Sunday, June 10, 2012

Who Is Nelson Acosta-Roque? (Part II)

Back in March, Who Is Nelson Acosta-Roque? (Part I) described the pending appeal of an Alaska waiter being deported because the government believes he is really Victor Antonio Pecheca-Aromboles — a previously deported Pennsylvania cocaine dealer from the Dominican Republic who changed his name and returned to the U.S. The Department of Justice resolved the question of the true identity on the basis of “trust-me testimony.” Below are some key questions and answers in the hearing in 2010, along with some thoughts about "100% certain" testimony. Susan R. Blei, supervisor of Alaska’s Criminal Records and Identification Bureau, is testifying over the telephone:
Q. [W]hat is your occupation?
A. I'm a fingerprint analyst and I supervise the criminal records and identification bureau. * * *
Q. Okay now, how do you — if you can answer this, how do you actually identify someone by means of fingerprints?
A. I have two impressions that I'm comparing, I use — I personally use two fingerprint magnifiers and two pointers is what we call them, and I'll have a glass over each impression, and I use the pointer to what I call walk through the ridges and count my points. * * *
Comment: That’s it? Just count the number of Level 2 details in common, with no concern for which ones are common and which ones are rare? How many common points were there in this case? How many fingers were compared? The lawyer for the Department of Homeland Security and the immigration judge never thought to ask. And neither did Mr. Acosta-Roque, who was representing himself. You would think that Ms. Blei knows, but she has not responded to my emails.
Q. What is the likelihood of two people possessing the same fingerprints?
A. That won't happen. No two people have the same fingerprints.
Q. Okay, is that — and how — I mean, how do you know that? Is there a scientific study?
A. There have been scientific studies, yes.
Q. So even identical twins wouldn't have the same fingerprints?
A. No, they would not.
* * *
Comment: These answers attracted the scorn of the 39 scholars and scientists, whose amicus brief stated that “The record does not report what studies the expert was referring to, and amici have found no study that proves that ‘No two people have the same fingerprints.’ Amici are skeptical that it would be possible to even conduct a study that proved that proposition." [1]

Impossible to prove? Well, it cannot be proved directly, by a census of all fingerprints that exist, that have existed, and that will exist [2], but studies of identical twins show that their fingerprints are distinguishable, and statistical analyses purport to show very small probabilities of duplicating a fingerprint pattern. For example, the NIST report, released in February, observes that “[o]ne study reports that in 100,000 randomly chosen fingerprints of exemplar quality, there is only a 10–14 probability that some pair of them will match in regard to both minutiae and ridge shape.” [3] Of course, there are more than 100,000 fingerprints in the world’s population, but the same generative model estimates the probability of at least one match among 10 billion to be less than 0.0001. [4]

To be sure, the fact that some scientific studies purport to establish the uniqueness of complete fingerprints does not conclusively prove the point. But it does make the statement that “[t]here have been scientific studies, yes,” understandable as a response to the lawyer’s question.

The problem with the expert’s answer, then, is not that it refers to a fact outside the domain of potential scientific knowledge (which is always tentative in the sense of being subject to revision) [5].  The problem is that it fails to inform the judge that the expert's belief can be and has been disputed. The assertion that "there have been scientific studies, yes" may be the literal truth, but it is one-sided, and an expert forensic scientist or technician has some responsibility to tell the whole truth. The NIST report explains:
During a trial, counsel’s questioning frames the expert’s presentation. The expert may not simply decide what information to discuss but must answer counsel’s questions. This can create some tension between the goal of being complete and the need to be responsive. In resolving this tension, “ethical considerations and professional standards properly place a number of constraints on the expert’s behavior.” One such constraint is “a requirement of candor. While an expert is ordinarily under no legal obligation to volunteer information, professional ethics may compel this ... when the expert believes that withholding information will change dramatically the picture that his … analyses, properly understood, convey.” Thus, the “guiding principles” proposed by the American Society of Crime Laboratory Directors/Laboratory Accreditation Board (ASCLD/LAB) advise forensic scientists to “attempt to qualify their responses while testifying when asked a question with the requirement that a simple ‘yes’ or ‘no’ answer be given, if answering ‘yes’ or ‘no’ would be misleading to the judge or the jury.” [3, p. 115 (notes omitted)]
In Acosta-Roque, it would not have been hard for the expert to answer, “Yes, there have been scientific studies that bear on the question of whether every person has distinguishable fingerprints. Although the conclusion is accepted in most forensic science textbooks and fingerprint identification manuals, some commentators have questioned it.”

