What is good sense in Hume's essay?

The supervising faculty: reason keeping the judge's other capacities honest — noticing purpose, consistency, and context, and guarding against one's own errors. A judge with delicate perception but no sense of when to trust it still judges badly.

How does calibration measure good sense?

On performance items you attach a confidence level (95%, 70%, or 50%) to each answer. Plotting confidence against actual accuracy yields a calibration curve, and Brier scores summarize it: well-calibrated judges are right about as often as they claim to be. Overconfidence and underconfidence both show up as measured distances from the diagonal.

HUME'S CRITERIA · GOOD SENSE

Good sense

Good sense is Hume's supervising faculty — reason, standing behind perception and checking its work. The other criteria can all misfire without it: delicate ears with no judgment about when to trust themselves, practice that rehearses a bias into a habit, breadth that collects exposure without weighing it. Good sense is the part of a judge that knows when their own verdict is reliable and when it isn't.

That sounds unmeasurable — a faculty about faculties. It isn't. Decision science has a precise, boring name for it: calibration. A judge is well calibrated when their confidence matches their accuracy — when the answers they'd stake 95% on are right about 95% of the time, and the coin-flip feelings are right about half the time. Overconfidence and underconfidence are both failures of exactly the thing Hume was pointing at: knowing the reliability of your own judgment.

So the gym measures it. On performance items — trials with objectively right answers, like the Delicacy Trials — you attach a confidence level to each answer: 95%, 70%, or 50%. Plot claimed confidence against actual accuracy and you get a calibration curve; a Brier score summarizes how far you sit from the diagonal where confidence and reality agree. The result is Hume's most abstract criterion turned into arithmetic: a curve you can read, and one number for how far it sits from the line.

One honesty note, because it's the house rule: confidence input never inflates or weights your scores — it's measured against your accuracy, never blended into it. A confident wrong answer costs you calibration; it cannot buy you points. The gym opens with the Prestige Test; the full measurement rules live in the methodology.

Questions, answered straight

What is good sense in Hume's essay?: The supervising faculty: reason keeping the judge's other capacities honest — noticing purpose, consistency, and context, and guarding against one's own errors. A judge with delicate perception but no sense of when to trust it still judges badly.
How does calibration measure good sense?: On performance items you attach a confidence level (95%, 70%, or 50%) to each answer. Plotting confidence against actual accuracy yields a calibration curve, and Brier scores summarize it: well-calibrated judges are right about as often as they claim to be. Overconfidence and underconfidence both show up as measured distances from the diagonal.