Error Rates in Users of Automatic Face Recognition Software

In recent years, wide deployment of automatic face recognition systems has been accompanied by substantial gains in algorithm performance. However, benchmarking tests designed to evaluate these systems do not account for the errors of human operators, who are often an integral part of face recognition solutions in forensic and security settings. This causes a mismatch between evaluation tests and operational accuracy. We address this by measuring user performance in a face recognition system used to screen passport applications for identity fraud. Experiment 1 measured target detection accuracy in algorithm-generated ‘candidate lists’ selected from a large database of passport images. Accuracy was notably poorer than in previous studies of unfamiliar face matching: participants made over 50% errors for adult target faces, and over 60% when matching images of children. Experiment 2 then compared performance of student participants to trained passport officers–who use the system in their daily work–and found equivalent performance in these groups. Encouragingly, a group of highly trained and experienced “facial examiners” outperformed these groups by 20 percentage points. We conclude that human performance curtails accuracy of face recognition systems–potentially reducing benchmark estimates by 50% in operational settings. Mere practise does not attenuate these limits, but superior performance of trained examiners suggests that recruitment and selection of human operators, in combination with effective training and mentorship, can improve the operational accuracy of face recognition systems.     

Neural Tuning Size in a Model of Primate Visual Processing Accounts for Three Key Markers of Holistic Face Processing

Faces are an important and unique class of visual stimuli, and have been of interest to neuroscientists for many years. Faces are known to elicit certain characteristic behavioral markers, collectively labeled “holistic processing”, while non-face objects are not processed holistically. However, little is known about the underlying neural mechanisms. The main aim of this computational simulation work is to investigate the neural mechanisms that make face processing holistic. Using a model of primate visual processing, we show that a single key factor, “neural tuning size”, is able to account for three important markers of holistic face processing: the Composite Face Effect (CFE), Face Inversion Effect (FIE) and Whole-Part Effect (WPE). Our proof-of-principle specifies the precise neurophysiological property that corresponds to the poorly-understood notion of holism, and shows that this one neural property controls three classic behavioral markers of holism. Our work is consistent with neurophysiological evidence, and makes further testable predictions. Overall, we provide a parsimonious account of holistic face processing, connecting computation, behavior and neurophysiology.     

Face Recognition by Metropolitan Police Super-Recognisers

Face recognition is used to prove identity across a wide variety of settings. Despite this, research consistently shows that people are typically rather poor at matching faces to photos. Some professional groups, such as police and passport officers, have been shown to perform just as poorly as the general public on standard tests of face recognition. However, face recognition skills are subject to wide individual variation, with some people showing exceptional ability—a group that has come to be known as ‘super-recognisers’. The Metropolitan Police Force (London) recruits ‘super-recognisers’ from within its ranks, for deployment on various identification tasks. Here we test four working super-recognisers from within this police force, and ask whether they are really able to perform at levels above control groups. We consistently find that the police ‘super-recognisers’ perform at well above normal levels on tests of unfamiliar and familiar face matching, with degraded as well as high quality images. Recruiting employees with high levels of skill in these areas, and allocating them to relevant tasks, is an efficient way to overcome some of the known difficulties associated with unfamiliar face recognition.     

I Choose,Therefore I Like: Preference for Faces Induced by Arbitrary Choice

Behavioral choice alters one’s preference rather than simply reflecting it. This effect to fit preferences with past choice, is known as “choice-induced preference change.” After making a choice between two equally attractive options, one tends to rate the chosen option better than they initially did and/or the unchosen option worse. The present study examined how behavioral choice changes subsequent preference, using facial images for the choice options as well as blind choice techniques. Participants rated their facial preference for each face, and chose between two equally preferred faces and subsequently rated their facial preference. Results from four experiments demonstrated that randomly chosen faces were more preferred only after participants were required to choose “a preferred face,” (in Experiment 1) but not “an unpreferred face,” (in Experiment 2) or “a rounder face” (in Experiment 3). Further, preference change was still observed after participants were informed that choices were actually random (in Experiment 4). Our findings provide new and important implications characterizing the conditions under which random choice changes preference, and show that people are tempted to make a biased evaluation even after they know that they did not make the choice for themselves.     

