Multiple confidence estimates as indices of eyewitness memory

Area of Science:

• Cognitive Psychology
• Forensic Psychology
• Human Memory

Background:

• Eyewitness identification is prone to errors, leading to wrongful convictions.
• Current identification procedures rely on binary decisions, which may not fully capture memory certainty.

Purpose of the Study:

• To investigate an alternative eyewitness identification procedure using confidence estimates.
• To assess if classification algorithms applied to confidence data can improve accuracy.

Main Methods:

• Participants' confidence ratings in face recognition tasks were analyzed using classification algorithms.
• Experiments involved standard old-new tasks, face mini-lineups, and standard lineup paradigms.
• Accuracy was compared between a confidence-based classification group and a binary decision control group.

Main Results:

• The confidence-based method showed superior accuracy in identifying target-absent trials (innocent suspects).
• While target-present accuracy varied, sophisticated algorithms improved performance.
• This approach demonstrated better discrimination between correct and incorrect identifications.

Conclusions:

• Confidence estimates provide accessible evidence of recognition memory strength.
• Classification of confidence data offers a more sensitive index of memory quality than binary decisions.
• This method has potential to reduce false identifications in eyewitness testimony.