New Tech for Composite Sketches

An article in this month’s issue of Forensic Magazine covers some interesting new software-based approaches to the development of facial composites being developed in the UK. Two research groups have been independently developing facial composite software that attempts to apply existing knowledge about how human memory of faces works.

According to Dr. Charlie Ford at the University of Stirling in Scotland, co-developer with Dr. Peter Hancock of EvoFIT:

Peter knew that the normal method used to make composites does not work very well. We are not good at describing and selecting individual facial features, but are very good at selecting whole faces which look like someone we’ve seen.

EvoFit is a software application that employs a genetic algorithm to generate a series of “generations” of facial composites, based on the theory that humans remember faces holistically rather than piecemeal, and from each “generation” of new faces, the witness selects the one that most resembles the witness’s memory of the perpetrator.

A typical EvoFIT session looks something like this:

From a randomly-generated selection of eighteen faces, a witness is asked to choose the six faces that most resemble the suspect. These six faces become the parents of eighteen offspring faces generated by combining the features of the parents. The witness then chooses another six from the offspring population to become parents, and so on. The “features” being selected and mutated are values of around fifty “principle components” that describe the structure of the face. The process can proceed indefinitely, but usually produces a likeness acceptable to the witness in about four generations, as long as the initial selection (or “search space”) was adequate. This likeness can be saved and embellished with haircuts and clothing using normal photo editing software.

Dr. Chris Solomon at the University of Kent in England has been developing a similar genetic algorithm-based application called EigenFIT, which is similar in concept but applies a different algorithm for generating each new crop of faces.

The results of both applications appear to be fairly impressive, compared to the often “bizarre and inaccurate” composites that can result from the traditional method of having a witness select individual facial features and then stitching them all together to create a single face that often bears little resemblance to the perpetrator.

Here’s one example from EigenFIT, involving a test run in which a witness was shown the top row of images, and the EigenFIT algorithm was able to generate the composites in the bottom row:

Given what we know about wrongful convictions resulting from bad composite images, this seems like reason for hope on that front, not to mention reason to insist that more reliable methods are used in police departments in the U.S.

Popular Science also ran a piece on EvoFIT last month, and noted that a large part of the problem with traditional approaches to composite sketches spawns from the inability of witnesses to provide detailed or accurate descriptions:

As often happens during a crime, a victim gets only a brief glance at the assailant. Later, when police ask him for a description of the perpetrator, he has trouble recalling details. But now, with new identification software developed by two researchers in Scotland, victims no longer have to worry about describing their assailants. A computer does it for them.

Dr. Frowd admits, however, that the genetic algorithm is also capable of producing serious errors.

One drawback of genetic algorithms generally, is that they can converge on the wrong solution if the initial population is poorly chosen. “This may happen,” Frowd admits, “but if it does, the system is rolled back and started again.”

The question then may become how to know when such an error has occurred, which ultimately takes us back to the reliability of the witness’s memory, which of course we know to be highly fallible.

NACDL’s Response to Ruling (Mostly) Denying Access to Illinois Lineup Data

Following yesterday’s ruling in Cook County, NACDL has offered its official response, which reads in part:

“The Court has dealt a blow to the public today. We will appeal the decision and continue to demand that the Chicago Police turn over all of the data and case files behind a report that contradicts vast amounts of research on lineup procedures,” said Scott Ehlers, state legislative director of NACDL, the plaintiff in the lawsuit. “We can’t conduct a thorough, scientific review of this information without access to all of the data used to reach the conclusions set forth in this report.”

The Roderick MacArthur Justice Center, of Northwestern University, had originally requested the data, but was met with silence.  From its director:

“The fact that this data will continue to remain a secret is shameful,” said Locke Bowman, legal director of the Roderick MacArthur Justice Center. “We will continue to do everything in our power to get access to this information.”

The court did order disclosure of a limited set of materials relating only to closed cases that were included in the study, but undermining any value to that ruling is the fact that the photos used in the arrays will be redacted, rendering it impossible to assess the reliability of the lineups themselves, notwithstanding the police procedures by which they were administered.

