Comparing eyewitness identification procedures with information-theory measures

Eyewitness identifications play a key role in many criminal investigations. Investigators have a wide range of options for how they conduct an identification attempt, and eyewitness researchers have explored many of the relevant variables, such as whether a suspect is shown to a witness individually (a “showup”) or together with a number of fillers (a “lineup”). Unfortunately, different measures of lineup effectiveness often support different research conclusions and policy recommendations. We show that existing measures are incomplete in the sense that they do not use all of the information from the reference population defining witness performance. We introduce a complete measure, Expected Information Gain (EIG), by applying information-theory principles to identification data. EIG identifies the procedure that produces the most information about suspect guilt or innocence across all of the possible witness responses. Thus, EIG is a useful measure for policy-focused research.