More Realistic Face Model Surface Improves Relevance of Pediatric In-Vitro Aerosol Studies

Background

Various hard face models are commonly used to evaluate the efficiency of aerosol face masks. Softer more realistic “face” surface materials, like skin, deform upon mask application and should provide more relevant in-vitro tests. Studies that simultaneously take into consideration many of the factors characteristic of the in vivo face are lacking. These include airways, various application forces, comparison of various devices, comparison with a hard-surface model and use of a more representative model face based on large numbers of actual faces.

Aim

To compare mask to “face” seal and aerosol delivery of two pediatric masks using a soft vs. a hard, appropriately representative, pediatric face model under various applied forces.

Methods

Two identical face models and upper airways replicas were constructed, the only difference being the suppleness and compressibility of the surface layer of the “face.” Integrity of the seal and aerosol delivery of two different masks [AeroChamber (AC) and SootherMask (SM)] were compared using a breath simulator, filter collection and realistic applied forces.

Results

The soft “face” significantly increased the delivery efficiency and the sealing characteristics of both masks. Aerosol delivery with the soft “face” was significantly greater for the SM compared to the AC (p< 0.01). No statistically significant difference between the two masks was observed with the hard “face.”

Conclusions

The material and pliability of the model “face” surface has a significant influence on both the seal and delivery efficiency of face masks. This finding should be taken into account during in-vitro aerosol studies.     

Eye Tracking the Face in the Crowd Task: Why Are Angry Faces Found More Quickly?

Among a crowd of distractor faces, threatening or angry target faces are identified more quickly and accurately than are nonthreatening or happy target faces, a finding known as the “face in the crowd effect.” Two perceptual explanations of the effect have been proposed: (1) the “target orienting” hypothesis (i.e., threatening targets orient attention more quickly than do nonthreatening targets and (2) the “distractor processing” hypothesis (i.e., nonthreatening distractors paired with a threatening target are processed more efficiently than vice versa, leading to quicker detection of threatening targets). Using a task, with real faces and multiple identities, the current study replicated the face in the crowd effect and then, via eye tracking, found greater support for the target orienting hypothesis. Across both the classical search asymmetry paradigm (i.e., one happy target in a crowd of angry distractors vs. one angry target in a crowd of happy distractors) and the constant distractor paradigm (i.e., one happy target in a crowd of neutral distractors vs. one angry target in a crowd of neutral distractors), fewer distractors were fixated before first fixating angry targets relative to happy targets, with no difference in the processing efficiency of distractors. These results suggest that the face in the crowd effect on this task is supported to a greater degree by attentional patterns associated with properties of target rather those of the crowd.     

Examination of Cognitive Fatigue in Multiple Sclerosis using Functional Magnetic Resonance Imaging and Diffusion Tensor Imaging

The present study investigated the neural correlates of cognitive fatigue in Multiple Sclerosis (MS), looking specifically at the relationship between self-reported fatigue and objective measures of cognitive fatigue. In Experiment 1, functional magnetic resonance imaging (fMRI) was used to examine where in the brain BOLD activity covaried with “state” fatigue, assessed during performance of a task designed to induce cognitive fatigue while in the scanner. In Experiment 2, diffusion tensor imaging (DTI) was used to examine where in the brain white matter damage correlated with increased “trait” fatigue in individuals with MS, assessed by the Fatigue Severity Scale (FSS) completed outside the scanning session. During the cognitively fatiguing task, the MS group had increased brain activity associated with fatigue in the caudate as compared with HCs. DTI findings revealed that reduced fractional anisotropy in the anterior internal capsule was associated with increased self-reported fatigue on the FSS. Results are discussed in terms of identifying a “fatigue-network” in MS.     

Is Beauty in the Face of the Beholder?