The amicus brief also chastises the expert for not listing specific studies, stating that “forensic expert witnesses have an ethical obligation to specify to courts the studies they rely upon in forming their conclusions and should refrain from the dubious practice of invoking ‘studies’ without specifying what those ‘studies’ are.” However, experts should not be expected to cite studies on every general proposition that is part of their reasoning. Certainly, an expert should not try to bolster testimony with uncalled for references to unnamed “studies”; however, it is fair to answer a lawyer’s unanticipated question, “Is there a scientific study?” in the affirmative even without being able to name the study that the expert knows exists.

Here, the expert could have replied, “Yes, there have been scientific studies that bear on the question of whether every person has distinguishable fingerprints. Because I did not know you would be asking about these, I do not have a list of them at hand.”
Q. Were you able to obtain the Pennsylvania fingerprints?
A. Yes, I was.
* * *
Q. And did you compare all four sets of prints?
A. Yes, I did.
Q. And what was your conclusion upon comparing the prints?
A. They were made by one and the same individual.
Comment: This is source-attribution testimony for ten-print cards. It is based on the theory that an examiner can distinguish every possible pair of prints from the same individual’s ten fingers from every possible pair of ten-print cards from different individuals. The recent reports of the National Academy of Sciences and the NIST Expert Working Group emphasized in the amicus brief do not focus on this claim. Rather, they concern latent print identification, in which far less information is available. In that context, the NIST report concludes that “[g]iven the current state of scientific and professional knowledge, . . . it is best to avoid [source-attribution] testimony based on the theory of global general uniqueness.” [3, p. 138] The same point could be made for ten-print identifications. There are viable alternatives to source attributions on the part of the testifying expert and to reliance on the difficult-to-prove “theory of global general uniqueness.”
Q. Okay, now how did you go about comparing these sets of prints?
A. I began with the Pennsylvania prints and I compared them one at a time to each of the other cards.
Q. And why did you use the Pennsylvania prints?
A. They were the ones that had been requested. You could do it in any order. That was just the way I did it.
Q. Okay, and you determined that all four matched up to the same person?
A. Yes, they did.
Q. And did you have — was there any sort of back-up check done on this?
A. I did have another employee do comparisons on the very same cards, and that employee came to the same conclusion.
* * *
Comment: The record is skimpy about the verification. Was it blind? Appellate counsel for Acosta-Roque asserted that “Neither the initial analysis nor the verification were blind,” [6], but how does he know what the first examiner told the second? The shallow hearing reveals nothing other than that “another employee” looked at the “same cards and came to the same conclusion.”
Q. And so what are the chances that this could — that the fingerprints from Mr. Acosta-Roque maybe don't match Mr. Victor Arumboles?
A. They do match.
Q. Okay, so I mean, can you state — I mean, what are they — is there a percentage? Can you quantify it, or ...
A. No. No, I'm 100 percent sure the prints are the same and made by the same individual.
* * *
Comments: Despite the inarticulate questions, the witness professes 100% subjective certainty in her source attribution. The amicus brief denigrates all claims of certainty as irresponsible because human error is possible:
Claims of absolute certainty go beyond mere expressions of the degree of an expert’s confidence, and are, in essence, an unsupported form of bolstering the expert’s own testimony. In addition to failing to recognize the possibility of an erroneous match, they also fail to recognize that there might be other errors that could occur, such as mislabeling of a record within a database, or a misattribution of a source print to the incorrect individual. No human endeavor is always utterly error-free, and assertions of 100 percent certainty in this context are both misleading and scientifically unwarranted.
I know what the brief is driving at, but its rationale for eschewing "assertions of 100 certainty" is unconvincing. The analyst attributed the four prints she examined to the same source — "the prints [were] made by the same individual." This conclusion does not turn on how the exemplar prints were labelled.

