Simulated Viewing Distance Impairs The Confdence–accuracy Relationship For Long, But Not Moderate Distances: Support For A Model Incorporating The Role Of Feature Ambiguity Part 1

Abstract

There is an increasing need in eyewitness identification research to identify factors that not only influence identification accuracy but may also impact the confidence–accuracy (CA) relationship. One such variable that has a notable impact on memory for faces is viewing distance, with faces encoded from a shorter distance remembered better than faces encoded at longer differences. 

With the continuous development of modern society, the number of crime witnesses has gradually increased. In this case, witnesses need to be able to effectively identify and remember key information such as the suspect's appearance and movements. There is a relationship between eyewitness recognition and memory.

First, good memory is the key for witnesses to recognize suspects. If witnesses can clearly remember the suspect's appearance, body characteristics, movement habits, and other information, the success rate of solving the case will be greatly improved. It can be seen that good memory is very important for witnesses to identify suspects.

Secondly, the psychological state of witnesses will also affect their recognition and memory abilities. For example, if the witness is scared or nervous, attention will be distracted and memory will be affected. In contrast, if a witness remains calm and focused, he or she will be better able to remember information about the suspect. Therefore, witnesses need to overcome negative emotions and remain calm to better perceive and observe.

In addition, the life experience and cognitive abilities of witnesses also play an important role in recognition and memory. By regularly observing people and things around you, learning psychological knowledge, etc., you can improve your observation and understanding abilities. At the same time, actively exercising the brain and maintaining good living habits can also help improve memory and cognitive abilities.

In short, the identification and memory abilities of eyewitnesses play a crucial role in solving crimes. Only by ensuring a good mental state, actively exercising, and improving cognitive abilities can we better play the role of witnesses and help the police maintain social peace and stability. It can be seen that we need to improve our memory. Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material with many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the many active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in a variety of ways.

Click know 10 ways to improve memory

In four pre-registered experiments, using both laboratory and online samples, we compared faces viewed at a simulated viewing distance at different levels (medium and far) to faces that were viewed at a very near-simulated distance Distance was simulated using a Gaussian blur function with higher levels of blur corresponds ending to greater simulated distance. We found that both medium and far simulated distances impaired memory performance overall relative to no simulated distance, with increased distance resulting in poorer memory. 

However, far-simulated distances impaired the CA relationship. In a fourth experiment, we found that a pre-test warning did not ameliorate this impairment of the CA relationship for faces viewed at far-simulated distance. The findings suggest that high-confidencenideidentificationss made for faces viewed from long distances should be disregarded and that estimator variables that impact memory may degrade the CA relationship when memory is reduced to a critical level significance statement

When an eyewitness views a crime unfold, several variables infuence the likelihood that they will be able to later make an accuridentificationtion. Past research claimed that the relationship between confidence and accuracy in the identification was tenuous at best, but a more recent analysis strategy has changed that perception. Confidence may be a good indicator of accuracy so long as the circumstances of the encoding event are ideal. What is less well-known is whether the relationship between confidence and accuracy may be preserved even in non-ideal circumstances. 

Although it is well established that eyewitnesses’ memooffor faces are degraded with increasing distance, whether this factor impacts the relationship between confidence and accuracy is less well-understood. Some studies have found that the relationship between confdence and accuracy may be conserved at some distances but not others. Researchers have suggested that this is driven by eyewitnesses’ failure to appreciate how difficult the task of identifying a face is as viewing distance increases. 

In four experiments, we employed a within-subjects design which eliminates the possibility that individuals can change their strategy to become more or less conservative but does them into the fact that recognition will be harder for some faces than others. We found evidence that only long distances impaired the confidence–accuracy relationship. Moderate distances did not, suggesting that feature ambiguity at encoding leads to the affected effect rather than overconfidence confidence in recognition performance.

Due to a relatively recent change in the way that eyewitness memory researchers quantitatively characterize the relationship between and accuracy, there has been a revived interest in identifying factors that do and do not influence the ability of eyewitnesconfidencence in identification to predict their likelihood of accuracy (Wixted & Wells, 2017). As such, we are interested in the way that certain estimator variables, that is, affect identification accuracy but are not under the control of the criminal justice system (Wellsinfluence influence this relationship. 

Whereas a great deal of research in the field of psychology and law focuses rightly on system variables or those factors that the re under the control of the criminal justice system, estimator variables are just as important to understand (S Miler et al., 2018). That is because we can generate evidence-based predictions of how likely a particular witness is to successfully identify a perpetrator biospecific a specific eyewitness context if we understand the impact of that context on memory performance. For the present studies, we focused on one estimator variable, viewing distance (the distance between the observer and the target face at encoding), to better understand the influences of the relationship between witconfidencedence and accuracy.

It should come as no surprise to a layperson that as the distance between an eyewitness and the face they need to later identify the accuracy decreases. The science is clear on this fact (e.g., Lindsay et al., 2008; Nyman et al., 2019; Lockamyier et al., 2020), and funding has been recognized by official policymakers. For example, Lindsay et al. approached students on a college campus and asked them to encode the face of a research assistant who appeared between 4 and 15 m away or 20–50 m away. 