The results of the infamous Illinois/Mecklenburg “study” sent shockwaves through the community of social scientists who had studied the same questions for decades, primarily because the results were touted as showing that the status quo procedures — i.e., the procedures that led to the majority of the 54 documented wrongful convictions in Illinois alone — were actually superior, in direct contradiction to all other research on the topic.  Foul play was suspected, and since then the police department and its then-General Counsel Sheri Mecklenburg have refused to turn over the data and subject the study to actual peer review, as is required of actual scientific studies.

Leading eyewitness researcher Dr. Nancy Steblay also weighed in on the outcome:

“Legal evidence and public policy should be based on sound science. The Illinois Study has again failed to meet standards for scientific credibility,” said Nancy Steblay, expert of eyewitness identification procedures and professor at Augsburg College. “A Blue-Ribbon Panel of eminent scientists previously judged the study’s method to be fatally flawed and thereby incapable of answering its research question. Now the Chicago Police Department is unwilling to share the complete data necessary for evaluation of the study’s purported findings. It is time to set aside the Illinois study and to firmly reject the unsupported claims of its proponents.”

Locke Bowman, legal director of the Roderick MacArthur Justice Center, summed it up succinctly:

Wrongful convictions in Illinois have been far too commonplace.  If the data supports the report’s findings that traditional lineups work better than reform methods, the Chicago Police have nothing to hide. But if the data doesn’t support those findings, it’s time for the Chicago Police Department and departments around the state to change the way they handle eyewitness identifications.

The Cook County judge’s decision is here (PDF).

NACDL Denied Access to Chicago Sham Lineup Study Data

A state court in Illinois today just denied NACDL access to the data underlying the controversial Illinois/Mecklenburg Report on police identification procedures, which means the widely denounced results remain cloaked in secrecy.

NACDL has stated that they will appeal the decision. We’ll report further as we have more info.

UPDATE: More news coverage here. Of particular interest from the comments at the Chicago Tribune from an individual who claims first-hand experience with the way lineups are conducted:

Here is first hand knowledge of how this works.
About ten years ago while preforming my job, I notice two guys breaking into a house. It was doing a holiday season. Its a lot of detail so I will spare you all the descriptions.
About a month later I was called to go the courthouse and identify the robbers. One id at the scene, the other got away, but they did catch him later when the other guy squealed.
When I showed up at the courthouse, I was let upstairs to their lineup. Mind you it had been a month later when I got called to come to id the suspect. I couldremember curtains facts, like his height, what he wore and what he had on his head,I just had to see the person to be sure, once you see them your memory comes back. That being said, even though I could pick the guy out, here are some of the sehemes the cops used.
1 they bring in about 6 guys
2 five of them are laughing
3 the guilty one is always center with the brightest light on him
4 police suggest to you in certain ways to look closely at a suspect, even when you are sure.
5 when you pick out the one they suspect, expect a pat on the back.
That was my experience.

And then, with unintentional irony, from a user tagging him/herself “the truth,” this response:

“4 police suggest to you in certain ways to look closely at the suspect in the middle, even when you are sure.
5 when you have picked out the one they suspect, expect a pat on the back.
That was my experience. “

So? The right guy went to jail. Sounds good to me.

Unfortunately, that’s exactly the problem with wrongful convictions, in which Illinois is second only to Texas: it’s always the police suspect who gets picked, but it’s not always the police suspect who committed the crime.  Unfortunately this “more convictions = more justice” mentality is all too common, and yet this is exactly the scenario that has sent the wrong people to prison in Illinois and across the country since the advent of the police lineup.

This is also exactly what reformists, like those seeking the data from this field study, are seeking to change — by implementing procedures that are specifically engineered to test the independent memory of witnesses, rather than sham procedures that merely rubber-stamp the suspicion of the police, who are in many cases simply wrong.

“Place your finger here” (Or: Police Lineups in Louisiana)

Naturally law enforcement officers are insisting that this wasn’t a “real” photo array, but WWLTV in Louisiana is reporting that a photo array was uncovered in a Louisiana police department which included, above the suspect’s photo, the words:

“Place your finger here”

and

“He did it”

The photo was discovered on a bulletin board inside the police department by a reporter examining the scene, a police department still locked down after being ravaged by Katrina two and a half years ago.

