Chapter 29 — Further Reading

Grouped by the book's three citation tiers (see _style-bible.md §7). Tier 1 = verified canonical sources we stand behind. Tier 2 = real ideas/literatures attributed honestly without a pinned-down exact citation. Tier 3 = illustrative/constructed material used for teaching. Annotations say what each is good for and, where relevant, its limits. The unifying use of all of these: practicing the §29.6 discipline of evaluating an emerging method before trusting it.

Tier 1 — Verified canonical

  • National Research Council (National Academy of Sciences), Strengthening Forensic Science in the United States: A Path Forward (2009). The yardstick this chapter applies prospectively. Read it not only as an audit of established methods but as the source of the question every emerging method must answer: has it been validated, and what is its error rate? Its central finding — that, apart from nuclear DNA, forensic methods had largely entered courtrooms without rigorous validation — is exactly the mistake §29.6 is designed to prevent the field from repeating with rapid DNA, microbial forensics, isotopes, and AI.

  • President's Council of Advisors on Science and Technology (PCAST), Forensic Science in Criminal Courts: Ensuring Scientific Validity of Feature-Comparison Methods (2016). Sharpens the NAS question into foundational validity — proof, by well-designed studies, that a method does what it claims, with a known error rate. The decisive lens for this chapter: it is the standard a microbial-PMI method has not yet met, the standard an AI tool's "99% accurate" marketing claim does not satisfy, and the standard rapid DNA meets only within its validated envelope.

  • The Rapid DNA Act of 2017 (United States, Public Law 115-50). The federal statute that created a pathway for approved rapid-DNA instruments to generate profiles from arrestee reference samples and, under defined conditions and approved processes, to search CODIS. Read it to understand the intended, validated scope of rapid DNA — clean reference samples in a booking context — and thus to recognize the scope creep (evidentiary samples in the field) the chapter warns against.

  • Federal Rules of Evidence, Rule 702; Daubert v. Merrell Dow Pharmaceuticals (1993); Kumho Tire Co. v. Carmichael (1999). The admissibility gate (Chapter 5). Directly relevant to every method in this chapter, but especially to the AI black box: a method whose reasoning cannot be examined is hard to square with a gatekeeping standard built around testability, error rate, and peer review. Kumho Tire extends the gatekeeping duty to all expert testimony, algorithmic included.

  • The public record of the Golden State Killer investigation and the arrest of Joseph James DeAngelo (2018). The anchor case (Case Study 1; advanced from Chapter 8's treatment). Valuable for the mechanics of investigative genetic genealogy — SNP profile, GEDmatch, family-tree reconstruction, STR confirmation on abandoned DNA — and for the privacy/ethics questions that crystallized afterward.

  • The public record of the Grim Sleeper investigation and the conviction of Lonnie David Franklin Jr. (Los Angeles). Case Study 2; widely cited as the first U.S. case in which a deliberate familial DNA database search led to an arrest in a major investigation. The textbook example of familial searching (as opposed to IGG) and of the equity questions rooted in who is already in the criminal database.

  • U.S. Department of Justice, interim policy on forensic genetic genealogical DNA analysis and searching (the federal "interim policy," effective 2019). The policy that set conditions on forensic genetic genealogy in federally supported investigations — generally limiting it to violent crimes and unidentified-remains cases and requiring conventional database searching first. Read it as the field's first formal attempt to govern a powerful technique whose law had lagged the science.

  • The Innocence Project (innocenceproject.org), case and policy record. Background for the chapter's governing argument. Note the structural point: the methods that have driven the most exonerations were those that overstated individualization without validation (bite marks, microscopic hair). The emerging methods most at risk of repeating that error are the ones offered as identifiers without measured error rates — which §29.6 is built to catch.

Tier 2 — Attributed, specifics unverified

  • The validation and performance literature on rapid-DNA instruments. A real body of work characterizes the performance of rapid-DNA systems on reference samples and the (much more limited) performance on challenging samples. We attribute the consensus — strong on clean single-source reference DNA, unsuited to complex evidentiary samples without laboratory interpretation — without citing a specific study; an agency adopting the technology should rely on its instrument's actual validation data.

