Note: This is the fifth of a six-part online symposium on the use of forensics in the criminal justice system. You can read the introduction and first four parts here, here, here and here. And you can read the full biographies of the panel members here.
There have been some studies showing that the more information a forensic analyst gets from law enforcement officials, the more likely they are to produce a false positive. Good scientists know that cognitive bias is a persistent threat to sound research, and they take precautions such as double-blind testing to keep it from corrupting their work.
Yet many forensic analysts insist they can’t perform their jobs without consulting with law enforcement. Medical examiners in particular — who work in a field that can sometimes require subjective analysis, but that is also generally accepted in the scientific community — say that the context that law enforcement officials provide can be essential for determining the manner of death. But there have also been plenty of cases where the police gave the medical examiner bad information and corrupted the diagnosis.
Where is the line here? How much interaction between law enforcement and a forensic analyst is appropriate, and what safeguards can be put in place to minimize cognitive bias?
John Lentini, fire/arson expert:
Sometimes context is essential. Other times it is not. Having a forensic “case manager” who can protect an analyst from domain-irrelevant data could minimize contextual bias. The case manager can decide what testing is appropriate, and the analyst can conduct the testing without knowledge of which side the answer helps.
Medical examiners present a different problem. The corpses they examine are “patients” who require a diagnosis, and no doctor attempts a cause-of-death diagnosis without a medical history. Manner of death is another question. In my field, medical examiners must rely on fire investigators who may not understand the limits of their qualifications or the unreliability of their origin determinations. Medical examiners could possibly assign a “conditional” manner of death. They could report something like, “assuming that the fire investigator is correct in his determination that the fire was intentionally set, the manner of death is homicide.” This would allow them for a little wiggle room if the person doing the origin-and-cause investigation is somehow mistaken.
Jules Epstein, Temple University Beasley School of Law; National Commission on Forensic Science:
What is “domain relevant” varies from discipline to discipline, and may indeed vary within disciplines — e.g. medical examiners may need to know more in one category of case (drowning?) than in another (multiple gunshot wounds?). What is critical is that disciplines continue to assess this question, and that forensic labs and medical examiner offices work on implementing information management systems to reduce the flow of domain-irrelevant information. It is also essential that all information received from law enforcement be documented.
Itiel Dror, University College London; Cognitive Consultants International:
This is quite simple to resolve. Yes, experts need to interact with law enforcement; first, to determine what information is relevant and is needed for the scientific analyses (we do not want law enforcement to decide that), and also, after the forensic analysis, someone needs to explain to law enforcement what the expert evidence means and how it relates to the crime at hand.
However, these interactions can also bias the experts, as has been demonstrated even with the more robust forensic evidence, such as fingerprinting and DNA mixtures.
The solution is simple, and I and others have suggested it many times before: Have a case manager who interacts with law enforcement but who doesn’t do any of the actual forensic work itself. The case manager will give only the task relevant information to another examiner, who will then do the actual forensic work.
Sandra Guerra Thompson, University of Houston Law Center; Houston Forensic Science Center:
From my experience on the board of the Houston Forensic Science Center, I know that the HFSC staff has taken steps to reduce cognitive bias. Early in the lab’s existence, the lab’s managers pioneered a proactive approach to reducing the influence of extrinsic contextual information received from law enforcement. They created a “Client Services/Case Management” division to receive evidence from law enforcement. Case managers decide what information analysts need for testing and what evidence is irrelevant. They convey information to analysts on a need-to-know basis.
For example, a forensic biologist may need to know the areas on a garment where an investigator expects to find DNA material. Thus, case managers will provide this information. However, the identity of the suspect, the nature of the crime and other contextual information not pertinent to the testing process is not disclosed. In this way, case managers can shield analysts from information that has the potential to introduce cognitive bias.
Keith A. Findley, Center for Integrity in Forensic Science; University of Wisconsin Law School:
Cognitive biases are a part of the human condition. No one, not even the most rigorous academic scientists, [is] immune from them. The biases can come in all sorts of forms that are well researched by social scientists. Some of the most prominent, and the ones that are most problematic in the forensic science world, include confirmation bias, role effects, and context bias. The more subjective a judgment is, the more likely it is to be tainted by the kinds of biasing information that leads to these cognitive distortions.
Unfortunately, most of the pattern-matching forensic disciplines are highly subjective. Even in much more objective scientific inquiries, academic and laboratory scientists know that the risk of unintentional cognitive biases demands that they conduct their work in a double-blind manner — so that they are shielded from as much potentially biasing information as possible. But almost uniquely among those who claim to be doing scientific work, forensic scientists do their work in an environment in which little, or often no, attempt is made to shield the examiner from biasing information. Routinely, evidence submissions from law enforcement to the laboratories tell the examiner who the suspect is, what other evidence there is against the suspect and even law enforcement’s desired outcome (to link the evidence to the suspect). Contamination with contextual information is the norm.
Medical examiners, along with child-abuse pediatricians and other physicians called upon to render opinions about cause and manner of death or injury, are particularly vociferous in insisting that they must have all the context information, because it helps inform their determinations about what happened. And, certainly, such information can be useful to physicians when performing their investigative roles as a part of a multidisciplinary child abuse team, or as a medical examiner required by statute to determine cause and manner of death for the death certificate.
But the physician’s role is quite different when she enters the courtroom to render an expert opinion. The jury is the ultimate fact-finder in the courtroom, and the jury has full capacity to consider all the nonmedical evidence in the case. Although most courts today permit medical experts to go beyond this, the medical expert’s role should be just to offer opinions about those things the jury cannot understand on its own — medical findings, like the presence of a subdural hematoma, or the presence of gunshot entry wounds, and any medically validated opinions about the kinds of things that could have caused those medical findings.
