Subtle adversarial image manipulations influence both human and machine perception.

Although artificial neural networks (ANNs) were inspired by the brain, ANNs exhibit a brittleness not generally observed in human perception. One shortcoming of ANNs is their susceptibility to adversarial perturbations-subtle modulations of natural images that result in changes to classification decisions, such as confidently mislabelling an image of an elephant, initially classified correctly, as a clock. In contrast, a human observer might well dismiss the perturbations as an innocuous imaging artifact. This phenomenon may point to a fundamental difference between human and machine perception, but it drives one to ask whether human sensitivity to adversarial perturbations might be revealed with appropriate behavioral measures. Here, we find that adversarial perturbations that fool ANNs similarly bias human choice. We further show that the effect is more likely driven by higher-order statistics of natural images to which both humans and ANNs are sensitive, rather than by the detailed architecture of the ANN.

Haboob Dust Storms and Motor Vehicle Collision-related Trauma in Phoenix, Arizona.

BACKGROUND: The Sonoran Desert region, encompassing most of southern Arizona, has an extreme climate that is famous for dust storms known as haboobs. These storms lead to decreased visibility and potentially hazardous driving conditions. In this study we evaluate the relationship between haboob events and emergency department (ED) visits due to motor vehicle collisions (MVCs) in Phoenix, Arizona. METHODS: This study is a retrospective analysis of MVC-related trauma presentations to Phoenix, AZ, hospitals before and following haboob dust storms. These events were identified from 2009-2017 primarily using Phoenix International Airport weather data. De-identified trauma data were obtained from the Arizona Department of Health Services (ADHS) Arizona State Trauma Registry (ASTR) from seven trauma centers within a 10-mile radius of the airport. We compared MVC-related trauma using six- and 24-hour windows before and following the onset of haboob events. RESULTS: There were 31,133 MVC-related trauma encounters included from 2009-2017 and 111 haboob events meeting meteorological criteria during that period. There was a 17% decrease in MVC-related ED encounters in the six hours following haboob onset compared to before onset (235 vs 283, P = 0.04), with proportionally more injuries among males (P < 0.001) and higher mortality (P = 0.02). There was no difference in frequency of presentations (P = 0.82), demographics, or outcomes among the 24-hour pre-and post-haboob groups. CONCLUSION: Haboob dust storms in Phoenix, Arizona, are associated with a decrease in MVC-related injuries during the six-hour period following storm onset, likely indicating the success of public safety messaging efforts. Males made up a higher proportion of those injured during the storms, suggesting a target for future interventions. Future public-targeted weather-safety initiatives should be accompanied more closely by monitoring and evaluation efforts to assess for effectiveness.

Predicting the Ease of Human Category Learning Using Radial Basis Function Networks.

Our goal is to understand and optimize human concept learning by predicting the ease of learning of a particular exemplar or category. We propose a method for estimating ease values, quantitative measures of ease of learning, as an alternative to conducting costly empirical training studies. Our method combines a psychological embedding of domain exemplars with a pragmatic categorization model. The two components are integrated using a radial basis function network (RBFN) that predicts ease values. The free parameters of the RBFN are fit using human similarity judgments, circumventing the need to collect human training data to fit more complex models of human categorization. We conduct two category-training experiments to validate predictions of the RBFN. We demonstrate that an instance-based RBFN outperforms both a prototype-based RBFN and an empirical approach using the raw data. Although the human data were collected across diverse experimental conditions, the predicted ease values strongly correlate with human learning performance. Training can be sequenced by (predicted) ease, achieving what is known as fading in the psychology literature and curriculum learning in the machine-learning literature, both of which have been shown to facilitate learning.

Highlights as an Early Predictor of Student Comprehension and Interests.

When engaging with a textbook, students are inclined to highlight key content. Although students believe that highlighting and subsequent review of the highlights will further their educational goals, the psychological literature provides little evidence of benefits. Nonetheless, a student's choice of text for highlighting may serve as a window into her mental state-her level of comprehension, grasp of the key ideas, reading goals, and so on. We explore this hypothesis via an experiment in which 400 participants read three sections from a college-level biology text, briefly reviewed the text, and then took a quiz on the material. During initial reading, participants were able to highlight words, phrases, and sentences, and these highlights were displayed along with the complete text during the subsequent review. Consistent with past research, the amount of highlighted material is unrelated to quiz performance. Nonetheless, highlighting patterns may allow us to infer reader comprehension and interests. Using multiple representations of the highlighting patterns, we built probabilistic models to predict quiz performance and matrix factorization models to predict what content would be highlighted in one passage from highlights in other passages. We find that quiz score prediction accuracy reliably improves with the inclusion of highlighting data (by about 1%-2%), both for held-out students and for held-out student questions (i.e., questions selected randomly for each student), but not for held-out questions. Furthermore, an individual's highlighting pattern is informative of what she highlights elsewhere. Our long-term goal is to design digital textbooks that serve not only as conduits of information into the reader's mind but also allow us to draw inferences about the reader at a point where interventions may increase the effectiveness of the material.

