A few forensic tid-bits, some more forensic-y than others. Still getting back on track after the American Academy of Forensic Sciences meeting. It’s amazing how much work doesn’t get done while you’re working…
American Bar Association: The vast majority of criminal cases end in plea bargains
Remember the old quality aphorism: “Fast, cheap, good—pick two”? Well, it seems that U.S. prosecutors, defenders, and judges picked the first two. A new report from the ABA says that 98% of criminal cases in the federal courts end with a plea bargain. This compromises fairness, as innocent people are coerced into plea agreements by prosecutors looking for a quick win. Trials are “rare legal artifacts” in the federal system and even at the state level: Pennsylvania, Texas, and New York have trial rates of less than 3%. In Santa Cruz County, Ariz., there were no trials from 2010 to 2012. This skimpy number of trials is probably a fallout from the U.S. government deciding in the 1970s that top executives could commit terrible financial crimes and go to prison. Financial crimes are exceedingly complicated and difficult to litigate, so the culture tipped towards plea agreements rather than taking chances of losing at trial. Prosecutors too scared to go to trial and flummoxed by the legal obstacles in doing so were dubbed members of the “Chickenshit Club.” Attorneys do learn from each other, after all, and who wants to “lose”? Especially the innocent people who get bullied into bad pleas…
Forensic Science* is a Distinct and Separate Discipline
Yep, I’m back on this horse again, just changed saddles. The distinction between basic and applied sciences is old hat, a tired argument in the face of discovery, application, and cross-pollination of disciplines. Our understanding of the natural world, long the bastion of “basic science,” can come from anywhere, if you have an open mind. Forensic science* can contribute to that understanding and bring a bit more science (sans *) with it. For example, conducting experiments on ballistic wound trauma on bones is difficult to conduct, for any number of anatomical, biomechanical, and ethical reasons. Synthetic bone material can be substituted but there is insufficient data in the literature to create models that allow extrapolation to real bone. Enter this recent study in the International Journal of Numerical Methods in Biomedical Engineering (and you thought I was wonky) that demonstrates the numerical model for Synbone® with a relative error lower than 10%. Check out the article: It’s impressive. Now, that is forensic science.
A Forensic Without the Science: Face Recognition in U.S. Criminal Investigations
“As a biometric, forensic investigative tool, face recognition may be particularly prone to errors arising from subjective human judgment, cognitive bias, low-quality or manipulated evidence, and under-performing technology.”
No kidding.
“…it is important to point out that face recognition doesn’t work well enough to reliably serve the purposes for which law enforcement agencies themselves want to use it.”
Couldn’t have said it better myself.
These are the key findings of a report by the Georgetown Law Center on Privacy & Technology, with the title listed above, are:
In its current form, face recognition is an unreliable source of identity evidence.
The algorithmic and human steps in a face recognition search each may compound mistakes.
Faces are inherently biasing, we are hard-wired to recognize them, and, therefore, it may be impossible to remove the risk of bias and error.
Even though we’ve been told they haven’t, face recognition has been used as probable cause to make arrests.
Face recognition searches have been used as evidence in criminal cases, and the accused have been deprived of the opportunity to challenge it.
The harms resulting from wrongful arrests and investigations are real, even if they are hard to quantify.
Like other questionable technologies, face recognition is forensic-adjacent and we need to be up to date on the science (and science*), issues, and outcomes.
Read the report here.