Forensic Delusions

A groundbreaking report by the National Academy of Sciences (NAS) last year recognized that much of forensic science is not rooted in solid science. Many forensic disciplines — such as hair microscopy, bite-mark comparisons, fingerprint analysis, firearm testing and tool-mark analysis — were developed solely to solve crimes.  They evolved mainly in the context of individual cases, which often had significant variation in resources and expertise. They have not been subjected to rigorous experimental scrutiny, and there are no standards or oversight in the United States or elsewhere to ensure that validated, reliable forensic methods are used consistently. With the exception of DNA analysis, no forensic method has been proved to reliably and accurately demonstrate a connection between evidence and a specific source. …

We advocate the creation of an office of forensic science improvement and support (OFSIS) within the US Department of Commerce to spur independent research, develop standards and ensure compliance. … [Some] within the forensic-science and law-enforcement fields … argue that an OFSIS is not necessary and that laboratory accreditation is sufficient. … They argue that an OFSIS would cost too much … and that it could create chaos in the US justice system by reopening countless old cases. … Political and criminal-justice ends — rather than research imperatives — have taken forensic science off course. … See

More here.  The primary social pressure on law court practices is for courts to give the appearance of punishing guilty folks.  Observers have much less info on who is actually guilty.  So the main pressure on legal standards is that officially-accepted evidence seem to the public, juries, and judges to indicate guilt, not that it actually indicate guilt.  We expect the law to be overconfident about its evidence.

Requiring that legal evidence standards stand up to independent experimental scrutiny would create a more accurate legal system, but at the expense of reducing the apparent rate at which the guilty are punished.  So I expect, sadly, this proposal to be rejected.

GD Star Rating
Tagged as: ,
Trackback URL:
  • Robin, some of what you are asking for already exists with respect to fingerprints at NIST. Check out

  • Dave
  • Pingback: uberVU - social comments()

  • komponisto


    As you point out, the problem with criminal justice systems is more fundamental than the inadequacy of any particular forensic technique. It’s that people aren’t all that wedded to the idea that verdicts need to reflect true beliefs in the first place. Instead, they’re mostly about certifying that the prosecuting authorities “followed the rules”; if so, juries feel, then they are owed deference.

    (The sentencing report in the Knox case, released earlier this month, hasn’t yet made it online; but reports about its content indicate that the jury had a hard time believing that Knox and Sollecito were really capable of the crime, but more or less felt an obligation to convict based on the fact that the prosecution went to the trouble of collecting all the various pieces of “evidence” against them.)

    • komponisto

      The sentencing report…hasn’t yet made it online

      Correction: here it is.

  • bellisaurius

    The legal system is more about preventing revenge and lex talonis than finding the truth. I’d have to agree, and on the whole, this may not be a bad thing (for individuals, it’s awful of course). It keeps order, which benefits the greatest number the most.

    The part that irks me on this topic has more to do with how people look at science in general, as if each test is accurate and there are no outliers; but hey, I guess that basically matches it up with most science fiction. Plus, I don’t think the system gives enough credit to juries for figuring things out. People tend to get smarter when they think a task is important, and they’re being judged by others for their wisdom. If there’s some uncertainty in the analysis, then say so. Plus, if a person’s getting convicted on one piece of evidence, they were screwed anyway (sort of like In my job, a single alarm or indication gets attention, two independent pieces of data though, get immediate reaction). It should take an entire case, which creates a solid narrative to convince people of someone having done something.

  • John Maxwell IV

    Maybe this is just a PR issue. Maybe if several well-received movies were made about innocent people being convicted and serving horrible sentences then the general public would better balance their concern to punish the guilty and their concern to exonerate the innocent.

    • komponisto

      This was my first thought too; unfortunately in most cases it’s absurdly difficult to convince people that someone convicted of a crime is actually innocent. (The main exceptions are when the case is a proxy for some larger political battle.)

  • Lo Statuz

    The application of DNA testing to old convictions revealed that a lot of those convicted were not guilty. This provides a rare research opportunity. It’s rare because there’s an independent standard against which to check the actual verdicts.

    Some questions I can think of:
    Do juries give better or worse results than bench trials?
    Do public defenders give better or worse results than private defense lawyers?
    How often do not-guilty defendants accept plea bargains?
    Are there any demographic characteristics of defendants that make results better or worse?
    Do appointed judges give better or worse results than elected judges?

  • Steven Schreiber

    It depends. The issue with scientific reports which indicate that a forensic method does not have any real backing is that they can directly effect court decisions.

    Forensic reform would arrive because too many cases are thrown out when defense attorneys produce research showing that the testimony is unreliable because analysis of the forensic evidence is necessarily faulty. In that case, no amount of sputtering about accreditation will do any good because you may as well have presented spectral evidence.

    • If we don’t approve a process to create the evidence that would throw out such evidence, it will continue to be accepted.

  • Robin,

    Click on my name for a different perspective on forensic science errors. IMHO we need checks and balances, not command and cotrol.

  • Are there any non governmental approaches available to try to address these problems? I’m concerned that a new agency would inherently be political, and the constituency that most cared about the results would be investigators and prosecutors, and so the results and findings would be skewed toward them. I realize that most funding for research comes from federal coffers, so it’s pretty hard to think of an alternative source for underwriting new research, but it’s pretty dispiriting to think that the only way to fight this broken system is to argue for funding that is likely to be highjacked by the people with the most to gain from allowing them to continue using flawed techniques.

  • Hamish Barney

    This reminds me of Leeson’s work the effectiveness of medieval trial by ordeal. That they worked because most people believed that they worked. Maybe we haven’t made as much progress in criminal justice as we would like to think.

  • Robin,

    We can refine your point about merely appearing to get the bad guys. Conviction rates are an important criterion for judging police and prosecution. But that criterion discourages them from doing the very thing we most wish for them to do, namely, discriminate between the guilty and the innocent. All potential suspects become fungible and the police and prosecution have an incentive to go after anyone they think might be easy to convict. (We have a multitask problem in criminal justice.) Basic integrity and decency in police and prosecutors mitigate the problem greatly, but it’s still there. And some police and prosecutors are unscrupulous, of course. There is some evidence that police sometimes just go after poor, uneducated, or mentally weak persons, who may be easier to convict. Since crime labs are organized under the police, they have an incentive to support the police and prosecution position. These organizational problems in the criminal justice system tend to create forensic science errors.

    I address some of the issues in this comment on a paper by Richard A. Posner. Posner’s reply at page 140.

    This way of looking at it leads you to different solutions than those proposed by Scheck and Neufeld. It is hard to exaggerate the good they have done in this area, but their reform ideas are built around the idea of oversight, which carries the risk of regulatory capture and misses many of the organizational issues at the root of the problem.

  • Roger and Chris, I’m not endorsing particular solutions to produce more skepticism and rigor relative to other ones here; here I’m more endorsing that entire class of solutions relative to the status quo.