Sympathetically mediated increases in cardiac output, not restraint of peripheral vasodilation, contribute to blood pressure maintenance during hyperinsulinemia.

Vasodilatory effects of insulin support the delivery of insulin and glucose to skeletal muscle. Concurrently, insulin exerts central effects that increase sympathetic nervous system activity (SNA), which is required for the acute maintenance of blood pressure (BP). Indeed, in a cohort of young healthy adults, herein we show that intravenous infusion of insulin increases muscle SNA while BP is maintained. We next tested the hypothesis that sympathoexcitation evoked by hyperinsulinemia restrains insulin-stimulated peripheral vasodilation and contributes to sustaining BP. To address this, a separate cohort of participants were subjected to 5-s pulses of neck suction (NS) to simulate carotid hypertension and elicit a reflex-mediated reduction in SNA. NS was conducted before and 60 min following intravenous infusion of insulin. Insulin infusion caused an increase in leg vascular conductance and cardiac output (CO; P < 0.050), with maintenance of BP (P = 0.540). As expected, following NS, decreases in BP were greater in the presence of hyperinsulinemia compared with control (P = 0.045). However, the effect of NS on leg vascular conductance did not differ between insulin and control conditions (P = 0.898). Instead, the greater decreases in BP following NS in the setting of insulin infusion paralleled with greater decreases in CO (P = 0.009). These findings support the idea that during hyperinsulinemia, SNA-mediated increase in CO, rather than restraint of leg vascular conductance, is the principal contributor to the maintenance of BP. Demonstration in isolated arteries that insulin suppresses α-adrenergic vasoconstriction suggests that the observed lack of restraint of leg vascular conductance may be attributed to sympatholytic actions of insulin.NEW & NOTEWORTHY We examined the role of sympathetic activation in restraining vasodilatory responses to hyperinsulinemia and sustaining blood pressure in healthy adults. Data are reported from two separate experimental protocols in humans and one experimental protocol in isolated arteries from mice. Contrary to our hypothesis, the present findings support the idea that during hyperinsulinemia, a sympathetically mediated increase in cardiac output, rather than restraint of peripheral vasodilation, is the principal contributor to the maintenance of systemic blood pressure.

Role of the carotid chemoreceptors in insulin-mediated sympathoexcitation in humans.

We examined the contribution of the carotid chemoreceptors to insulin-mediated increases in muscle sympathetic nerve activity (MSNA) in healthy humans. We hypothesized that reductions in carotid chemoreceptor activity would attenuate the sympathoexcitatory response to hyperinsulinemia. Young, healthy adults (9 male/9 female, 28 ± 1 yr, 24 ± 1 kg/m2) completed a 30-min euglycemic baseline followed by a 90-min hyperinsulinemic (1 mU·kg fat-free mass-1·min-1), euglycemic infusion. MSNA (microneurography of the peroneal nerve) was continuously measured. The role of the carotid chemoreceptors was assessed at baseline and during hyperinsulinemia via 1) acute hyperoxia, 2) low-dose dopamine (1-4 µg·kg-1·min-1), and 3) acute hyperoxia + low-dose dopamine. MSNA burst frequency increased from baseline during hyperinsulinemia (P < 0.01). Acute hyperoxia had no effect on MSNA burst frequency at rest (P = 0.74) or during hyperinsulinemia (P = 0.83). The insulin-mediated increase in MSNA burst frequency (P = 0.02) was unaffected by low-dose dopamine (P = 0.60). When combined with low-dose dopamine, acute hyperoxia had no effect on MSNA burst frequency at rest (P = 0.17) or during hyperinsulinemia (P = 0.85). Carotid chemoreceptor desensitization in young, healthy men and women does not attenuate the sympathoexcitatory response to hyperinsulinemia. Our data suggest that the carotid chemoreceptors do not contribute to acute insulin-mediated increases in MSNA in young, healthy adults.

Obtaining psychological embeddings through joint kernel and metric learning.

