JevMade hello@JevMade.com
← Back to guides

JevMade field notes / Primitive usage guide

Rate one quality on a clearly described scale

Score places an item along ordered, described levels. Its number is the average position formed from the probability Jev assigns to every level, not an exact measurement.

Original by TypeSafe AIEvaluation

Listen to this guide

JevMade’s plain-English explanation

Expanded guide unavailable: permission has not been verified.

0:00 /

AI narration

Our summary

Score helps when possible answers have an order, such as calm, frustrated, and very angry. Jev assigns a probability to every described level, then returns an average position based on those probabilities. The number summarizes the spread rather than measuring the subject directly.

A program can compare scores or combine several questions. Before combining scales of different lengths, it should put them onto the same comparable range. Clear descriptions and examples can sharpen boundaries, but examples should resemble real inputs and be tested on separate material.

A result between two levels does not create a newly measured fact. It only reflects probability spread across the levels supplied. Do not treat that between-level result as an exact measurement, and do not assume a tightly grouped answer is correct without checking known examples.

Key takeaways

  1. Describe each ordered level in words before using its numbered position.
  2. Read the score as an average of the probability assigned to each level.
  3. Put different scales onto a comparable range before combining them, and do not treat a between-level result as an exact measurement.

The official examples illustrate recorded behavior. Higher confidence alone does not show that a level description is correct.

Official TypeSafe documentation · Source reviewed

Read the original guide Opens the author’s site in a new tab.