Participants then made a lineup decision from a target-absent or target-present lineup immediately or 24 hours a 24-h delaConfirming the intuitive prediction, correct targidentificationion was poorer at the longer distances. In a similar vein, Nyman et al. had participants encode in a more controlled (but still naturalistic) setting different different distances ranging from 5 to 110 m. Participants similarlyidentificationtifcation from a lineup, and the same pattern of results emerged.

However, the real-world nature of the testing conditions used in both Lindsay et al. (2008) and Nyman et al. (2019) introduces higher variability than what would be ideal by laboratory standards. 

Teseofferdies offer an excellent illustratiotrade-offstrade-ofs that researchers must make between simulating natural conditions closely and exerting rigorous experimental control. While there is no right answer as to which approach is best, other researchers have attempted to simulate viewing distance in the laboratory using digital facsimiles of faces (rather than real, live, faces) corresponding to some physical distance in the natural world. Fortunately, distance can be simulated in several empirically established manners first is to record the witnessed event on video and simply vary the distance between the target and the camera (Lockamyier et al., 2020). Researchers may alternatively opt to present participants with photographs of faces but reduce their size to approximate distances distance (e.g., Loftus & Harley, 2005). 

However, both approaches translate poorly to remote data collection given the lack of standardization concerning the display size that remote participants opt to use. To address this issue, we chose to blur faces to a greater or lesser extent translating to greater or lesser simulated distance. Face blurring can be accomplished in multiple ways (e.g., Gaussian blur, Lampinen et al., 2015; filters filters, Loftus & Harley, 2005). Similar to the logic behind showing images of reduced size, blurring the image of a face simulates the fact that distant faces are represented by fewer photoreceptors in the retina, meaning that each feature of the face must be encoded by fewer cells, resulting in something similar to a pixelated image.

Loftus and Harley (2005) employed such an approach and found that using a low-filterer perfectly simulated in vivo changes in viewing distance (comparable to reducing photograph size). Following this study, Lampinen et al. (2015) used Gaussian blur (which produces a similar low-pass visueffectect) and found evidence consistent with more naturalistic research procedures—a negative monotonic trend for accuracy as simulated distance Therefore, the use of a blur function to simulate distance apeffectiveective and allows researchers to maintain tight laboratory cortisol. Tis method has an advantage over size manipulations because participants can adjust the visual angle of their screen to ensure that they encode the blurred face as intended, but this does not ameliorate the intendeeffectr effect.

Such a blurring procedure allowed us to investigate how (if at all) distance impacts the relationship between confidence and accuracy in eyewitness memory  (Lampinen et al., 2014). Eyewimisidentificationcation is the single largest contributor to cases that have been overturned in the last several decades by the Innocence Project (Innocence Project, 2019). According to researchers need to identify the circumstances under whimisidentificationsons is more liConfidencefdence judgments commonly accompany wiidentificationsations and may be prompted (“How sure are you that this was the guy?”) or unprompted (“That was for sure the guy I saw!”). 

Given the frequency with which such information is tied to identificationsons, it behooves the criminal justice system to determine their usefulness for triers of fact (Smalarz et al., 2021; Wells et  al, 1981, 2002). Not even thirty years ago, eyewitness memory researchers were generally in agreement that eyewitness confidence was, at best, weakly related to eyewitness accuracy (Sporer et al., 1995). That sentiment, however, has since shifted; Wixted and Wells (2017) recently made the case in favor of usiconfidencence as a predictor of accuracy in legal contexts, arguing that previous suppositions of thconfidencence and accuracy were unrelated were based on the incorrect analytical approach (i.e., the point-biserial correlation, see Juslin et al., 1996 for a discussion).

Teir proposal did come with a caveat: confidence should only be used as a predictor of accuracy as long as the conditions at the time of encodinidentificationcation of the face were favorable to a posidentificationcation (e.g., estimator variables at the time of encoding are associated with better memory performance system-level level factors knoinfluencefuence eyewitness accuracy or confidence are absent). However, some have further argued that these favorable conditions may be insufficient, but not necessary, condition confidence to be a valid and reliable predictor of accuracy (Mickes et al., 2017). 

In essence, there may be several suboptimal conditions that impair eyewitness memory overall but leave the tconfidencence–accuracy relationship intact (see Wixted et al., 2016, for a fielded study confirming this precision). Tis bodes well for confidence as probative evidence in legal contexts, as it prefilters a filtering mechanism whereby the test low-confidence confidence witnesses can be prevented from proceeding to trial (because confidence suggests a low likelihood of an identification).

In the context of viewing distance, relatively few studies have attempted to examine confidence–accuracy specifically. Semmler et al. (2018) re-analyzed the tconconfidencee accuracy-relationship from Lindsay et al. (2008) by constructing calibration curves and found that even though viewing distance had large effects on memory (as indexed by d prime, a measure of discriminability), the confidence–accuracy relationship itself remained largely intact (although note that this was true only when deidentification was delayed; the same was not true when deidentification was immediate). They posed the argument that simply because a given estimator variable affects memory does not mean that the variable will necessariinfluencenconfidencence judgments, and identifications made by highly confident witnesses are still likely to be accurate. 