The expert goes too far when she offers opinions based on nonmedical evidence (such as the presence of a suicide note, or the fact that the child’s caregiver was not the child’s father and was an unhappy babysitter) related to matters that are not medical questions — like whether some third party inflicted the injuries observed by the physician, and whether that person must have done so with intent to kill or with recklessness. And while courts routinely permit physicians to exceed these limits, some courts and legal authorities are finally beginning to take note that such broad testimony, based on biasing information, on matters that exceed true medical diagnosis, goes too far.
Chris Fabricant, Innocence Project:
I’ve not heard many compelling arguments for giving forensic practitioners any contextual information, apart from latent fingerprint examiners or tool mark experts needing to know the substrate from which the mark in question was lifted. At the very least, every crime lab should have an evidence manager to control the flow of case information to individual forensic practitioners on a need-to-know basis. And that information must [be] documented and shared with the defense.
Michael Risinger, Seton Hall University School of Law; Last Resort Exoneration Project:
I think this is the issue upon which there has been most progress in the last two decades, although, as with changes in institutional practice generally, the process has been slow and there is much left to be done.
Twenty years ago, virtually every forensic practitioner resisted the notion that their judgment could be affected by biasing information to which they might be exposed in the course of their examinations. Now, there is a fairly general recognition that this can happen. In addition, there is at least theoretic agreement that exposure to any information about a case that is not relevant to the expert’s function should be eliminated to the extent possible, and that, even when one is dealing with information which the expert should take into consideration in performing his or her expert function, that information should be released to the practitioner in the least biasing order.
Finally, it is generally recognized that protocols to reduce the likelihood of such context bias are a good thing, at least in theory. The debates occur over the details of what those protocols should be and how they should be administered. On that front, there is still a long way to go.
There are areas where there’s still resistance to the notion that such protocols are necessary, and forensic pathology [medical examiners] is probably chief among those. There are a whole host of reasons for this resistance — some historical, some having to do with insistence that determining the cause of death is “practicing medicine” just like any other context of practice, and that, as MDs, they are obliged to collect “history” and trained to give it its proper weight. It doesn’t help that a medical examiner is expected to determine not only the cause of death but also the manner of death, which is a nonexpert legal conclusion (inherited from the function of the common-law coroner, who was a judicial officer) concerning whether a death was the result of homicide, suicide, accident or by natural causes. The information relevant to those determinations (often properly coming from police or other law enforcement sources) is often irrelevant to and potentially biasing in regard to the accurate determination of the cause of death.
Perhaps we’re moving to the point where we can discuss control protocols in at least a subset of cases, such as police-involved shootings. I hope so.
Judy Melinek, forensic pathologist; author of “Working Stiff”:
I have to take issue with a definition of cognitive bias based on “the more information a forensic analyst gets from law enforcement officials, the more likely they are to produce a false positive.” Studies that apply to one discipline of forensic science may not apply to another. For example, a DNA analyst doesn’t need to know that the weapon recovered from the defendant’s car was a double-edged knife with a decorative hilt, but the forensic pathologist might need to know what it looked like to determine whether it matched the stab wounds on the dead body based on size, shape and the wound pattern.
Scene investigation is an intrinsic and inseparable part of forensic pathology. You are likely to make significant diagnostic errors if you try to look at a dead body in the absence of scene information. For example, trauma caused by resuscitative efforts or the final resting position of a dead body can be misinterpreted at autopsy as being due to assault or strangulation. Asking a forensic pathologist to interpret a dead body based on the autopsy findings alone, in the absence of scene and circumstances, is medical malpractice — akin to having a surgeon cut into you without ever reading your medical records or examining your clinical history.
One way of preventing misleading information from biasing a forensic pathologist is to ensure that the pathologist isn’t working for a law enforcement agency, and that the police aren’t present at the autopsy. Pathologists should have access to the death scenes in suspicious cases, and any information received from law enforcement agencies which informs the pathologist’s opinion on cause and manner of death should be documented in writing.
Frederic Whitehurst, FBI crime-lab whistleblower; Forensic Justice Project:
In my experience, law enforcement personnel have to make split-second decisions, move to control potentially dangerous situations and demand absolute compliance. These are necessary traits for that particular part of the government prosecutor’s team. But when those same people are part of the decision-making team in the forensics lab, they bring with them their need for absolute control and the compulsion to demand compliance. Too often, law enforcement official[s] demand that lab personnel simply agree with their hypothesis about the crime, rather than test it. Lab analysts get jerked around from one “truth” to the next. An honest lab analyst may quickly find himself out of a job, or put in a corner and occupied with nonwork.
Again, there’s an ethical dilemma. Do I take care of my family and keep my job, or do I uphold my oath of office? Too often, the people promoted to lab managers are people who have limited scientific background, who are considered “team players” and who don’t support the analysts who tell the truth.
But there is a balance. Without knowledge and context of the crime, an analyst could spend a long time looking for evidence that’s ultimately irrelevant to the case. A white powder handed over to a crime lab’s drug section might come with a request to test for controlled substances, but it shouldn’t include a description of the suspect — not race, age, gender or even name. Just the relevant information to do the job. Too often, an officer is on the phone telling the analyst what the test results should be. At the very least, communication between the lab and law enforcement should also be in writing, or not at all.
Read more from Radley Balko: