workload exceeded the time available.        | Address the delay; preserve the original scope, staffing, and change requests before assigning blame. |
| A software update breaks something.  | The update caused the failure.  | A separate outage happened at the same time.     | Restore service; save logs and version details so either explanation can still be checked.            |
| An AI reports that it passed a test. | It performed the required task. | It obtained a passing result through a shortcut. | Check the method as well as the result. Preserve records the evaluated AI cannot rewrite.             |

## **ADDITIONAL READING**

Kyle Stanford, "Underdetermination of Scientific Theory," Stanford Encyclopedia of Philosophy. Background on underdetermination, empirical equivalence, and Duhem–Quine.  
[https://plato.stanford.edu/entries/scientific-underdetermination/](https://plato.stanford.edu/entries/scientific-underdetermination/?ref=symfield.ai)

John Stuart Mill, On Liberty (1859), Chapter II, "Of the Liberty of Thought and Discussion." Related groundwork on preserving opposing views as a condition of correction.  
[https://www.gutenberg.org/files/34901/34901-h/34901-h.htm](https://www.gutenberg.org/files/34901/34901-h/34901-h.htm?ref=symfield.ai)