DNA testing in recent years has resulted in many wrongly-accused citizens being exonerated of crimes they didn’t commit. For example, investigations conducted by the Innocence Project, an organization started in New York City, led to the exoneration of more than 220 citizens convicted of serious felonies. An analysis of the first 130 exonerated revealed that mistaken eyewitness identification was a contributing factor in 101 (77 percent) of those cases. It is believed that many cases of mistaken identity are caused by police inadvertently influencing eyewitnesses as they look through photo-arrays, searching for the perpetrators. Enter ‘Officer Garcia’ – an impartial virtual detective who guides witnesses through the identification process without the risk of imparting any influence.
Prosecutors agree that a good identification test is one that allows the investigator to conclude that the witness’ identification of the suspect is the product of his/her memory, and not the result of inadvertent influence by the investigator who conducted the procedure. This is called a ‘blind photo-array’, where the officer guiding the witness doesn’t know if the offender is in the photo-array or not.
The North Carolina legislature recognized this idea and incorporated it into its 2007 Eyewitness Identification Reform Act. The Act specifies the use of blind identification procedures, but allows the use of computer programs in lieu of blind procedures.
Meet avatar 'Officer Garcia'
Brent Daugherty at the University of North Carolina in Charlotte, together with computer science colleagues, has developed a new technology that enables citizens to interact with computers through voice commands and without the aid of a mouse and keyboard. Daugherty developed Officer Garcia, a virtual human program designed to conduct photo-arrays in accordance with established guidelines.
Officer Garcia has been modeled on a plain-clothes detective, who converses by identifying keywords from the witness’s speech and responds appropriately in a speech-synthesized voice. The virtual officer framework combines speech, graphics, discourse modeling, and planning to guide the witness through the identification procedure.
The first test
Researchers recruited 259 undergraduate students from a state university who were each shown a mock wallet theft and then asked to identify the suspect by studying a series of photographs. First, they were led through the identification process by a student with no knowledge of the suspect, then by the avatar. The results were very similar, proving to the team that Officer Garcia performed as well as an impartial human.
Daugherty plans to test Officer Garcia in field trials.
Watch the video below to see how Officer Garcia conducts the procedure.
Via New Scientist