As for the charge of "bolstering," a claim of total certainty is no less an “expression of the degree of an expert’s confidence” than is a claim of near certainty. Any assertion of a high degree of confidence could be overstated because of an inadequate perception of the risk of human error. Any assertion of great confidence “gives a boost to” a more tentatively stated  conclusion that all the fingerprint impressions on the four cards come from the same set of ten fingers. There is nothing wrong with this type of "bolstering" if (1) the asserted degree of confidence is warranted, and (2) the expert acknowledges that, like all statements about the world, it is subject to some degree of doubt. An expert who testifies that “I am as sure about this as I am about my own name or the fact that the earth orbits the sun” also could be mistaken. Degree-of-confidence testimony is misleading only if it is misunderstood as a statement that the objective probability that the same fingers are represented on all the cards is the specified percentage.

That said, it would be best for experts to avoid 100% even as an expression of subjective probability. From a Bayesian perspective, this level of certainty corresponds to accepting infinite odds on a bet that the match is false. Would Ms. Blie be willing to accept eternal damnation and unbearable torture if she were proved wrong? Total certitude means that no contrary evidence, no matter how convincing could change her mind. Thus, the use of 100% smacks of dogma or advocacy rather than a neutral statement of the evidence. Most likely, this extreme probability is what the amicus brief means by “bolstering.”

The more fundamental and general issue, however, is whether an assertion of a number bordering on 100% for the subjective certainty of identity is “scientifically unwarranted.” Science certainly permits a witness to report probabilities conditioned on explicit assumptions. Do scientific studies permit an examiner to assert that, having extensively analyzed sets of ten-print cards, she is (nearly) 100% confident that the only plausible hypothesis to explain the comparable patterns is a common source? If so, was her examination sufficient to warrant such a claim? Without more information on the actual examination than the government and the expert deigned to provide, the last question cannot be answered.
Q. Okay, and you concluded that they're one and the same to the exclusion of all others?
A. Yes, sir.
Q. Okay and I just — one last time, how certain are you about your conclusion that these prints belong to the same individual?
A. I'm 100 percent certain.
Comment: Objection — asked and answered.

This is not the end of the testimony. I'll pick up the thread and wind it up soon with a discussion of the amici's criticism of the expert's desire to have at least eight matching points before making a source attribution.


1. Brief of Scientists and Scholars of Fingerprint Identification as Amici Curiae in Support of Petitioner and in Favor of Reversal, Acosta-Roque v. Holder, No. 11-70705 (9th Cir. Mar. 8, 2012).

2. Michael Saks & Jonathan J. Koehler, The Individualization Fallacy in Forensic Science, 61 Vanderbilt L. Rev. 199 (2008).

3. NIST Expert Working Group on Human Factors in Latent Print Analysis, Latent Print Examination and Human Factors: Improving the Practice through a Systems Approach: The Report of the Expert Working Group on Human Factors in Latent Print Analysis 82 (David H. Kaye ed., 2012). The report adds that “[w]ithout the assumption of independence, however, the computed probability could be orders of magnitude higher.” Id.

4. Su, C. and S. Srihari. “Generative Models for Fingerprint Individuality Using Ridge Models.” in Proceedings 19th International Conference on Pattern Recognition, Tampa, Florida, Dec. 8–11, 2008 (tbl. 3).

5. David H. Kaye, Probability, Individualization, and Uniqueness in Forensic Science Evidence: Listening to the Academies, 75 Brooklyn L. Rev. 1163 (2010).

6. Caleb Mason, Scientific Validity of Fingerprint Comparison: New Case and Amicus Brief (Mar. 5, 2012),

No comments:

Post a Comment