Entries categorized "Junk Science"

The Blogosphere Reacts to the NAS's Report on Junk Science in the Courtroom

Norm Pattis has posts here, here, and here.  Scott Greenfield comments here.  Radley Balko's contribution is here

Somewhat relatedly: Here is an August '08 post noting that so-called criminal profiling is junk science.

Perhaps the next Blawg Review will have a section linking to various reactions to the NAS Report? 

Other reactions?  Please leave a comment and I'll update this post.

Criminal Profiling is Junk Science

The latest issue of Skeptic has a fascinating critique of criminal profiling entitled, "Criminal Profiling:     Granfalloons and Gobbledygook."  Unfortunately, it's available to subscribers only.  Fortunately, the article's author, psychologist Paul Taylor, has another article on criminal profiling available online:

Criminal profiling (CP) is the practice of using crime scene evidence to infer the personality, behavioural, and demographic characteristics of the offender who committed the crime.  The increased usage of CP by police investigators around the world over the past three decades suggests that police officers generally believe that CP works.  Given the lack of compelling scientific evidence that CP is a reliable, valid, or useful psychologically-based investigative technique, it seems prudent to ask officers about their perceptions of CP and its application.

The article is available here.

Science and the Law

For the first time in several years, I spent a week not reading anything law.  No cases, no work, no books.  Nothing.  If you think you see a lot wrong with the system when reading law, put the books down and read some science: Then you'll really realize how screwy the system is.

Let's take eyewitness identifications, for example.  Let's see how they're done, and see why a freshman wouldn't pass Chemistry 101 if he performed his science the way we perform eyewitness IDs.

In the typical police line-up, police put a suspect's photograph on a sheet with five other people with similar characteristics.  It's called a "six pack," since there are six photos on the page.  The police then tell an eyewitness to "Pick one."

Or you'll have the traditional "Law and Order" style line-up.  The suspect will stand in line with other people.  Again, the witness will be told to "Pick one."

Based on chance alone, there's a one-in-six chance the wrong person will be picked.  That aside, there is no replication of the results of the line-up.

One requirement of the scientific method is replication.  Simply put, someone else should be able to get the same result as you.  Here, in a separate line-up, the eyewitness should be able to pick out the suspect.

Yet we only have one line-up.  That's it.  There is no replication.

Now some might say that replication is pointless.  After all, if the eyewitness spotted the same people in another line-up, she'd pick the same person again.  That's actually not true.

The most-recent research on human memory shows that, well, people have really bad memories.  People often forget faces.  If the line-up were replicated one-month later, it's unlikely that the eyewitness would pick the same person based on her memory of the first line-up.

Why don't we require replication?  You'd think that prosecutors and police would welcome this.  After all, it's further evidence that the eyewitness is reliable.

Sadly, however, juries view eyewitness identifications as inherently reliable.  So, as the current system stands, prosecutors and police could only lose if we required replication.

Then again, if prosecutors and police really "search for the truth," and "seek justice," wouldn't they win if an eyewitness were exposed as unreliable?