Skip to content

Chapter 1 templates · The "Is This Book for You" Self-Test

How to use. Answer the seven questions below honestly, one by one, and score each by the rule it gives. Total time about one minute. This checklist is the full version of the four questions in section 1.5 of Chapter 1. It does not test your level. It tests your situation. The value of this book depends on whether a wrong answer from you really hurts something. The first six questions are scored. The seventh is not scored, and it decides how you read Part II.

Self-test questions

1. The last "conclusive output" you handed over (a report / analysis / judgment / review). If its core conclusion were wrong, what would happen? - Someone would make a costly decision based on it (spend money, set an architecture, launch a project, seek treatment, invest) → +2 - I would lose face, but no real decision depends on it → +1 - Nothing would happen → 0

2. Are you already using AI (or being asked to use AI) to produce this kind of content? - Yes, and the frequency is rising → +2 - Not yet, but colleagues / competitors / upstream suppliers are → +1 - Not at all, and not in the near term → 0

3. Have you received material with deep AI involvement (reports, reviews, due diligence) where the sender did not say so? - Definitely yes → +2 - Cannot rule it out → +1 - Definitely not (and I can say why I am sure) → 0 - Hint: if you want to answer "probably not," that is "cannot rule it out," not "definitely not."

4. Can you accept the premise that "verification takes work"? - Yes, I want the work made affordable, not made zero → +2 - I hope one day it will be reliable in one click → 0 (this book will disappoint you)

5. Are you willing to accept honest but unsatisfying answers like "still exploring / nobody knows"? - Yes, I would rather have an honest map than a categorical one → +2 - I need a definite conclusion for every question → 0

6. Do you have a real, unsolved problem in hand right now that can carry the exercises through the book? - Yes, and I can write it as one sentence right now → +2 - No, but I want to find one → +1 - No, and I do not plan to → 0

7. Is your evidence something that can be rerun? (not scored) - Yes, code, data, a ledger that can be recomputed, and a bad run can be redone → you are the core reader this book writes the full process for, follow Part II as written - Partly, for example I have data but the collection cannot be redone → read Chapter 3, section 3.6, check the five premises one by one, and convert wherever one breaks - No, wet lab, one-shot interviews, clinical data → the steps still apply, but the shape of the case has to be converted, Chapter 3, section 3.6 is your exchange-rate table, read it before entering Part II

Scoring

  • 8–12 points. This book's primary reader is you. Start with Start Here and produce your first artifact the same day.
  • 4–7 points. This book is useful to you, but read Part I (the frame) and Part III (trust) first, and enter Part II when a real problem shows up in your hands.
  • 0–3 points. What you need right now is probably a good search or Q&A tool, not this book. That is not a criticism. Paying verification cost for answers that hurt nothing is waste.

Self-check (Chapter 1 failure modes)

  • [ ] I have not taken "AI can make scientific discoveries" (AlphaFold-style headlines) as a reason to believe "the AI report on my desk is reliable." No credibility passes between the two.
  • [ ] I know a report's persuasiveness and its correctness are decoupled. To judge whether it is reliable, I look at the process that produced it, not at how smoothly the text reads.
  • [ ] Before forwarding any AI output, I have picked at least one number that would change the conclusion and asked, "where did this number come from?"
  • [ ] I have not abandoned the whole thing over one fabricated citation, and I have not signed and forwarded because it was "mostly right." Chapter 1 names both reactions.