How do you double-blind the part where the lineup gets put together in the first place? A lot of the bias examples you showed earlier had to do with making the suspect stick out, and somebody’s got to start with a photo of the suspect alone if they’re going to find a bunch of similar-looking people. Are computers good enough at face recognition to help with this yet?
Yes, with today’s advances in facial recognition software, it would be easy enough to have a computer sort an array of similar faces to use as distractors. http://www.popularmechanics.com/technology/how-to/g1731/8-weird-ways-people-are-using-facial-recognition-software/
One (possibly too time consuming way) would be to have one person get the picture of the original suspect and have to pick a matching photo. Then it gets passed on the a second person who doesn’t know which of these faces is the suspect who picks a photo to add to the lineup. Those three faces are passed on to another and so on. There’s still potential for bias and collusion, but hopefully less so.
Or you could have one person take the photograph of the suspect and add two similar photographs then give all three photographs to someone who doesn’t know which one is the suspect and have that person add three more photographs. The first person will still unconsciously bias the first set of three photographs in favor of the suspect. However, the second person won’t know which photograph to compare the photographs that they add to. So the second person will cancel out the unconscious biases of the first person.
I think it’s central to the idea of double-blindness to not let what you think who the suspect is or looks like influence the selection in the lineup. So I would say that picking a lineup that looks like the suspect is a bad idea.
For example, maybe the eyewitness actually can’t be sure if it was a tall person or short person. If your suspect is a tall person, and your lineup is all deliberately tall people, then you are subtly hinting to the witness that the suspect is a tall person. Pretty sure you do not want that.
My only concern is that purely random selection might choose weird lineups, especially with small sample sizes. One way around this is to randomly pick more people, insert your suspect(s) in to the group. Then shuffle the order in which people appear, and then show 10(?) people at a time to the witness. That way the suspect only appears in one of the lineups and the witness is shown many people with different features. If the witness can still pick out the suspect from that procedure, then you can be reasonably confident about how certain the witness is.
As I understand it, they usually only fill the lineup with tall people if the witness (or another witness) has previously said the person they’re looking for is tall. If they told you the perp was tall and blonde, you found a tall and blonde suspect, then you went put the suspect in a line of people who aren’t tall or blonde, you would increase the chances that they’d identify that suspect as the one, regardless of whether or not it was them.
The whole purpose of the lineup is to avoid that, so instead, you give them numerous faces that match the description they gave, only one of which has any chance of being the person you’re looking for. If they can pick that person out, it proves they actually remember the face of the perpetrator. If they can’t (either choosing wrong or not recognizing anyone), then either the suspect is innocent or the witness can’t remember well enough to be relied upon.
Ideally, then, if I’ve understood right, you don’t try to make them all look like your suspect, per se. Instead, you pick a bunch of other people that match the description the witness gave (the suspect probably matches it too), so the others will look similar, but you avoid biasing the witness toward a certain trait.