What does A Zero-Leak Socratic Examiner conclude, in numbers, about “Bayesian Knowledge Tracing: Four Parameters, the Mastery Line and Identifiability”?

Answer: Chapters 3 and 4 cover the four-parameter update of knowledge tracing, the mastery line and identifiability, and show that when a fifth of learners hold the answers, false certificates make up 0.2534 of all certificates, while a fit after the fact reads this as a stronger intake.

Scope of the numbers: Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface, contain no keys or server addresses and include no concrete extraction prompts. This is not individual business advice.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it.

step respP(L|obs)P(L next)P(correct next)mastered
0--0.30000.4100
100.05080.14580.3020no
210.43430.49090.5436no
310.81270.83140.7820no
400.38140.44330.5103no
510.78180.80360.7625no
610.94850.95360.8675yes
710.98930.99040.8933yes
810.99780.99810.8986yes
910.99960.99960.8997yes
mastery 0.95 first reached at step 6
Show the full program
# Chapter 3. p31: Bayesian knowledge tracing, one student, one skill.
# The four parameters of Corbett and Anderson's model (values are
# illustrative, of the size often reported):
#   L0  probability the skill is known before the first item  0.30
#   T   probability of learning at each opportunity           0.10
#   G   probability of a correct answer without the skill     0.20
#   S   probability of a wrong answer despite the skill       0.10
# After each graded response:
#   correct: L|obs = L(1-S) / (L(1-S) + (1-L)G)
#   wrong:   L|obs = LS / (LS + (1-L)(1-G))
#   then     L' = L|obs + (1 - L|obs) T
# Only the verdict bit enters the update; the answer itself never does.
L0, T, G, S = 0.30, 0.10, 0.20, 0.10
MASTERY = 0.95


def update(p, correct):
    if correct:
        post = p * (1 - S) / (p * (1 - S) + (1 - p) * G)
    else:
        post = p * S / (p * S + (1 - p) * (1 - G))
    return post, post + (1 - post) * T


def p_correct(p):
    return p * (1 - S) + (1 - p) * G


seq = (0, 1, 1, 0, 1, 1, 1, 1, 1)
p = L0
print("step resp  P(L|obs)  P(L next)  P(correct next)  mastered")
print("0    -     -         %.4f     %.4f" % (p, p_correct(p)))
done = None
for i, r in enumerate(seq, 1):
    post, p = update(p, r)
    flag = "yes" if p >= MASTERY else "no"
    if p >= MASTERY and done is None:
        done = i
    print("%-4d %-5d %-9.4f %-10.4f %-16.4f %s"
          % (i, r, post, p, p_correct(p), flag))
print("mastery %.2f first reached at step %s" % (MASTERY, done))

Single source and citation

Fan, Y. (2026). A Zero-Leak Socratic Examiner: Server-Side Isolation of Answers and Bayesian Mastery Probing Trees (Version 1.0.0), Chapter 3: Bayesian Knowledge Tracing: Four Parameters, the Mastery Line and Identifiability. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-262 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026262,
  author    = {Fan, Yafu},
  title     = {A Zero-Leak Socratic Examiner: Server-Side Isolation of Answers and Bayesian Mastery Probing Trees},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book