Opposing forces influence assortative mating so that one seeks a similar mate while at the same time avoiding inbreeding with close relatives. Thus, mate choice may be a balancing of phenotypic similarity and dissimilarity between partners. In the present study, we assessed the role of resemblance to Self’s facial traits in judgments of physical attractiveness. Participants chose the most attractive face image of their romantic partner among several variants, where the faces were morphed so as to include only 22% of another face. Participants distinctly preferred a “Self-based morph” (i.e., their partner’s face with a small amount of Self’s face blended into it) to other morphed images. The Self-based morph was also preferred to the morph of their partner’s face blended with the partner’s same-sex “prototype”, although the latter face was (“objectively”) judged more attractive by other individuals. When ranking morphs differing in level of amalgamation (i.e., 11% vs. 22% vs. 33%) of another face, the 22% was chosen consistently as the preferred morph and, in particular, when Self was blended in the partner’s face. A forced-choice signal-detection paradigm showed that the effect of self-resemblance operated at an unconscious level, since the same participants were unable to detect the presence of their own faces in the above morphs. We concluded that individuals, if given the opportunity, seek to promote “positive assortment” for Self’s phenotype, especially when the level of similarity approaches an optimal point that is similar to Self without causing a conscious acknowledgment of the similarity.     

Explicating the Face Perception Network with White Matter Connectivity

A network of multiple brain regions is recruited in face perception. Our understanding of the functional properties of this network can be facilitated by explicating the structural white matter connections that exist between its functional nodes. We accomplished this using functional MRI (fMRI) in combination with fiber tractography on high angular resolution diffusion weighted imaging data. We identified the three nodes of the core face network: the “occipital face area” (OFA), the “fusiform face area” (mid-fusiform gyrus or mFus), and the superior temporal sulcus (STS). Additionally, a region of the anterior temporal lobe (aIT), implicated as being important for face perception was identified. Our data suggest that we can further divide the OFA into multiple anatomically distinct clusters – a partitioning consistent with several recent neuroimaging results. More generally, structural white matter connectivity within this network revealed: 1) Connectivity between aIT and mFus, and between aIT and occipital regions, consistent with studies implicating this posterior to anterior pathway as critical to normal face processing; 2) Strong connectivity between mFus and each of the occipital face-selective regions, suggesting that these three areas may subserve different functional roles; 3) Almost no connectivity between STS and mFus, or between STS and the other face-selective regions. Overall, our findings suggest a re-evaluation of the “core” face network with respect to what functional areas are or are not included in this network.     

Joint Cognition: Thought Contagion and the Consequences of Cooperation when Sharing the Task of Random Sequence Generation

Generating random number sequences is a popular psychological task often used to measure executive functioning. We explore random generation under “joint cognition” instructions; pairs of participants take turns to compile a shared response sequence. Across three studies, we point to six key findings from this novel format. First, there are both costs and benefits from group performance. Second, repetition avoidance occurs in dyadic as well as individual production settings. Third, individuals modify their choices in a dyadic situation such that the pair becomes the unit of psychological function. Fourth, there is immediate contagion of sequence stereotypy amongst the pairs (i.e., each contributor “owns” their partner’s response). Fifth, dyad effects occur even when participants know their partner is not interacting with them (Experiment 2). Sixth, ironically, directing participants’ efforts away from their shared task responsibility can actually benefit conjoint performance (Experiment 3). These results both constrain models of random generation and illuminate processes of joint cognition.     

Categorical Biases in Perceiving Spatial Relations

We investigate the effect of spatial categories on visual perception. In three experiments, participants made same/different judgments on pairs of simultaneously presented dot-cross configurations. For different trials, the position of the dot within each cross could differ with respect to either categorical spatial relations (the dots occupied different quadrants) or coordinate spatial relations (the dots occupied different positions within the same quadrant). The dot-cross configurations also varied in how readily the dot position could be lexicalized. In harder-to-name trials, crosses formed a “+” shape such that each quadrant was associated with two discrete lexicalized spatial categories (e.g., “above” and “left”). In easier-to-name trials, both crosses were rotated 45° to form an “×” shape such that quadrants were unambiguously associated with a single lexicalized spatial category (e.g., “above” or “left”). In Experiment 1, participants were more accurate when discriminating categorical information between easier-to-name categories and more accurate at discriminating coordinate spatial information within harder-to-name categories. Subsequent experiments attempted to down-regulate or up-regulate the involvement of language in task performance. Results from Experiment 2 (verbal interference) and Experiment 3 (verbal training) suggest that the observed spatial relation type-by-nameability interaction is resistant to online language manipulations previously shown to affect color and object-based perceptual processing. The results across all three experiments suggest that robust biases in the visual perception of spatial relations correlate with patterns of lexicalization, but do not appear to be modulated by language online.     

New Tests to Measure Individual Differences in Matching and Labelling Facial Expressions of Emotion,and Their Association with Ability to Recognise Vocal Emotions and Facial Identity

Although good tests are available for diagnosing clinical impairments in face expression processing, there is a lack of strong tests for assessing “individual differences” – that is, differences in ability between individuals within the typical, nonclinical, range. Here, we develop two new tests, one for expression perception (an odd-man-out matching task in which participants select which one of three faces displays a different expression) and one additionally requiring explicit identification of the emotion (a labelling task in which participants select one of six verbal labels). We demonstrate validity (careful check of individual items, large inversion effects, independence from nonverbal IQ, convergent validity with a previous labelling task), reliability (Cronbach’s alphas of.77 and.76 respectively), and wide individual differences across the typical population. We then demonstrate the usefulness of the tests by addressing theoretical questions regarding the structure of face processing, specifically the extent to which the following processes are common or distinct: (a) perceptual matching and explicit labelling of expression (modest correlation between matching and labelling supported partial independence); (b) judgement of expressions from faces and voices (results argued labelling tasks tap into a multi-modal system, while matching tasks tap distinct perceptual processes); and (c) expression and identity processing (results argued for a common first step of perceptual processing for expression and identity).     

The Embodiment of Success and Failure as Forward versus Backward Movements

People often speak of success (e.g., “advance”) and failure (e.g., “setback”) as if they were forward versus backward movements through space. Two experiments sought to examine whether grounded associations of this type influence motor behavior. In Experiment 1, participants categorized success versus failure words by moving a joystick forward or backward. Failure categorizations were faster when moving backward, whereas success categorizations were faster when moving forward. Experiment 2 removed the requirement to categorize stimuli and used a word rehearsal task instead. Even without Experiment 1’s response procedures, a similar cross-over interaction was obtained (e.g., failure memorizations sped backward movements relative to forward ones). The findings are novel yet consistent with theories of embodied cognition and self-regulation.     

Occurrence of Coliforms, Fecal Coliforms, and Streptococci on Vegetation and Insects

This study considers the sanitary significance of coliforms, fecal coliforms, and streptococci isolated from 152 species of plants and 40 samples of insects. These specimens were collected from various ecological environments and grouped into several categories. Results indicate that typical coliforms of the warm-blooded animal gut contribute a relatively small percentage of the organisms associated with vegetation (14.1%) and insects, (14.9%). A total of 1,203 coliform strains from vegetation and 1,084 coliform strains from insects were classified as to IMViC type and fecal coliform. No type was predominant in either the vegetation or insect groupings. The biochemical results for 646 streptococci from vegetation and 226 cultures from insects were reported. The predominant group, Streptococcus fecalis, as defined by Sherman criteria, constituted a majority of all strains from vegetation and insects. The “Completed Coliform Test” is recommended for the examination of plant and insect specimens to eliminate the many anaerobic and aerobic sporeforming bacteria that frequently produce false positive reactions by the “Confirmed Test” procedure. These findings support the current interpretation of the significance of the fecal coliform test for stream investigations or for surface water quality evaluations.     

The Emergence of Three Human Development Clubs

We examine the joint distribution of levels of income per capita, life expectancy, and years of schooling across countries in 1960 and in 2000. In 1960 countries were clustered in two groups; a rich, highly educated, high longevity “developed” group and a poor, less educated, high mortality, “underdeveloped” group. By 2000 however we see the emergence of three groups; one underdeveloped group remaining near 1960 levels, a developed group with higher levels of education, income, and health than in 1960, and an intermediate group lying between these two. This finding is consistent with both the ideas of a new “middle income trap” that countries face even if they escape the “low income trap”, as well as the notion that countries which escaped the poverty trap form a temporary “transition regime” along their path to the “developed” group.     