Metro Crime Commission officials Rafael Goyeneche and Tony Radosti, upon investigation, “raised serious concerns about a photograph they found.”

It’s a picture of six men in an apparent police lineup with one man circled and the phrases “place your finger here” and “he did it” written above.

“To have something like this in any police department, I think it’s embarrassing and potentially a civil rights violation,” said Goyeneche.

The picture is on a bulletin board in the damaged police station. There is no way to determine how long it has been there, who put it there or who wrote on it, but Goyeneche said the public is owed an explanation.

“This is an indication of a grander problem,” he said. “(It’s) a problem that requires an immediate and complete investigation.”

No one seems to know the context in which the photo emerged (it has also been described as a “sick joke”), and certainly no one has admitted any wrongdoing, but its mere existence is reason for concern, at the very least, about the tone of law enforcement with respect the ongoing, systemic problem of wrongful convictions resulting from suggestive police lineup procedures.

Another Mis-ID Wrongful Conviction Revealed in Illinois

Dean Cage was released from prison in Chicago last night, after serving nearly 14 years in prison for a crime he didn’t commit. Mr. Cage’s exoneration marks the 29th wrongful conviction revealed in the state of Illinois, since the Innocence Project began investigating cases in which DNA evidence was preserved and held the potential to prove innocence.

Like 20 of the other wrongful convictions uncovered so far in Illinois, Mr. Cage was wrongfully convicted on the basis of erroneous eyewitness evidence. After a composite sketch was circulated around the neighborhood, an anonymous individual reported that Mr. Cage was a possible suspect, and reported his place of employment.

In a blatantly unnecessary and excessively suggestive identification procedure, the police then took the victim to the meat-packing plant where Mr. Cage worked, to see if she could identify her attacker. Naturally, she picked out the man who presumably looked most like her attacker. But as it turned out, resembling the attacker was Mr. Cage’s only link to that crime.

This is one more case study in favor of Illinois evolving its police procedures into the 21st century, rather than fighting tooth and nail to preserve its antiquated — and demonstrably flawed — practices.

Flaws in the ABA’s Jury Instruction on Cross-Race

At a groundbreaking two-day eye-ID litigation conference in NYC in March co-sponsored by the Eyewitness ID Reform Litigation Network, we addressed the issue of cross-race instructions, including the ABA version referred to this week at the Kansas Defenders blog. In short, while it is certainly good news that the ABA is paying attention to this issue, since the average instructions on eye-ID are woefully inadequate, the ABA instruction is far from ideal for the following reasons.

First, the initial language is conditional. “You ‘may’ consider, if you think it is appropriate …,” instead of you “should” consider.

Second, there is no mention of the numerous scientific studies that have shown, empirically, that cross-racial bias exists. In cases where experts have not testified at trial on the subject (i.e., most cases), jurors are left ignorant of the copious social science research on the topic. Instead of stating that “scientific studies have shown,” the court cites the amorphous “ordinary human experience.”

Third, the instruction suggests erroneously that sufficient contacts with another race may well overcome cross-race bias. Again, no science is cited. Moreover, it appears that contacts alone do not offset cross-race bias; rather, it is exposure combined with differentiating tasks (i.e., merely living in a neighborhood versus teaching a class of 30 in which one must distinguish among members) that might mitigate the effect, but even then the mitigation appears to be minimal.

Fourth, there is no discussion of prejudice (i.e., that cross-race bias exists in people seemingly without prejudice).

A better instruction (though not a perfect one), and one which I would encourage lawyers to propose instead of the ABA version, can be found in Johnson, Cross-Racial Identification Errors in Criminal Cases, 69 Cornell L. Rev. 934 (1984):

“In this case the identifying witness is of a different race than the defendant. In the experience of many it is more difficult to identify members of a different race than members of one’s own. Psychological studies support this impression. In addition, laboratory studies reveal that even people with no prejudice against other races and substantial contact with persons of other races still experience difficulty in accurately identifying members of a different race. Quite often people do not recognize this difficulty in themselves. You should consider these facts in evaluating the witness’s testimony, but you must also consider whether there are other factors present in this case that overcome any such difficulty of identification.”

