The DEA's investigations produce a huge volume of wiretaps and other monitored conversations. Many are in other languages. Back when I was a fed, the DEA agents were free to pick whichever professional translator they wanted to produce English-language transcripts. Many translators advertised to the DEA agents. In the late 1990s, the criminal defense bar managed to get copies of some of those advertisements, and used them to devastating effect in cross-examinations. See, some translators would pitch their services by suggesting that they would skew the translations to support the government's theory of the case, saying things like "pick the right translator, because one word can mean the difference between guilty or not guilty."
This produced unpleasant tension between federal prosecutors and the DEA. The AUSAs didn't want to try cases based on tainted translations, and the DEA agents were loyal to their pet translators and saw nothing wrong with translators as advocates. Sometimes the federal prosecutors were obligated to pay with their own budget to have the tapes re-translated by a translator without such baggage.
That's always going to be a hazard when cops and prosecutors seek expert services in cases, whether those services are linguistic or forensic. So when I saw that Radley Balko was using his new platform at HufPo to suggest privatization of crime labs, I was skeptical. Balko has done great work on crime lab issues at Reason for years, but how could privatization work?
Fortunately his proposal is structured to prevent the sort of incentives and biases I'm concerned about. Balko is advocating a system suggested by Reason Foundation writer Roger Koppl:
Instead, Koppl's plan would use competition to remedy the incentive and cognitive bias problems that occur when analysts who are supposed to be objective work for and report to the same government agencies that then use their results to try to put defendants in prison.
Under Koppl's plan, a city or state would create a position of "evidence handler." The evidence handler's job would be to distribute the testable evidence in a case to the appropriate crime lab. Under a fully privatized system, the evidence handler would distribute it to one of a rotating series of private labs. Under a partially-privatized system, there would still be a state lab, but under both systems, in every third case or so, the evidence would be sumbitted to a second or third lab for verification. The original lab would not know when it was being checked by other labs.
This system, which Koppl calls "rivalrous redundancy," flips the incentive problem upside down. For the individual crime lab worker, the incentive is no longer to please prosecutors or police, but to do the most thorough, sound, objective analysis possible. For the private labs, the incentive is to catch the state labs — or another private lab — making a mistake. When there's conflict over test results, a third or fourth lab could come into the mix.
I think it's a start. The key would be to expose the private lab selection process to exacting scrutiny, to avoid selection based on promises of "helpful" results. We'd also need to ensure that when private labs expose shoddy work by state labs, there would be real consequences for state lab workers — no easy task when those workers belong to government employee unions. We'd have to watch for revolving-door systems where "helpful" performance at a private lab can lead to a job at a public one, or vice-versa — the sort of rewards system that taints Congress and our federal agencies.
Mostly, we'd have to continue to work to change public attitudes about the role of law enforcement, hardened by two generations of insipid law-and-order rhetoric. The glut of forensic shows on television only encourage the public to view lab workers as righteous advocates, actively involved in investigations and questing for information that will confirm their suspicions about defendants. The public is still far more upset about crime labs that fail to catch criminals — as in the cases of untested rape kits — than they are about crime labs that bungle, stretch, or falsify results to produce questionable convictions. In short, for now, too few people give a shit.