http://for-sci-law.blogspot.com/2017/08 ... other.html
an excerpt below:
The Scientific Method
From the outset, the authors -- all leading researchers in forensic science -- express dissatisfaction with "standard, shallow statements such as 'nature never repeats itself'" and "the tautological argument that every entity in nature is unique." (P. 1). They also dispute the claim, popular among latent print examiners, that the "ACE-V protocol" is a deeply "scientific method":
Indeed, it is hard to know what to make of claims that "standard hypothesis testing" is the "scientific method." Scientific thinking takes many forms, and the source of its spectacular successes is a set of norms and practices for inquiry and acceptance of theories that go beyond some general steps for qualitatively assessing how similar two objects are and what the degree of similarity implies about a possible association between the objects.ACE-V is a useful mnemonic acronym that stands for analysis, comparison, evaluation, and verification ... . Although [ACE-V was] not originally named that way, pioneers in forensic science were already applying such a protocol (Heindi 1927; Locard 193]). ... Its. It is a protocol that does not, in itself give details as to how the inference is conducted. Most authors stay at this descriptive stage and leave the inferential or decision component of the process to "training and experience" without giving any more guidance as to how examiners arrive at their decisions. As rightly highlighted in the NRC report (National Research Council 2009, pp. 5-12): "ACE-V provides a broadly stated framework for conducting friction ridge analyses. However, this framework is not specific enough to qualify as a validated method for this type of analysis." Some have compared the steps of ACE-V to the steps of standard hypothesis testing, described generally as the "scientific method" (Wertheim 2000; Triplett and Cooney 2006; Reznicek et al. 2010: Brewer 2014). We agree that ACE-V reflects good forensic practice and that there is an element of peer review in the verification stage ... ; however, draping ACE-V with the term "scientific method" runs the risk of giving this acronym more weight than it deserves. (Pp. 34-35).
I would offer a slightly different take on the matter. Identity has always been an elusive topic. As history has unfolded, scientific explanation of it has shifted as well. This dates back to pre-Socratic notions of mereology, to Gotlob Frege’s second order derivative logic, Ship of Theseus Paradox, type and token theories of identity, and up to more recent notions of John Searle’s social constructions of identity through status functions. I tend to agree with Searle and would offer up the notion that as it relates to fingerprints, identity is an artefact, or one created by institutions of man. This would be contrasted with brute facts, which are independently observable facts of nature. It is this difference (Institutional vs Brute facts) which delineates physical from social science, as social science is not based solely on physical objects, but the attitudes people have about them. It is in some sense, metaphysical as all questions of value are and thus requires the cognitive apparatus of recognition and moral decision making as best described by Joshua Greene or Daniel Kahneman.The reason is simple, fingerprints do not compare themselves. It takes humans or algorithms (read: mathematical formulas which embody human values) to do so.
Fingerprint identification definitely has at its core an antiquated notion of philosophy of science, namely the deductive-nomological (DN) model. This model assumes that from law like (nomological) conditions, one can deduce the outcome. Obviously, the most famous notions of this model of scientific explanation are Newtwon’s laws of motion. E.g. ‘An object in motion tends to stay in motion unless acted upon by an outside force’. No one has ever demanded proof of such a statement nor is one possible, given what we know about quantum mechanics. Namely, that there is no such thing as an object existing without force being exerted onto it. However, this law was used to successfully make predictions about orbits, further Kepler’s work and confirm the Copernican theory of heliocentrism. The only components required of this model is the law itself, the necessary components, and a criteria of sufficiency. Historically speaking as applied, the laws were Uniqueness and permanence, the necessary components of fingerprint identification were Galton points and what was sufficient was approximately 8. From this we deduced identification. However, as it is quickly apparent, notions of what counts as necessary and the criteria of sufficiency are hard pressed against attitudes of Examiners. The question of sufficiency has always loomed large over the discipline and to a lesser degree what features in fingerprints are necessary. (ie level 2, level 3). This ultimately led to an explicit recognition by the IAI in 1973 that "No valid basis exists at this time for requiring that a predetermined minimum number of friction ridge characteristics must be present in two impressions in order to establish positive identification."
The shift from the DN model to ACE-V as an articulated method further reinforces this notion. It is a tacit recognition that perception and recognition of similarity is an intuitive ability of humans. ACE-V represents what Jurgen Habermas/Imre Lakatos referred to as a Rational Reconstruction. This rational reconstruction is an attempt to explicitly systematize and structure behaviors and interpretations. While there is and has been an attempt to port the notion of absoluteness of the DN model to the rational reconstruction of ACE-V (i.e. If applied correctly, the error rate of ACE-V is zero), ultimately this approach has been found to shift the conversation to an inductive method of identification. This is expressed as patterns of similarity and/or patterns of experience (i.e. I have not seen nor would I expect to see two prints with such similarity to have come from a different source). Now the attempt is to shift to an inductive statistical method.
The attempt to statistically quantify identity has been threaded through this conversation for some time, although with less prominence. Originating with the Galton’s attempt to quantify statistical odds of finding matching points and moving through several generations as computational speed and accuracy has advanced with technology, it has recently made its way to the forefront again. Citing the success of DNA’s use of likelihood ratio, an attempt to formalize the statistical nature of fingerprint identitification has been met with limited success and a healthy dose of skepticism. However, there’s four large pressing issues at hand. Firstly, it is not just about a purely inductively statistical method, it is about a statistically relevant model. Fingerprints are not deterministic and there is no Hardy-Weinberg equivalent of frequency for Galton points in a population. Secondly, there is a debate about whether or not there is a frequentist or Bayesian approach to the best type of model. Even within the Bayesian camp there’s a debate as to whether there is a global Bayesian model or a personal Bayesian model. Thirdly, looking back to our Rational Reconstruction notion of explanation, there is an interpretive element to feeding any statistical model and statistics have no litmus test to say what is relevant given interpretation, or how much relevance at any single point of similarity should be imported into any likelihood ratio. (i.e. priors). Lastly, there is an inherent under-determination of the models that is best captured by the famous attributed to George Box “All models are wrong, but some are useful”. Statistical models only make use of limited information, which cannot account for objective vs interpretive assessments, and humans can outperform models on prints that have limited clarity.
The practical tension and nature of such an argument is outlined in
this document by NIST. It states:
And then goes on to state“We know that when humans analyze a crime scene fingerprint, the process is inherently subjective,” said Elham Tabassi, a computer engineer at NIST and a co-author of the study. “By reducing the human subjectivity, we can make fingerprint analysis more reliable and more efficient.”
So, to become more objective, scientists use human subjectivity to train their machine learning algorithms? Sounds about right.....The researchers used machine learning to build their algorithm. Unlike traditional programming in which you write out explicit instructions for a computer to follow, in machine learning, you train the computer to recognize patterns by showing it examples.
To get training examples, the researchers had 31 fingerprint experts analyze 100 latent prints each, scoring the quality of each on a scale of 1 to 5. Those prints and their scores were used to train the algorithm to determine how much information a latent print contains.