  • The forensic microbiology / necrobiome literature on postmortem-interval estimation. A genuine and active research field proposes reading microbial community succession (on the body and in the underlying soil) as a PMI estimator, conceptually parallel to insect succession. We attribute the field's promise and its current limitations — sensitivity to temperature, soil, moisture, and individual variation, and the incomplete state of condition-spanning reference data and measured error rates — honestly, without overstating readiness.

  • The microbial-trace and skin-microbiome-transfer literature. Research documenting that individuals carry somewhat distinctive skin microbiomes that are shed onto touched objects has prompted proposals for microbial trace association (a complement to touch DNA). Attributed as an active research direction, not an established forensic method.

  • The stable-isotope literature on geographic provenancing of human tissues. A substantial scientific literature establishes that hydrogen/oxygen, carbon/nitrogen, sulfur, and strontium isotope ratios in tissues reflect water, diet, and geology, and vary geographically, enabling regional provenance and life-history reconstruction. We attribute the science's soundness and its forensic limits — coarse, reference-map-dependent, region-level inference — without a pinned citation; applied casework should rest on appropriate regional reference data.

  • The literature on demographic bias in facial-recognition and other forensic AI systems. Multiple evaluations have documented differential error rates across demographic groups in commercial facial-recognition systems, and a broader literature examines bias and validation gaps in algorithmic forensic tools. We attribute the existence and direction of these findings (worse performance on underrepresented groups; uneven and often undisclosed validation) in general terms, without claiming a specific figure.

  • The forensic and policy literature on familial searching and investigative genetic genealogy. Real scholarship addresses the kinship statistics of familial searching, the methodology of IGG, and the privacy, consent, equity, and Fourth Amendment questions both raise. Attributed as a developing literature and policy debate; specific jurisdictional rules vary and change.

Tier 3 — Illustrative / constructed

  • The Mill Creek cold case (the Case File, and the cold-case exercises). The gas-can DNA mixture, the rapid-DNA/IGG confirmation that the minor contributor is not an unknown stranger, and the resulting "stranger theory excluded" status are constructed teaching material, built on the frozen facts established in Chapters 7–8. Used to practice claiming an exclusion at its true strength and refusing to convert it into an identification. Clearly fictional; the persons of interest are invented.

  • Figure 29.1 ("What the tissues remember"), Figure 29.2 ("The confident algorithm"), and Figure 29.3 (the two-routes-to-a-relative schematic). Constructed teaching examples. The isotope life-history sketch, the facial-recognition score, and the schematics are illustrative; real casework uses actual reference data, validated tools, and measured error rates. The "96%" and "99%" figures in Figure 29.2 are illustrative round numbers chosen to make the score-versus-validation distinction transparent, not real performance figures.

  • The schematic family tree in Case Study 1. A generalized reconstruction of the IGG triangulation logic, labeled as such; the specific Golden State Killer tree is not reproduced, and the diagram teaches the method's structure, not the case's exact genealogy.

Where to go next in this book

  • For the validated DNA foundations these methods build on or hand off to, see Chapters 7 (STR, CODIS, RMP), 8 (touch DNA, mixtures, IGG), and 9 (likelihood ratio, probabilistic genotyping, the prosecutor's fallacy).
  • For the entomological clock the necrobiome is a cousin of, see Chapter 13; for the pathology PMI it would extend, Chapter 11.
  • For the admissibility gate every emerging method must pass, see Chapter 5 (Daubert); for the field's reckoning that built the validity yardstick, Chapter 6 (NAS/PCAST).
  • For how an expert presents an AI lead, an isotope region, or a genealogy hand-off without overstating, see Chapter 30; for the cognitive-bias safeguards (including automation bias), Chapter 31.
  • For the privacy, independence, and reform debates these technologies sharpen, see Chapter 38; and for the capstone assembly of the cold case — where the "stranger theory excluded" finding takes its place among every other thread — Chapter 39.