Search

Previous research on visual comparison tasks has shown that low non-match base-rates impair non-match detection and result in more a decision-making bias (C) in face-matching and fingerprint-matching tasks (known as the “low prevalence” effect, or LPE). This study is a follow-up to one which found that completing feature ratings (e.g., rating the similarity of different fingerprint regions between two prints) reduced the LPE in fingerprint comparisons, compared to completing no ratings. This study will investigate this same question with an improved control group that will complete image quality ratings rather than no ratings). Further, no research has investigated the LPE in fingerprint comparison in non-novice samples. This project will also recruit a group of forensic science trainees that will act as a comparison group to novices to investigate whether feature-comparison may also be an effective training tool to reduce bias in individuals with forensic training.

Previous research has shown that low non-match base-rates impairs non-match detection and results in more a conservative decision-making bias (C) in face-matching and fingerprint-matching tasks (known as the “low prevalence” effect). To address this, studies of visual search have shown that strategies which cause observers to slow down and minimize rote responding diminishes the low prevalence effect (Fleck & Mitroff, 2007). T his study will explore another potential method of preventing the low prevalence effect through making feature-feature ratings. Rating the feature of features when performing pairwise comparisons has been shown to increase accuracy on face-matching tasks, particularly on non-match trials, through careful consideration of the similar and dissimilar features in each image. We aim to explore the impact of ratings (no ratings, feature ratings, or an image quality ‘control’ condition) and prevalence (low: 90/10 match/nonmatch; or equal: 50/50 match/nonmatch) on fingerprint matching performance.

This study will test for a prevalence effect (cf. Wolfe et al., 2005) on fingerprint identification judgments by exploring the impact of match/non-match base-rates on fingerprint identification accuracy. Previous research has shown that when a target stimulus appears infrequently over many visual search trials (e.g., 10% of the time), people exhibit more “misses” (i.e., they fail to detect the target when it is present) and more conservative decision-making (C) in face-matching and other perceptual tasks. Conversely, when the target stimulus appears frequently (e.g., 90% of the time), people exhibit more “false alarms” (i.e., believing that the target is present when it is not) and more liberal decision-making (C). We aim to explore the impact of base-rates (10/90 match/nonmatch; or 50/50 match/nonmatch; or 90/10 match/nonmatch) and trial-by-trial feedback (present or absent) on fingerprint matching performance.

Testing whether knowledge of an expert's performance on blind proficiency testing affects trust in the expert witness, the evidence (fingerprint or bitemark), and verdicts.

The low prevalence effect is the phenomenon whereby low frequency targets are more likely to be missed in visual search and visual comparison tasks, such as screening luggage to detect weapons or comparing faces to detect fake IDs. Growns and Kukucka (2020) recently replicated the low prevalence effect in fingerprint identification, such that participants were more likely to “miss” matching fingerprint pairs when such pairs were rare, and were more likely to misidentify non-matching pairs as matches when such mismatches were rare. This is an important topic for study because these types of errors could lead forensic examiners to misidentify a guilty person as innocent or an innocent person as guilty, respectively. Prior research has attempted various ways of eliminating the low prevalence effect. Of relevance to the current study, Miyazaki (2015) found that monitoring participants’ performance in real time reduced the low prevalence effect in a computer-based visual search task. The current study will test whether participants who believe that their performance is being monitored show an attenuated (or no) low prevalence effect in judgments of fingerprint pairs. As a secondary aim, we will test whether our monitoring manipulation has a stronger effect in individuals who are dispositionally apprehensive about having their performance evaluated.

Knowledge of task-irrelevant information (i.e., a suspect's confession) undermines the probative value of forensic evidence (i.e., forensic confirmation bias). Cross-examination may sensitize jurors to bias—but do attorneys recognize when bias has tainted evidence against their client and adjust their cross-examination strategy accordingly? To address this question, 130 defense attorneys imagined representing a man charged with manslaughter and reviewed a case file that included, among other things, an autopsy report from a medical examiner who was either aware or unaware of their client’s recanted confession before ruling the death a homicide. When the examiner knew of the confession, attorneys rated the autopsy as no less probative or reliable, they were no less confident in their client’s guilt, and only 46% raised the possibility of confirmation bias on cross-examination. Our findings suggest that defense attorneys underappreciate the impact of forensic confirmation bias, such that biased forensic testimony would be better avoided via procedural reform.