Under a global memory and metamemory impairment framework, any variable that hurts memory overall would also reduce the threshold that an eyewitness might have for making a high-confidence judgment. For example, if one were to witness a burglar late at night under poor viewing conditions, we would expect the likelihood of correctly identifying the burglar later to be quite low. A global framework would predict that on average, calibration would also be impaired and an eyewitness in this scenario (compared to one witnessing under ideal viewing conditions) would be more likely to erroneously assign higher-confidence judgment tomisidentificationion, making tconfidencence judgment an overestimator of performance.

What Semmler et al. (2018) and we (Davis et al., 2019) have proposed is that factors that affect memory performance overall do not necessarily impair metamemory judgments (e.g., Co et al., 1997; Leippe, 1980; Penrod & Cutler, 1995). In essence, the idea is that eyewitnesses know implicitly or explicitly that their memory performance is poor and scale the confidence ratings downward as they make identifcationThisTis leaves the confidence–accuracy relationship largely intact, meaning that triers of fact can still place their faith in high-confidence eyewitness judgments even when the circumstances surrounding identification on are suboptimal. 

In contrast, Man et al. (2019) analyzed– the accuracy relationship in their study using an in vivo testing procedure in a science center and found tconfidenceonfdence–accuracy relationship appeared to be preserved for shorter distances, but not longer distances. Lockamyier et al. (202identifiedfed a similar pattern in the theconfidencence–accuracy characteristic curves, with good calibration observed at short distances (3  m) but not at longer distances (10 and 20 m), again showing that increasing distance may be associated with impairments to tconfidencence–accuracy relationship. Findings have also been demonstrated for estimator variables other than viewing distance (e.g., the croeffectce effect; Dodson & Dobolyi, 2016).

According to Nyman et al.’s (2019) threshold model, participants in conditions where discriminability is especially low may fail to realize just how difficult the task is, and that success is unliThisy. Tis leads them to approach making their metacognitive judgments in the way that they would for only a moderately difficult task, which means that the majority of their judgments will overestimate their actual performance. Research on metacognition in the education literature supports this idea; learners tend to be more confident about difficult material compared to difficult material because they tend to erroneously assume that their level of performance will remain the same even when difficulty increases (e.g., Kelemen et al., 2000; Maki t  al., 2005; Schraw & Roedel, 1994).

However, both Nyman et al. (2019) and Lockamyier et al. (2020) acknowledged that there were relatively few positiidentificationsons made the very long distances, making it difficult to calculate the appropriate inferential statistics at this highest levels of confidence and casting into doubt whether the threshoeffectect observed was simply due to high variability in high-confidence judgments at long distancesThisTis methodological problem is exacerbated by the lineup procedure, which has high applied value but by necessity only yields a single data point per condition, further increasing variability. In the present studies, we examined the impact of viewing distanceconfidenceonfdence–accuracy relationship using a face recognition paradigm, which generates dozens of observations per conditioThisTis in turn yields more reliable estimates of accuracy at each level of confidence. Whereas single-face lineidentificationsons and old/new decisions for a large set of previously presented faces are nsuperficiallylly identical tasks, they both rely on similar cognitive architectures of facial processing (Morgan et al., 2007), with face recognition often being useful as a basic research tool for testing theoretical positions. 

Although there are differences between face recognition and lineup paradigms (see Weber & Brewer, 2006, for a discussion of the similarities between the paradigms), we believe that the increased power and stability of the estimates outweigh the disadvantages of reduced ecological validity.

In addition to maximizing statistical power by using a face-recognition paradigm, we conducted the current studies in a more powerful within-subjects design meant to maximize the reliability of point estimates. 

Whereas man et al. (2019) did require participants to make fouridentificationsfcations of possible targets, these were subdivided into both target-present and target-absent lineups, and each task was completed sequentially (e.g., participants studied and were tested single-fatigue face at a time), making it more likely that participants could adjust their global criteria for choosing and for asconfidenceonfdence judgments from task to task. We argue that while the data provided by Nyman et al. provide some evidence in favor of a threshold account, more rigorous laboratory testing should be done with a sample and design created to detect situations of a confidence–accuracy relationship.

We also conducted our studies using both college students and workers on Amazon’s Mechanical Turk (MTurk). Because eyewitnesses in the USA can probabilistically be any of several demographic profiles present in the country, participant samples must match this level of diversity. Thus, the MTurk sample allows a more general estimate of the abilities of the US population,1 rather than just relying on a highly educated, homogenous sample of college students to generalize to the population of interest. Finally, we implemented a simple procedural tactic (i.e., an instructional warning) meant to correct overestimates of confdence in Experiment 3, as it is important to determine whether any negative infuence of estimator variables can be corrected by policy changes that can easily be implemented in the feld.

For more information:1950477648nn@gmail.com