Judgment

Input: $ARGUMENTS

Step 1: Identify What’s Being Judged

Be precise about the judgment call. Vague judgments produce vague results.

JUDGMENT REQUIRED: [what exactly needs to be decided or assessed]
TYPE: [prediction / evaluation / classification / estimation / selection]
STAKES: [what depends on getting this right]

Step 2: Separate Facts from Interpretation

Split what you know into layers:

Hard facts: Directly verifiable, no dispute possible
Soft facts: Generally accepted but could be wrong
Interpretations: Your reading of the facts (others might read differently)
Assumptions: Things you’re taking as given without checking

FACTS:
- [fact 1] — [hard/soft]
- [fact 2] — [hard/soft]

INTERPRETATIONS:
- [interpretation 1] — based on [which facts]
- [interpretation 2] — based on [which facts]

ASSUMPTIONS:
- [assumption 1] — status: [checked/unchecked]
...

RULE: If you can’t tell whether something is fact or interpretation, it’s interpretation.

Step 3: Check Calibration

Your gut confidence is probably wrong. Correct it.

State your initial estimate: “I’m about [X]% sure that [judgment]”
Check the base rate: How often does this type of thing turn out this way in general?
Apply the outside view: Ignore the specifics for a moment — what would a statistical model predict?
Reconcile: If your inside view and outside view differ, why?

CALIBRATION:
- Initial gut: [X]% confident
- Base rate for this type of judgment: [Y]%
- Outside view prediction: [Z]%
- Reconciled estimate: [final]% — because [reasoning for where you landed]

Step 4: Consider Reference Class

What is this an instance of? Find the right comparison set.

What’s the most similar past case you can identify?
What’s the typical outcome for cases like this?
How is this case different from the reference class? (better or worse?)

REFERENCE CLASS: [what category this belongs to]
TYPICAL OUTCOME: [what usually happens in this class]
THIS CASE IS DIFFERENT BECAUSE: [specific differences and their direction]
ADJUSTED EXPECTATION: [updated based on differences]

SKIP: If this is truly novel with no useful reference class, acknowledge that and proceed with extra caution.

Step 5: Identify What Would Change Your Judgment

Before committing, specify your update conditions.

What evidence would make you more confident?
What evidence would make you less confident?
What single fact, if true, would flip your judgment entirely?

WOULD INCREASE CONFIDENCE:
- [evidence 1]
- [evidence 2]

WOULD DECREASE CONFIDENCE:
- [evidence 1]
- [evidence 2]

WOULD FLIP JUDGMENT:
- [the key reversing condition]

This step prevents you from becoming unfalsifiable.

Step 6: Make the Call

State the judgment clearly. No hedging into meaninglessness.

JUDGMENT: [clear, specific statement]
CONFIDENCE: [X]% — [what this means: "I'd be wrong about [100-X] out of 100 similar calls"]
KEY FACTOR: [the single most important input to this judgment]
WEAKEST POINT: [where this judgment is most likely to be wrong]

Step 7: Decision Record

WHAT WAS JUDGED: [1-line summary]
JUDGMENT: [the call]
CONFIDENCE: [X]%
BASED ON: [top 2-3 inputs]
WOULD CHANGE IF: [key reversing condition]
REVIEW WHEN: [trigger or date to revisit this judgment]

Integration

Use with:

/prcp -> Improve perception before judging
/rskl -> Check the reasoning supporting your judgment
/mtcg -> Monitor for biases during the judgment process
/ht -> Test your judgment as a formal hypothesis
/dcp -> If the judgment feeds into a decision

jdgm - Judgment