Modelling how cleaner fish approach an ephemeral reward task demonstrates a role for ecologically tuned chunking in the evolution of advanced cognition

What makes cognition “advanced” is an open and not precisely defined question. One perspective involves increasing the complexity of associative learning, from conditioning to learning sequences of events (“chaining”) to representing various cue combinations as “chunks.” Here we develop a weighted graph model to study the mechanism enabling chunking ability and the conditions for its evolution and success, based on the ecology of the cleaner fish Labroides dimidiatus. In some environments, cleaners must learn to serve visitor clients before resident clients, because a visitor leaves if not attended while a resident waits for service. This challenge has been captured in various versions of the ephemeral reward task, which has been proven difficult for a range of cognitively capable species. We show that chaining is the minimal requirement for solving this task in its common simplified laboratory format that involves repeated simultaneous exposure to an ephemeral and permanent food source. Adding ephemeral–ephemeral and permanent–permanent combinations, as cleaners face in the wild, requires individuals to have chunking abilities to solve the task. Importantly, chunking parameters need to be calibrated to ecological conditions in order to produce adaptive decisions. Thus, it is the fine-tuning of this ability, which may be the major target of selection during the evolution of advanced associative learning.

What makes cognition ‘advanced’ is an open and not precisely defined question. In this study, a cognitive model of cleaner fish learning the ephemeral-reward task demonstrates how a critical step in cognitive evolution may be understood as the evolution of chunking and its tuning to fit ecological conditions.     

Reporting Quality of Systematic Reviews/Meta-Analyses of Acupuncture

Background

The QUOROM and PRISMA statements were published in 1999 and 2009, respectively, to improve the consistency of reporting systematic reviews (SRs)/meta-analyses (MAs) of clinical trials. However, not all SRs/MAs adhere completely to these important standards. In particular, it is not clear how well SRs/MAs of acupuncture studies adhere to reporting standards and which reporting criteria are generally ignored in these analyses.

Objectives

To evaluate reporting quality in SRs/MAs of acupuncture studies.

Methods

We performed a literature search for studies published prior to 2014 using the following public archives: PubMed, EMBASE, Web of Science, the Cochrane Database of Systematic Reviews (CDSR), the Chinese Biomedical Literature Database (CBM), the Traditional Chinese Medicine (TCM) database, the Chinese Journal Full-text Database (CJFD), the Chinese Scientific Journal Full-text Database (CSJD), and the Wanfang database. Data were extracted into pre-prepared Excel data-extraction forms. Reporting quality was assessed based on the PRISMA checklist (27 items).

Results

Of 476 appropriate SRs/MAs identified in our search, 203, 227, and 46 were published in Chinese journals, international journals, and the Cochrane Database, respectively. In 476 SRs/MAs, only 3 reported the information completely. By contrast, approximately 4.93% (1/203), 8.81% (2/227) and 0.00% (0/46) SRs/Mas reported less than 10 items in Chinese journals, international journals and CDSR, respectively. In general, the least frequently reported items (reported≤50%) in SRs/MAs were “protocol and registration”, “risk of bias across studies”, and “additional analyses” in both methods and results sections.

Conclusions

SRs/MAs of acupuncture studies have not comprehensively reported information recommended in the PRISMA statement. Our study underscores that, in addition to focusing on careful study design and performance, attention should be paid to comprehensive reporting standards in SRs/MAs on acupuncture studies.     

