New NIJ Stuff

[L.J.Steele]

Through the National Criminal Justice Reference Service, NIJ has made available the following final technical reports:

Cognitive and Contextual Influences in Determination of Latent Fingerprint Suitability for Identification Judgments (pdf, 36 pages)
https://www.ncjrs.gov/pdffiles1/nij/grants/241289.pdf

Quantitative Measures in Support of Latent Print Comparison (pdf, 63 pages)
See below for additional details.
https://www.ncjrs.gov/pdffiles1/nij/grants/241288.pdf

These reports are the results of NIJ-funded projects but were not published by the U.S. Department of Justice.

Title: Cognitive and Contextual Influences in Determination of Latent Fingerprint Suitability for Identification Judgments (pdf, 36 pages)
Authors: Peter Fraser-Mackenzie, Cognitive Consultants International & University of Southampton; Itiel Dror, University College London; Kasey Wertheim, Institute of Cognitive Neuroscience
Authors' Abstract:

We examined forensic fingerprint examiners' suitability determinations of latent fingerprints comparing situations in which the latent is assessed solo (in isolation) versus situations in which it is presented alongside a comparison (matching or non-matching) exemplar print. The presence of a non-matching comparison exemplar led examiners to be more inclined to draw the conclusion that the latent was suitable for comparison compared to when the latent was presented solo. This effect persisted even when the latent presented was highly unsuitable for comparison. The presence of a matching comparison exemplar led examiners to be less likely to decide that the latent was suitable and more likely to decide the latent was questionable compared to solo analysis. This effect persisted even when the latent presented was highly suitable, suggesting a strong main effect.

Knowledge of another examiner's previous determination that the latent was unsuitable was found to increase the likelihood that the examiner would conclude that the latent was unsuitable. However, knowledge of a previous "suitable" determination by another examiner did not increase the likelihood of a "suitable" conclusion by examiners. The finding that effects were weaker, although not entirely removed, in those with IAI certification suggests that training may be an appropriate route for reducing the effect of contextual influence and bias in suitability determinations. It was also shown that a latent prints that were previous classed as "unsuitable" in a non-biasing context tended to still be judged to be "unsuitable" by examiners that were presented with the latent in a strongly biasing context (a major case in which a previous examiner was purported to have made an Individualization).

Title: Quantitative Measures in Support of Latent Print Comparison (pdf, 63 pages)
Author: Sargur N. Srihari, University at Buffalo
Author's Abstract:

Latent prints of friction ridge impressions have long been useful in identification, and the methodology of examining latent prints, known as ACE-V (analysis, comparison, evaluation and verification), has been well documented. The need to quantify confidences within ACE-V has been articulated in several recent influential reports to strengthen the science of friction ridge analysis. This research addresses the evaluation of three quantitative measures: rarity of features, confidence of opinion and a probabilistic measure of similarity. The first of these, useful in the analysis phase of ACE-V, is to determine the rarity of observed features. Rarity is difficult to compute due to the large number of variables and high data requirements. The proposed solution uses probabilistic graphical models to represent spatial distributions of fingerprints represented at level 2 detail (minutiae). First, the minutia coordinate system is transformed into standard position based on a point of high curvature, viz., core point; statistical regression (based on a Gaussian process formulation and a training set of latent prints) is used to estimate the core point.

A directed probabilistic graphical model is constructed using inter-minutia dependencies and minutia confidences. The resulting model is used to determine the probability of random correspondence of the evidence in a database of n prints. The method is validated using statistical goodness-of-t tests and illustrated using: (i) a simple configuration of minutiae, (ii) randomly selected latent fingerprints in a database, and (iii) a well-known case of erroneous identification. The second quantitative measure addressed is that of determining confidence of opinion - which is relevant to the evaluation phase of ACE-V. The proposed computation determines a likelihood ratio as a product of rarity and the probability of similarity under the identification hypothesis. The third area concerns measuring similarity probabilistically, in a manner analogous to cognition. A Markov random field is used to jointly model the minutiae in both the evidence and the known. The algorithms developed are based on statistical machine learning using several publicly available fingerprint data sets for parameter learning and testing. The developed methods can be of potential use in examiner training, presentation of opinion and validating examination procedures.