Book Title: Computational Criminology: AI Applications in Forensic Science and Criminal Justice
Editors: Dr. Xavier Louis, Dr. Surbhi Girdhar, Ms. Aswathi Chandran Nair, Mr. Ravi Kumar, and Ms. Nandini Katare
Chapter: 11
DOI: https://doi.org/10.59646/704/11
Author: Labhini Lalit Rahangdale
Abstract
For fourdecades the dominant forensic use of DNA has been identification: comparing short tandem repeat (STR) profiles recovered from a crime scene with profiles in a reference database, or from a known suspect, to produce a probabilistic match statistic.[1] This use of DNA identity matching is by now methodologically mature, widely admitted in courts, and underwritten by decades of validation studies on inheritance, allele frequencies, and population genetics. It is the paradigm of forensic science success in the late twentieth and early twenty-first centuries. Forensic DNA phenotyping (FDP) is a different enterprise. Rather than asking whose DNA this is, it asks what does the contributor look like: the externally visible characteristics eye colour, hair colour, skin pigmentation, facial morphology, biogeographical ancestry, chronological age that can be inferred from the DNA recovered from a crime scene where no suspect or reference profile is yet available.[2] Where identification answers a yes/no question with high statistical precision, phenotyping produces a probabilistic prediction across multiple correlated traits, each with its own uncertainty bounds, calibration properties, and historical baggage of misuse. Linear regression on a handful of marker SNPs gave way to ensemble methods such as random forests and gradient boosting, which in turn are being supplemented by deep neural networks trained on whole-genome data and on three-dimensional craniofacial scans.[3] This chapter examines the methodological state of forensic DNA phenotyping, the predictive performance achievable on the traits most studied, the limits of the inference, and the substantial ethical and constitutional concerns that the technology raises.[4] We argue that FDP, even at its current technical best, is suited to investigative narrowing rather than to evidentiary identification, and that responsible deployment depends on institutional safeguards that no algorithmic improvement can substitute for.[5]
[1]Manfred Kayser and Peter de Knijff, “Improving Human Forensics through Advances in Genetics, Genomics and Molecular Biology” (2011) 12 Nature Reviews Genetics 179, 181.
[2]Manfred Kayser, “Forensic DNA Phenotyping: Predicting Human Appearance from Crime Scene Material for Investigative Purposes” (2015) 18 Forensic Science International: Genetics 33, 35.
[3]Susan Walsh and others, “IrisPlex: A Sensitive DNA Tool for Accurate Prediction of Blue and Brown Eye Colour in the Absence of Ancestry Information” (2011) 5 Forensic Science International: Genetics 170, 173.
[4]Susan Walsh and others, “The HIrisPlex System for Simultaneous Prediction of Hair and Eye Colour from DNA” (2013) 7 Forensic Science International: Genetics 98, 101.
[5]Sinha, A. K. (2026). Whistleblower Protection and Criminal Investigations in Corporations. Minnesota Journal of Business Law and Entrepreneurship (1) 452, 453.