What does Costly Signals and Hot-Cold Traps conclude, in numbers, about “Variable-Ratio Reinforcement and Resistance to Extinction: Why Responses That Come and Go Are the Hardest to Give Up”?

Answer: Chapter 3 uses the partial reinforcement extinction effect and a discrimination model to show that a variable schedule answering one attempt in four needs 21 times as many misses before quitting as steady responding, and that the mean size of the prediction error peaks at 0.5000 when responses come half the time. The small temporal-difference model of Chapter 4 shows that ranking partners by intensity of feeling almost reverses ranking by mean warmth, and that a partner who floods and then withdraws can produce a single rush of 0.887.

Scope of the numbers: Its parameters are illustrative; all calculations are generated by its programs; it is not psychological counselling, claims no effect of any kind, and teaches no technique of manipulation. Chapter 2 enters each interaction in a likelihood-ratio ledger; with illustrative parameters one missed plan weighs as much as 2.52 kept ones, and Wald's sequential test saves 0.459 of the interactions a fixed observation period needs.

中文版

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: 21

schedulepays/attemptmisses to quitmisses to quit
(pi = 0.05)(pi = 0.20)
continuous1.0011
variable, 1 in 20.5097
variable, 1 in 40.252116
variable, 1 in 100.105642
fixed, every 40.2544
fixed, every 100.101010
1-in-4 needs 21 times the misses of steady reward
Show the full program
# Chapter 3. p31: why partial reinforcement resists extinction, in the
# "discrimination" reading. A learner trained on a schedule that pays
# with probability p per attempt sees n unanswered attempts in a row.
# It concludes "the source has gone cold" once the posterior of a change
# reaches 0.95, with prior pi of a change. Posterior after n misses:
# pi / (pi + (1 - pi)(1 - p)^n). Fixed ratio k (pay every k-th attempt)
# is told apart after k misses. Exact; the schedules are illustrative.
import math


def misses_needed(p, pi, conf=0.95):
    if p >= 1.0:
        return 1
    bound = pi * (1 - conf) / ((1 - pi) * conf)
    return math.ceil(math.log(bound) / math.log(1 - p))


print("schedule            pays/attempt  misses to quit  misses to quit")
print("                                     (pi = 0.05)     (pi = 0.20)")
for lab, p in (("continuous", 1.0), ("variable, 1 in 2", 0.5),
               ("variable, 1 in 4", 0.25), ("variable, 1 in 10", 0.1)):
    print("{:18s}  {:12.2f}  {:14d}  {:14d}".format(
        lab, p, misses_needed(p, 0.05), misses_needed(p, 0.20)))
for k in (4, 10):
    print("{:18s}  {:12.2f}  {:14d}  {:14d}".format(
        "fixed, every " + str(k), 1.0 / k, k, k))
r = misses_needed(0.25, 0.05) / misses_needed(1.0, 0.05)
print("1-in-4 needs {:.0f} times the misses of steady reward".format(r))

Single source and citation

Fan, Y. (2026). Costly-Signal Screening in Early Intimacy, Intermittent-Reinforcement Dopamine Traps and Reciprocal-Floor Stopping Games (Version 1.0.0), Chapter 3: Variable-Ratio Reinforcement and Resistance to Extinction: Why Responses That Come and Go Are the Hardest to Give Up. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-181 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026181,
  author    = {Fan, Yafu},
  title     = {Costly-Signal Screening in Early Intimacy, Intermittent-Reinforcement Dopamine Traps and Reciprocal-Floor Stopping Games},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book