Download presentation
Presentation is loading. Please wait.
1
Fingerprints and Identity: Exploring the Validity and Certainty of Identification
Karrie L. Casada University of California, Irvine Undergraduate Research Symposium May 15, 2004 1/18/2019
2
Key Definitions Automated Fingerprint Identification System AFIS
Ten-Print or Rolled Impressions Latent Prints 1/18/2019
3
Key Definitions (cont.)
Candidate list Hit Match Minutia 1/18/2019
4
Look to the past…. Fingerprint evidence has been used for over 100 years Accepted without challenge Not tested Become part of our belief of reliable forensic evidence 1/18/2019
5
…and the evidence is flawed
How do we know it is reliable? How many people have been put in jail because of fingerprint evidence? What if we are wrong? 1/18/2019
6
Past research: Uniqueness
Most research supports uniqueness Uniqueness if often used to explain accuracy and validity Since fingerprints are created in the womb, they are permanent and unique (Babler, 1987) Because they are permanent and unique, the error rate is zero (Wertheim, 2001) 1/18/2019
7
Past research: Uniqueness (cont.)
Uniqueness cannot account for accuracy or validity The question is not about uniqueness but about human error: how much similarity exists between prints that might hinder distinguishing one print from another? How can an subjective observation lead to an absolute identification? 1/18/2019
8
Past research: Accuracy & Validity
Since 1892, ten statistical models have been presented in an attempt to provide a probabilistic model for fingerprint individuality (Stoney, 2001) Each model has flaws that damage the approach irrevocably 1/18/2019
9
Past research: Accuracy & Validity (Cont.)
“From a statistical viewpoint, the scientific foundation for fingerprint individuality is incredibly weak” “None of the models (have) been subjected to testing, which is of course the basic element of the scientific approach” (Stoney, 2001) 1/18/2019
10
Past research: Proficiency Testing
The International Association for Identification (IAI) certifies individual fingerprint examiners: pass rate of 50% The American Society of Crime Laboratory Directors (ASCLD): 20% of the examiners made an erroneous identification (Haber and Haber, 2003) 1/18/2019
11
Past research: Proficiency Testing (cont.)
Examiners are not required to be certified to be deemed an expert in a courtroom Test does not mimic real life identifications Use of clean, clear test prints Lack of blind testing Lack of generalalizablity Lack of AFIS (Haber and Haber, 2003) 1/18/2019
12
Past research: Proficiency Testing (cont.)
This “matching” game does not produce a zero error rate for examiner accuracy “If these comparisons had been presented in court as testimony, one in five latent print examiners would have provided damning evidence against the wrong person” (Haber and Haber, 2003) 1/18/2019
13
Past research: AFIS Meagher (2002), testifying on behalf of the FBI, estimated that 50% of fingerprint comparisons use AFIS “There are no published experimental tests on the accuracy with which human examiners compare latent prints to AFIS outputs, nor any on the effect of the number of candidates produced by AFIS on the accuracy of comparison” (Haber and Haber, 2003) 1/18/2019
14
Past research: AFIS (cont.)
FBI 50K study Copy of database print used as search print Searching for an exact duplicate Latent search print was a cropped section of database print Still searching for an exact duplicate (Kaye, 2002) 1/18/2019
15
Past research: AFIS (cont.)
“This number, 1097, is extremely large. We have no word for this number in any language, as it is beyond human comprehension. In the entire history of mankind, there have been only about 1011 fingerprints. It is possible that in the entire future of all mankind there will never be 1097 fingerprints” (Wayman, 2000) Because neither the method nor the calculations were made available, scholars can only guess as to how the final results were reached (Stoney, 2001) 1/18/2019
16
Past research: AFIS (cont.)
Test produced an unusual high Z score for 3 comparisons 3 prints were duplicates 3 prints were discarded 1/18/2019
17
Past research: AFIS (cont.)
The fingerprint community and scholars are aware that no two impressions obtained from rolled or latent prints from the same finger are exactly alike In real world fingerprint comparisons, a print would never be compared to itself Flawed statistical methods Test and its results were never published 1/18/2019
18
Past research: AFIS (cont.)
“…experimental tests are needed to demonstrate the accuracy with which examiners make identifications and elimination using AFIS outputs. Until such data is available, identifications made in court based on AFIS search outputs will continue to pose special concerns about error rates. At present, there are no data” (Haber and Haber, 2003) “…the real issue is whether fragments of prints, or smeared prints, or layers of prints one upon another, can be accurately linked to one and only one person on earth. It is time to examine the court’s antipodal opinions on this issue” (Kaye, 2003) 1/18/2019
19
Current Study Overview
Collect ten-prints Collect latent prints All prints are numerically linked Create an AFIS database Run searches 1/18/2019
20
Hypotheses H1: Can a candidate be located for each search print?
H2: Is the first candidate the match? H3: Can a statistical probability be derived? 1/18/2019
21
Methodology Ten-print cards collected: 166 Latent prints collected:
371 1/18/2019
22
Methodology (cont.) Ten-prints scanned into AFIS creating a searchable database Ten-prints prints are “auto extracted” Latent prints used as search prints Candidate list analyzed 1/18/2019
23
Results H1: Although each latent print has a
corresponding ten-print card in the system, 31% had “no hit” 1/18/2019
24
Results (cont.) H2: Although each latent print has a
corresponding ten-print card in the system, 60% had the correct candidate “hit” in the #1 position 1/18/2019
25
Results (cont.) H3: Based on the points used to locate the correct match print, perhaps a statistical probability can be developed 1/18/2019
26
Discussion Latent prints that mimic real-life crime scene prints are key to any fingerprint research 1/18/2019
27
Discussion (cont.) Error rate is not zero
If examiners rely on the first “match,” they might be excluding the correct candidate AFIS is a screening tool 1/18/2019
28
Discussion (cont.) Accuracy dependent on quality prints
Certainty is dependent on examiner abilities Uniqueness theory can hinder proper identifications 1/18/2019
29
Perspective Small database Peer review to explore 1/18/2019
30
Future research Amount of detail needed for a match
The role of minutia points Examiners ability to interpret a candidate list Examiner bench note availability as courtroom evidence 1/18/2019
31
Acknowledgements Dr. Simon Cole Dr. Valerie Jenness UROP AFIX Tracker
Social Science Lab Volunteer subjects 1/18/2019
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.