What does Three-Light Cut Points conclude, in numbers, about “Setting the Doubtful Aside, and Bayes: Two Classical Texts on Being Unsure”?
Answer: Chapter 7 examines the Analects' "hear much, set aside the doubtful, speak with care about the rest, and you will seldom err" and Bayes's paper of 1763: with 19 right out of 20 reviewed, the posterior mean is only 0.9091.
Scope of the numbers: Private customer service has a third action, a bot draft checked by a person, so it needs a six-cell loss matrix with two cut points; under illustrative parameters red lights below a chance of being right of 0.4000 and green from 0.9500. Parameters are illustrative, all conversations are fictional, and all numbers come from the book's programs, which read and write no file and connect to no customer-service system.
Data table
The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.9091, 19, 20
| n | k | raw rate | post mean | P(>0.95) | P(>0.84) | light |
| 20 | 19 | 0.950 | 0.9091 | 0.283 | 0.872 | amber |
| 40 | 40 | 1.000 | 0.9762 | 0.878 | 0.999 | green |
| 60 | 58 | 0.967 | 0.9516 | 0.594 | 0.998 | green |
| 200 | 192 | 0.960 | 0.9554 | 0.679 | 1.000 | green |
| 1,000 | 955 | 0.955 | 0.9541 | 0.741 | 1.000 | green |
| 5 | 2 | 0.400 | 0.4286 | 0.000 | 0.007 | amber |
| 12 | 4 | 0.333 | 0.3571 | 0.000 | 0.000 | red |
Show the full program
# Chapter 7. p72: Bayes's 1763 question asked of one score bin.
# Model: in one bin of the bot's score, k of n reviewed chats were
# answered right. With Bayes's uniform prior the chance theta that the
# bot is right in that bin has posterior Beta(k + 1, n - k + 1), and
# P(theta > x) = P(Binomial(n + 1, x) <= k) (exact identity).
# The losses of p21 are linear in theta, so the Bayes action uses the
# posterior mean (k + 1) / (n + 2): green from 0.95, red below 0.40.
# Bins are illustrative.
import math
def tail(k, n, x):
m = n + 1
terms = [math.comb(m, j) * x ** j * (1 - x) ** (m - j)
for j in range(k + 1)]
return math.fsum(terms)
def light(q):
return "green" if q >= 0.95 else ("amber" if q >= 0.40 else "red")
print(" n k raw rate post mean P(>0.95) P(>0.84) light")
for n, k in ((20, 19), (40, 40), (60, 58), (200, 192), (1000, 955),
(5, 2), (12, 4)):
mean = (k + 1) / (n + 2)
print("%5s %4d %9.3f %9.4f %8.3f %8.3f %s"
% (format(n, ","), k, k / n, mean, tail(k, n, 0.95),
tail(k, n, 0.84),
light(mean)))Single source and citation
Fan, Y. (2026). Three Lights for Handing Over between Bot and Human in Private Customer Service: Misclassification Loss Matrices, Chow's Reject Rule and Optimal Cut Points on Calibrated Confidence (Version 1.0.0), Chapter 7: Setting the Doubtful Aside, and Bayes: Two Classical Texts on Being Unsure. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-266 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026266,
author = {Fan, Yafu},
title = {Three Lights for Handing Over between Bot and Human in Private Customer Service: Misclassification Loss Matrices, Chow's Reject Rule and Optimal Cut Points on Calibrated Confidence},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · The Blind Spot of Accuracy: Chow's Reject Rule and the Cost Ratio
- Chapter 2 · A Six-Cell Loss Matrix: Bayes Risk in Three Zones and Two Cut Points
- Chapter 3 · Cost Curves: Why the More Accurate Bot Can Cost More
- Chapter 4 · Confidence Is Not Probability: Isotonic Calibration and the Price of Few Labels
- Chapter 5 · Big Sales and Drift: Bayesian Correction for Prior Shift and the Sampling Error of Cut Points
- Chapter 6 · Red Lights Beyond Price: Stop Means Stop, Vulnerable Customers Go to Staff, and the Staffing Cap
- Chapter 8 · The Monthly Recalculation Loop, the Distilled Tool and a Falsification Plan