Performance of Single and Concatenated Sets of Mitochondrial Genes at Inferring Metazoan Relationships Relative to Full Mitogenome Data

Mitochondrial (mt) genes are some of the most popular and widely-utilized genetic loci in phylogenetic studies of metazoan taxa. However, their linked nature has raised questions on whether using the entire mitogenome for phylogenetics is overkill (at best) or pseudoreplication (at worst). Moreover, no studies have addressed the comparative phylogenetic utility of mitochondrial genes across individual lineages within the entire Metazoa. To comment on the phylogenetic utility of individual mt genes as well as concatenated subsets of genes, we analyzed mitogenomic data from 1865 metazoan taxa in 372 separate lineages spanning genera to subphyla. Specifically, phylogenies inferred from these datasets were statistically compared to ones generated from all 13 mt protein-coding (PC) genes (i.e., the “supergene” set) to determine which single genes performed “best” at, and the minimum number of genes required to, recover the “supergene” topology. Surprisingly, the popular marker COX1 performed poorest, while ND5, ND4, and ND2 were most likely to reproduce the “supergene” topology. Averaged across all lineages, the longest ∼2 mt PC genes were sufficient to recreate the “supergene” topology, although this average increased to ∼5 genes for datasets with 40 or more taxa. Furthermore, concatenation of the three “best” performing mt PC genes outperformed that of the three longest mt PC genes (i.e, ND5, COX1, and ND4). Taken together, while not all mt PC genes are equally interchangeable in phylogenetic studies of the metazoans, some subset can serve as a proxy for the 13 mt PC genes. However, the exact number and identity of these genes is specific to the lineage in question and cannot be applied indiscriminately across the Metazoa.     

Acute Alcohol Consumption Impairs Controlled but Not Automatic Processes in a Psychophysical Pointing Paradigm

Numerous studies have investigated the effects of alcohol consumption on controlled and automatic cognitive processes. Such studies have shown that alcohol impairs performance on tasks requiring conscious, intentional control, while leaving automatic performance relatively intact. Here, we sought to extend these findings to aspects of visuomotor control by investigating the effects of alcohol in a visuomotor pointing paradigm that allowed us to separate the influence of controlled and automatic processes. Six male participants were assigned to an experimental “correction” condition in which they were instructed to point at a visual target as quickly and accurately as possible. On a small percentage of trials, the target “jumped” to a new location. On these trials, the participants’ task was to amend their movement such that they pointed to the new target location. A second group of 6 participants were assigned to a “countermanding” condition, in which they were instructed to terminate their movements upon detection of target “jumps”. In both the correction and countermanding conditions, participants served as their own controls, taking part in alcohol and no-alcohol conditions on separate days. Alcohol had no effect on participants’ ability to correct movements “in flight”, but impaired the ability to withhold such automatic corrections. Our data support the notion that alcohol selectively impairs controlled processes in the visuomotor domain.     

Do I Have My Attention? Speed of Processing Advantages for the Self-Face Are Not Driven by Automatic Attention Capture

We respond more quickly to our own face than to other faces, but there is debate over whether this is connected to attention-grabbing properties of the self-face. In two experiments, we investigate whether the self-face selectively captures attention, and the attentional conditions under which this might occur. In both experiments, we examined whether different types of face (self, friend, stranger) provide differential levels of distraction when processing self, friend and stranger names. In Experiment 1, an image of a distractor face appeared centrally – inside the focus of attention – behind a target name, with the faces either upright or inverted. In Experiment 2, distractor faces appeared peripherally – outside the focus of attention – in the left or right visual field, or bilaterally. In both experiments, self-name recognition was faster than other name recognition, suggesting a self-referential processing advantage. The presence of the self-face did not cause more distraction in the naming task compared to other types of face, either when presented inside (Experiment 1) or outside (Experiment 2) the focus of attention. Distractor faces had different effects across the two experiments: when presented inside the focus of attention (Experiment 1), self and friend images facilitated self and friend naming, respectively. This was not true for stranger stimuli, suggesting that faces must be robustly represented to facilitate name recognition. When presented outside the focus of attention (Experiment 2), no facilitation occurred. Instead, we report an interesting distraction effect caused by friend faces when processing strangers’ names. We interpret this as a “social importance” effect, whereby we may be tuned to pick out and pay attention to familiar friend faces in a crowd. We conclude that any speed of processing advantages observed in the self-face processing literature are not driven by automatic attention capture.