The Network has sample motions for jury instructions on both estimator and system variables. Lawyers should be encouraged to tailor these to the facts of their cases and submit.

Attorneys should also look at the Long and Brink cases in Utah, and contact the Network about Utah’s sample eye-ID instructions on various factors from which, particularly when combined, is very comprehensive (though the cross-race portions can be made more robust).

Zeke Edwards
The Innocence Project

Lineup “Wild Cards” as Precaution Against Mis-ID

We reported on a New Zealand study last year that seeks to test the effectiveness of including a “wild card” in a police lineup, as an additional protection to cue to the witness that “none of the above” is a perfectly valid choice. Personally, I think this is a fantastic idea — imagine the difference between a multiple choice question where “none of the above” is an option, vs. the exact same question without that option; as a test-taker, you’re far more likely to guess in the latter scenario. In the realm of police lineups, guesses often have grave consequences.

In an update on this study, researcher Dr. Rachel Zajac of Otago University notes that “The biggest misconception about lineups is that the perpetrator will be there.” Her early research suggests that providing a visual “none of the above” option, or “wild card,” in each lineup, increases the likelihood that a witness will not select a lineup member when he or she is uncertain, rather than guessing. In other words, adding the non-pick to the set of affirmative options better communicates the fact that not selecting anyone from the lineup is an acceptable choice, and may further reduce the likelihood of misidentification in a way that cautionary instructions alone do not.

On the prospects of her recommendations being adopted by law enforcement, Dr. Zajac reports: “In my experience, the police are very receptive to evidence-based ideas that would facilitate the accuracy of the evidence they solicit from witnesses.”

Another Flawed Lineup, Another Exoneration in Dallas County, and a Glimmer of Hope

Thomas Clifford McGowan Jr. has spent the last 23 years in prison because he was selected from a fundamentally flawed photo lineup, after his photo was collected from a traffic offense two days after a rape/burglary.

From the Dallas Morning News:

The 49-year-old would become the 16th Dallas County inmate to be cleared through DNA testing since 2001, the highest total for any county in the country. Like almost all of the other discredited convictions, Mr. McGowan’s was based primarily on the victim selecting his photograph from a police lineup.

The photo array from which the victim selected his photograph was a sloppy collection of black and white photos, color photos, and photocopies of photos. Not surprisingly, Mr. McGowan’s photo was a color original. Further, after the witness tentatively pointed to Mr. McGowan, the investigating officer insisted that the witness make a positive ID, rather than allowing her to describe her level of certainty in her own words.

Barry Scheck and the Innocence Project have worked for the year to free Mr. McGowan.

By not allowing the victim to describe her certainty in her own words, Mr. Scheck contended, Detective Corley implied Mr. McGowan was the attacker.

“I have no doubt that the officer wasn’t trying to do anything wrong here,” Mr. Scheck said. “It’s just a terrible, bad practice.”

Even the local prosecutors are beginning to agree that there is a systemic problem:

Dallas County prosecutor Mike Ware, who oversees the conviction integrity unit, said Tuesday that problems with photo lineups are “a common thread that runs through almost all of the wrongful conviction we’ve run across.”

The Dallas Morning News also reports that Mr. Ware is now publicly embracing the importance of mandatory double-blind lineup procedures for photo arrays, which is an important and significant development, at least to my knowledge. As we reported previously, Dallas County has been laying the groundwork for a field study on police lineups in a purported attempt to develop more reliable procedures, and until now there appeared to be reason for concern that the study would replicate the fundamentally flawed methodology of the now-infamous Illinois/Mecklenburg study that turned out to be little more than a rubber-stamping of the flawed status quo procedures that continue to cause innocent people to be put in prison.

