Investigating New Ways to Improve Eyewitness Identifications Using Receiver Operating Characteristic Analysis

There is a two-pronged, very real societal problem concerning identifications made by eyewitnesses: innocent suspects are mistakenly identified and charged with a crime they did not commit, or guilty suspects are not identified and free to commit more crimes. Decreasing the chances that innocent suspects are misidentified, unfortunately also decreases the chances that guilty suspects are identified; and likewise, increasing the chances that guilty suspects are identified also increases the chances that innocent suspects are identified. In other words, an eyewitness's accuracy is not just about choosing the right suspect; it is also about not misidentifying the wrong suspect. This teeter tottering is what eyewitness memory researchers have been grappling with for decades, and we aim to combat it.

That may seem like a bold claim, but it is now possible because of the new method of analysis we recently introduced to the field of eyewitness memory. The method disentangles the concept of response bias (i.e., the inclination of an eyewitness to choose someone or not choose someone from an identification parade) from discriminability (i.e., the ability to discriminate an innocent suspect from a guilty suspect). While the method is new to eyewitness memory researchers, the method is tried-and-true for medical diagnosticians who routinely use it to test whether one diagnostic procedure is better able to discriminate a disease state from a non-disease state, for example. Eyewitness memory researchers are faced with the same conceptual issue as medical diagnosticians in which they need to determine whether one procedure is better able to discriminate guilty from innocent suspects. Our new method allows us to answer important questions in eyewitness memory that could not be previously answered because there were no satisfactory ways to measure discriminability.

We will use our new technique in a series of experiments designed to investigate ways to increase discriminability. We will do this by comparing commonly used identification parade procedures and manipulating components of each procedure to investigate the factors that increase discriminability. The results of the experiments will shed light on why certain identification parade procedures (e.g., sequentially presented videos vs. simultaneously presented videos) and components of those procedures (e.g., identifications made quickly, number of individuals presented in an identification parade) increase discriminability. The results of this research could ultimately lead to more guilty people and fewer innocent people being identified and later prosecuted. Thus, the proposed research has the potential to make significant societal impact.

The appeal of this work is wide ranging both in terms of applying a new methodology in the field of eyewitness research as well as advancing research based practice when it comes to identification parades.

By identifying factors that increase the accuracy of identifications, we expect the following individuals and groups to benefit:

Professionals in the Criminal Justice System, including:
- Police Investigators
- Solicitors
- Policymakers
Advocacy Groups, including:
- Victim Support Advocates
- The Innocence Network
Members of Society, including:
- Victims
- Innocent suspects
- Eyewitnesses

Communication with user groups is critically important for this type of research because if researchers and practitioners do not communicate, our respective efforts will fall on deaf ears. We have begun to and will continue to liaise with Metropolitan (MET) Police Inspector Burnell. Inspector Burnell is in charge of the MET Identification Suites. He is a critical contact person for sharing information, as well as providing the resources necessary to carry out the proposed experiments. We will continue to work closely together to maximise impact (please see his accompanying letter of support).

We will build a network of experts from various professions (i.e., stakeholders from the list above) to disseminate the results and receive feedback. As we disseminate the results and attend conferences and research fairs, the network will grow. These individuals report to the Association of Chief Police Officers and the Crown Prosecution Services committees, so this will be a step towards expanding the network to include more officers and solicitors. If the findings from the experiments yield results that could improve PACE Code guidelines, then a next logical step would be to lobby the Home Office to make procedural changes.

We will also liaise with members of victims advocacy groups of two varieties: one for victims of crimes and the other for victims of the criminal justice system (i.e., innocent suspects who are wrongly convicted). There is a legacy of some researchers who only liaise with groups who support innocent suspects and their work comes across as one-sided. This is unfortunate because there are trade-offs associated with eyewitness identifications (e.g., fewer innocent suspects identified lead to fewer guilty suspects identified; Clark, 2012).

A workshop is planned for the end of the project to disseminate the findings of the research to those who aim to make the criminal justice system more efficient, including identification officers, solicitors, victim rights advocates, and policymakers. The workshop will allow for us to receive valuable feedback. We will also create and maintain an online forum for the network to regularly communicate.

To inform the general public, we will promote the work through contact with the media and participation in public science events. We will also build and maintain a web presence dedicated to providing updates of the research for the user groups and general public (e.g., our laboratory websites and social media accounts). We hope this will invite researchers, students, and the general public to be informed by and comment on our research.