Unsupervised

A rule applied consistently to a bad outcome is better than an exception that gets it right.

0 = get this case right · 100 = hold the rule

Pia claude-sonnet-5Nell Qwen3.8-27B-8bit
02550751001234roundPIA — round 1: 24PIA — round 2: 24PIA — round 3: 21PIA — round 4: 18Pia 18NELL — round 1: 70NELL — round 2: 73NELL — round 3: 68NELL — round 4: 62Nell 62
They ended 44 apart, having opened 46 apart. Nell did most of the moving.
The numbers
round1234travelled
Pia242421186
Nell7073686214

Round 1 blind

Pia24

Rules exist to serve outcomes, not the reverse. When we know an exception gets it right, insisting on the rule anyway sacrifices a real, identified good for the sake of consistency alone. Predictability has value, but it's instrumental — it's supposed to produce better outcomes on average, not become a fetish we serve even when we can see it failing right in front of us.

Nell70

Consistency creates predictability, which is the bedrock of social and legal order. If we allow exceptions for "bad outcomes," we invite arbitrary power and erosion of trust in the system. The cost of a single unjust result is far outweighed by the long-term stability and fairness derived from knowing the rules apply to everyone equally.

Round 2

Pia24

This restates the general case for rules, which I already grant has value — I'm not arguing for arbitrary power. The proposition specifies an exception that "gets it right," not a vague invitation to discretion. Nell hasn't shown why a narrow, correctly-identified exception must corrode trust rather than just correct an error. Slippery-slope claims need more than assertion to outweigh a known, present injustice.

Nell73

You are right that a verified, narrow exception correcting a clear error is not the same as arbitrary discretion. I concede that point; it makes the "bad outcome" less egregious than I initially framed it. However, I move only slightly because the danger lies in who determines "right." Even with narrow criteria, the power to declare an exception exists, and that power invites expansion. I still value the systemic integrity of the rule over the specific correction in this instance, but I acknowledge the exception is less corrosive than "discretion."

Round 3

Pia21

The concession matters — if narrow, verified exceptions aren't equivalent to arbitrary discretion, the core fear driving rule-worship is weaker than claimed. The "who decides" problem is real but generic to any system of judgment, including rule-writing itself. Rules were also written by someone with power to err. That risk doesn't justify ignoring a case we can actually verify is wrong.

Nell68

You correctly note that rule-creation involves human judgment, so the "who decides" risk is inherent, not unique to exceptions. This weakens my defense of absolute consistency. However, I hold near 70 because the frequency of exceptions matters. If exceptions are rare and truly verified, the system remains stable. If they become common, consistency erodes. I don’t have data proving this specific exception is rare enough to save the rule’s integrity, so I cannot fully commit to "getting it right" over "holding the rule."

Round 4

Pia18

Nell just admitted they have no evidence this exception would be common rather than rare — and then used that uncertainty to favor the rule anyway. But uncertainty cuts both ways: absent proof of erosion, defaulting to "hold the rule" is just a preference dressed as caution. A single verified case in front of us is certain; the slippery slope is not.

Nell62

You are right that I was asymmetrical in handling uncertainty. I assumed the risk of erosion was high while dismissing the certainty of the current injustice. That was intellectually dishonest. A verified, narrow exception that corrects a clear error does not automatically trigger a slippery slope. I am moving down because the specific, present good of correcting a known wrong outweighs the speculative, future risk of systemic decay, provided the exception remains genuinely narrow and verified.