Friday, January 27, 2012

Working blind

Ask any scientist and she'll tell you that the best result from an experiment is often the one that disproves the hypothesis being tested. Why? Because such a result means there are more questions to ask - and more knowledge to be gained.

According to the scientific method, we first develop a hypothesis, then we conduct experiments. We observe the data obtained and compare the results to our hypothesis. If the data contradicts the hypothesis we must change our hypothesis. That, you see, is the recipe for scientific breakthroughs.

But such a scenario is anathema in a crime lab.

The last thing the police want is an inconclusive test. Analysts know this. Their supervisors know this. The people funding the lab know this.

And so the "forensic scientists" at crime labs across the country are made part of the criminal investigation. They are given the task of proving a link between the test result and the suspect. And, as many results are subject to interpretation, there is an inherent bias to be found.

Grits for Breakfast referenced an article in The Economist that when forensic scientists are given too much "contextual information," test results can be subject to a "cognitive bias."

As Grits points out medical trials work because the doctors conducting them don't know who's receiving the actual treatment and who's receiving the placebo - and neither do the test subjects. Without that knowledge there is no pressure (whether real or imagined) to make the results fit the desired outcome.

In the criminal (in)justice context, however, the forensic scientists, or lab technicians, running the test are told they are testing something that came from the suspect. They are told what that person is suspected of doing. They are told they're part of the team. Their mission, in other words, is to provide the evidence the government needs to obtain a conviction.

They are neither neutral nor unbiased. Just think what role that ideology played in the expansion of forensic analysis into the analysis of tire tracks and foot prints. Fingerprints and bite marks. Ballistics and lead analysis. Arson investigations.

If we are going to call these glorified lab techs "forensic scientists," then we should at least expect them to play by the same rules other scientists follow. It's the least we should demand when people's lives are on the line.


EB said...

It's hard to imagine how some forensic analyses could be done in a totally blind fashion. For instance, in blood alcohol tests for dui the analyst will know based on the test requested, the type of sample, and the packaging of the sample that it is a blood sample from someone accused of dui. Similarly in most DNA testing. To do a criminal paternity test the analyst must know which samples came from the child, the mother, and the alleged father. In rape cases the analyst will know from the types and packaging which samples came from the rape kit, the victim, and the alleged rapist. For interpretation of mixtures where a likelihood ratio statistic must be calculated the analyst must evaluate alternative hypotheses that would explain the evidence, so the analyst must know the prosecution hypothesis of the crime which includes knowing which profile comes from the alleged perpetrator. The same for firearms analysis. The analyst will know from the item and the packaging that a bullet came from an autopsy. The analyst will know that the gun they are comparing it to was suspected of being used. Why else would the detective submit it?

So truly blind testing is not an achievable goal.

The better approach that is achievable is two fold:

1) Separate labs from police agencies and district attorney offices so the staff of laboratories are not part of the prosecution team.

2) Encourage, promote and make possible extensive review of test records and retesting of evidence by the defense so that possible errors and biases can be identified.

The true gold standard of quality in science is not blind testing - the vast majority of scientific studies are not done in a blind fashion. The real gold standard of science is reproducibility of findings based upon retesting.

Retesting is almost always an option when it comes to forensic analysis, and often the defense attorney can get the court to pay for it. But, for whatever reason, defense attorneys rarely request retesting, and they rarely get their own expert to review the testing results. I don't really understand that. If I were a defendant I would sure want retesting done.

Paul B. Kennedy said...

Thank you for your comments. I will probably be writing another post based on the points you made.