Chapter 29 — Exercises

Work these without looking back at the chapter first; then check yourself. Items marked † have full worked solutions in the answers appendix. There are no answers in this file. Mix of recall, applied reasoning, evidence interpretation, "spot the overstatement," ethics, and a cold-case extension. The through-line of the chapter — hold every emerging method to the same validity yardstick — should guide almost every answer.

A. Recall and definitions

  1. Define rapid DNA in one sentence, and state the single kind of sample it is validated for.

  2. † Define microbial forensics and the necrobiome, and name the established forensic clock (Chapter 13) that the necrobiome is conceptually a cousin of.

  3. Define forensic isotope analysis, and state the kind of question it answers (and the kind it cannot).

  4. Define AI/ML in forensics, and name two tools from earlier chapters that already rely on it.

  5. Define familial searching, and state the theory of inheritance it rests on (why a close relative produces a partial, not exact, match).

  6. In one sentence each, distinguish a stable isotope from a radioactive one, and explain why tooth enamel and hair record different time periods of a person's life.

  7. What is the Rapid DNA Act of 2017, and what use of rapid DNA did it create a pathway for?

  8. Name the three perils of AI/ML in forensics emphasized in §29.4, and give a one-phrase description of each.

B. Applied reasoning

  1. † A rapid-DNA instrument produces a clean profile from an arrestee's buccal swab in 95 minutes. The same instrument is then used on a low-template, heat-degraded swab from a crime scene. Explain, using the idea of the "validated envelope," why the first use is legitimate and the second is not — and name the specific safeguard that automation removed.

  2. Why does removing the human analyst (the feature that makes rapid DNA fast) also remove a safeguard? Name one specific problem an analyst would catch that the box will not.

  3. A microbial-PMI method claims to estimate time since death from the bacteria on a corpse. List three environmental or individual variables that could distort the estimate, and explain why "under given conditions" is the phrase that bounds the method.

  4. † An unidentified body yields isotope data: tooth enamel consistent with one region, bone with another, hair with a dietary shift in the final months. Write the single most defensible sentence an expert could say about this person's life history, and the single sentence that would overstate it.

  5. Explain why a facial-recognition system's "96% similarity score" is not the probability that the match is correct. What is it?

  6. A vendor advertises an AI tool as "99% accurate." List four questions you must answer before that figure means anything for court.

  7. Distinguish familial searching from investigative genetic genealogy on three axes: genetic markers, database searched, and degree of relatedness reached.

  8. † Walk through how the Golden State Killer was identified, in five steps, making explicit which step generated the lead and which step made the identification. Why does that division of labor make IGG a model of methodological honesty?

  9. Automation bias and bias laundering are two ways AI relocates rather than removes cognitive bias. Define each, and explain why each is harder to challenge than the human bias it replaces.

  10. Why is an algorithm that "cannot show its work" a problem for forensic testimony specifically (as opposed to, say, a weather forecast)? Connect your answer to cross-examination (Chapter 30) and the Daubert gatekeeper (Chapter 5).

C. Evidence interpretation

  1. † Re-read Figure 29.2 ("The confident algorithm"). The system reports a 96% score and the vendor claims 99% accuracy. State precisely (a) what the output does show, (b) what it does not establish, and (c) the strongest honest use of the result.

  2. Figure 29.1 ("What the tissues remember") reconstructs a layered geographic life history. List two things this evidence genuinely contributes to identifying the unidentified dead, and two things it specifically cannot do.

  3. An agency runs evidentiary (crime-scene) samples through a rapid-DNA box in the field, with no laboratory and no analyst, and reports the profiles as identifications. Identify the methodological error using §29.1, and name the earlier-chapter problem (touch DNA / mixtures) being ignored.

  4. A report estimates a postmortem interval from a microbial sample and presents it with the same confidence as a validated DNA result. Using §29.2 and §29.6, list the specific things the microbial estimate still lacks that the DNA result has.

D. Spot the overstatement / junk-science alert

  1. † A slide reads: "AI facial recognition has positively identified the defendant from surveillance footage." Identify two distinct problems with this statement and rewrite it honestly.

  2. An expert testifies: "The isotopes prove she grew up in this town." Name the specific overstatement (what is being claimed that isotope evidence cannot support) and give the defensible version.

  3. "It uses next-generation sequencing and deep learning, so it must be reliable." Name the rhetorical move (from the §29.6 Junk-Science Alert) and explain why sophistication is not validity.

  4. A company says its method "cracked a cold case that had stumped investigators for decades," offering that as proof the method is reliable in general. Explain why a single successful demonstration (with the answer ultimately known) is not a validation, and what would be.

  5. A detective says, "The computer flagged him, so it's basically confirmed." Using automation bias (§29.4), explain what has gone wrong and what the output actually was.

E. Ethics and reasoning

  1. † IGG and familial searching both extend DNA's reach from the suspect to the suspect's relatives. Explain the distinct privacy concern each raises — "consent at a distance" (IGG) versus the genetic surveillance of the already-databased population's families (familial searching) — and why each falls unequally across populations.

  2. A genealogy database changes its terms of service to permit law-enforcement matching after users have uploaded. Explain why this sharpens the "informed consent" problem, and whose interests are affected beyond the person who uploaded.

  3. You are asked to testify that an AI tool's output "identifies" a suspect, when the tool is an unvalidated black box. Explain why you should decline, what you can honestly say instead, and how this parallels the bite-mark error (Chapter 16, previewed).

  4. The chapter says a triumph (like the Golden State Killer case) is "the moment to examine costs, not to suspend judgment." Explain why a celebrated success is exactly when scrutiny is most needed — and most likely to be silenced.

F. Synthesis and validity spectrum

  1. † Place these on the NAS 2009 / PCAST 2016 validity spectrum (strong → emerging/unvalidated), justifying each: (a) rapid DNA on a clean reference swab; (b) the conventional STR confirmation that follows an IGG lead; (c) forensic isotope analysis as an investigative geographic lead; (d) a microbial-PMI estimate today; (e) an opaque, vendor-validated facial-recognition match offered as proof of identity.

  2. Explain how rapid DNA can be both near the top of the validity spectrum and effectively off it — what single variable decides which, and why this proves that "validity is a property of method-plus-sample-plus-question, not of the instrument alone."

  3. In one paragraph, explain how this chapter advances at least two of the book's four themes (exclusion over proof; the validity spectrum; cognitive bias; the CSI effect cutting both ways). Name which themes and how — and note specifically how the validity spectrum is used prospectively here rather than as a backward audit.

G. Cold-case extension

  1. † Cold Case. This chapter's Case File delivers an exclusion, not an identification: rapid DNA and IGG establish that the minor contributor to the gas-can DNA mixture is not an unknown stranger, excluding the "random intruder" theory. Write the entry you would add to the Mill Creek evidence log (Appendix I). State (a) the defensible inference at its true strength, (b) the honest verb, (c) at least three things this evidence specifically does not establish, and (d) why the analyst stops at an exclusion rather than naming the contributor.

  2. Cold Case extension. Explain how the chapter's exclusion ("stranger theory excluded") interacts with the earlier finding from Chapter 8 (the gas-can DNA is a mixture: victim + an unknown minor contributor). What did Chapter 8 leave open that this chapter narrows, and what does it still leave for the capstone (Chapter 39)?

  3. Cold Case, integrative. Suppose a detective says, "So the genealogy names the killer." Using the chapter's distinction between an investigative lead and a courtroom identification, explain why that statement is wrong even though the genealogy was decisive in narrowing the field — and what would have to happen for a name to be honestly spoken.

H. Short writing

  1. In 150–200 words, explain to a juror why "rapid DNA" and "AI identified him" are phrases to interrogate rather than trust — and what single question they should ask of each.

  2. † In 150–200 words, contrast the validity foundations of forensic isotope analysis with those of an unvalidated AI black box: what does each rest on, what does each honestly claim, and where is each most easily overstated in court?

  3. In 150–200 words, defend the §29.6 evaluation checklist as the "most durable thing this book can teach." Why is a forward-looking gate more valuable than any list of which current methods are good or bad?