Now, Mr. Ware is on record stressing the importance of double-blind procedures, which all social scientists (and reformers) have long agreed must be a component of any standard lineup procedure that hopes to reduce wrongful convictions.  Presumably this means that the Dallas study won’t repeat the futile endeavor of “testing” blind procedures against non-blind procedures, as was done in Illinois under the guidance of Sheri Mecklenburg, with the number of suspect picks as the measure of “reliability.”

This is big news on the eyewitness ID reform front.  We’ll try to stay on top of this as more information on the Dallas field study begins to surface.

“Clothing Bias”: More than Common Sense

Chief Judge Easterbrook in the 7th Circuit ruled on a case yesterday in which the central issue on appeal was whether an ID was properly admitted over a motion to suppress, where all the members of the lineup were wearing identical prison clothes with navy blue slippers, with the exception of the suspect, who wore white tennis shoes — where eyewitnesses had described one suspect as wearing white tennis shoes. (cite: United States v. Williams, — F.3d —-, 2008 WL 1701843 (7th Cir. 2008)).

At oral argument, both parties argued that “common sense” supported their respective positions — for the defense, common sense tells us that putting the suspect in white tennis shoes where everyone else has exactly identical clothing makes the suspect stand out. For the government, apparently it has the opposite effect, or no effect.

But the court wanted more than common sense arguments, and it specifically requested empirical data in support of the arguments. Unfortunately, none was provided, and the court was forced to rule on the basis of battling “lawyers’ talk” alone — despite the fact that empirical data do exist on the question of clothing bias as it relates to identification procedures.

Drs. Jennifer Dysart and Rod Lindsay published a chapter in the Handbook of Eyewitness Psychology (vol. 2, 2007) that focuses on clothing bias, and reports that, in target-absent lineups, “clothing bias does lead to a significant increase in false identifications of innocent suspects.” (p. 143, citing Lindsay et al., Do the Clothes Make the Man? An Exploration of the Effect of Lineup Attire on Eyewitness Identification Accuracy, 19 Canadian J. Behavioural Sci. 463)). The same chapter cites to a number of other studies on the same topic.

Dysart et al. published another study in 2006 that revealed the significant effect of clothing bias in show-ups (Dysart et al., Show-ups: The Critical Issue of Clothing Bias, 20 Applied Cognitive Psychol. 1009 (2006)), and Freire et al. published a study showing a serious problem with clothing bias among children subjected to a lineup condition (Freire et al., Lineup Identification by Children:
Effects of Clothing Bias, 28(3) L. & Hum. Behav. 339 (2004)).

While there may be room for more research on this issue, to say that no empirical data exist is an unfortunate oversight.

The redeeming part of the decision, unrelated to lack of empirical data entered into the record on clothing bias, is Easterbrook’s observations on common sense as it relates to eyewitness identifications:

If there is one thing known about eyewitness identification, it is that ‘common sense’ misleads more often than it helps…The problem with ‘common sense’ is that experience tells us what leads to confidence about whether we have seen a given person before but does not provide reliable ways to test whether that confidence is justified. People confuse certitude with accuracy and so are led astray. Psychologists have established that certitude often is unwarranted. It takes data rather than intuition to answer questions such as “can non-uniform footgear in a lineup lead to misidentification?

In support of his observations on the failures of common sense as a guide to understanding the factors affecting eyewitness identifications, Easterbrook cites to Gio Shay and Tim O’Toole’s great article on the shortcomings of Manson v. Brathwaite (PDF), which systematically dismantles the Supreme Court’s obsolete framework governing the admissibility of eyewitness evidence.

Thanks to Gio Shay for the heads up on this decision.

Justice Denied Again for Troy Davis

The AP is reporting that the Georgia Supreme Court has denied Troy Davis’s latest effort to obtain a new trial.  Mr. Davis’s attorneys filed a new motion, following the court’s March 17 denial of their earlier extraordinary motion for a new trial, this time moving for reconsideration of that earlier ruling.

A substitute opinion was also issued, with the same 4-3 split and the dissenting opinion remaining the same.  New opinion available here (PDF).

Thanks to Deirdre O’Connor at Emory Law School for the tip, and for tirelessly seeking a remedy for Mr. Davis.