Psychological embeddings provide a powerful formalism for characterizing human-perceived similarity among members of a stimulus set. Obtaining high-quality embeddings can be costly due to algorithm design, software deployment, and participant compensation. This work aims to advance state-of-the-art embedding techniques and provide a comprehensive software package that makes obtaining high-quality psychological embeddings both easy and relatively efficient. Contributions are made on four fronts. First, the embedding procedure allows multiple trial configurations (e.g., triplets) to be used for collecting similarity judgments from participants. For example, trials can be configured to collect triplet comparisons or to sort items into groups. Second, a likelihood model is provided for three classes of similarity kernels allowing users to easily infer the parameters of their preferred model using gradient descent. Third, an active selection algorithm is provided that makes data collection more efficient by proposing comparisons that provide the strongest constraints on the embedding. Fourth, the likelihood model allows the specification of group-specific attention weight parameters. A series of experiments are included to highlight each of these contributions and their impact on converging to a high-quality embedding. Collectively, these incremental improvements provide a powerful and complete set of tools for inferring psychological embeddings. The relevant tools are available as the Python package PsiZ, which can be cloned from GitHub ( https://github.com/roads/psiz ).

Deep neural network improves fracture detection by clinicians.

Suspected fractures are among the most common reasons for patients to visit emergency departments (EDs), and X-ray imaging is the primary diagnostic tool used by clinicians to assess patients for fractures. Missing a fracture in a radiograph often has severe consequences for patients, resulting in delayed treatment and poor recovery of function. Nevertheless, radiographs in emergency settings are often read out of necessity by emergency medicine clinicians who lack subspecialized expertise in orthopedics, and misdiagnosed fractures account for upward of four of every five reported diagnostic errors in certain EDs. In this work, we developed a deep neural network to detect and localize fractures in radiographs. We trained it to accurately emulate the expertise of 18 senior subspecialized orthopedic surgeons by having them annotate 135,409 radiographs. We then ran a controlled experiment with emergency medicine clinicians to evaluate their ability to detect fractures in wrist radiographs with and without the assistance of the deep learning model. The average clinician's sensitivity was 80.8% (95% CI, 76.7-84.1%) unaided and 91.5% (95% CI, 89.3-92.9%) aided, and specificity was 87.5% (95 CI, 85.3-89.5%) unaided and 93.9% (95% CI, 92.9-94.9%) aided. The average clinician experienced a relative reduction in misinterpretation rate of 47.0% (95% CI, 37.4-53.9%). The significant improvements in diagnostic accuracy that we observed in this study show that deep learning methods are a mechanism by which senior medical specialists can deliver their expertise to generalists on the front lines of medicine, thereby providing substantial improvements to patient care.

Rejecting salient distractors: Generalization from experience.

Distraction impairs performance of many important, everyday tasks. Attentional control limits distraction by preferentially selecting important items for limited-capacity cognitive operations. Research in attentional control has typically investigated the degree to which selection of items is stimulus-driven versus goal-driven. Recent work finds that when observers initially learn a task, the selection is based on stimulus-driven factors, but through experience, goal-driven factors have an increasing influence. The modulation of selection by goals has been studied within the paradigm of learned distractor rejection, in which experience over a sequence of trials enables individuals eventually to ignore a perceptually salient distractor. The experiments presented examine whether observers can generalize learned distractor rejection to novel distractors. Observers searched for a target and ignored a salient color-singleton distractor that appeared in half of the trials. In Experiment 1, observers who learned distractor rejection in a variable environment rejected a novel distractor more effectively than observers who learned distractor rejection in a less variable, homogeneous environment, demonstrating that variable, heterogeneous stimulus environments encourage generalizable learned distractor rejection. Experiments 2 and 3 investigated the time course of learned distractor rejection across the experiment and found that after experiencing four color-singleton distractors in different blocks, observers could effectively reject subsequent novel color-singleton distractors. These results suggest that the optimization of attentional control to the task environment can be interpreted as a form of learning, demonstrating experience's critical role in attentional control.

Forgetting of Foreign-Language Skills: A Corpus-Based Analysis of Online Tutoring Software.

We explore the nature of forgetting in a corpus of 125,000 students learning Spanish using the Rosetta Stone® foreign-language instruction software across 48 lessons. Students are tested on a lesson after its initial study and are then retested after a variable time lag. We observe forgetting consistent with power function decay at a rate that varies across lessons but not across students. We find that lessons which are better learned initially are forgotten more slowly, a correlation which likely reflects a latent cause such as the quality or difficulty of the lesson. We obtain improved predictive accuracy of the forgetting model by augmenting it with features that encode characteristics of a student's initial study of the lesson and the activities the student engaged in between the initial and delayed tests. The augmented model can predict 23.9% of the variance in an individual's score on the delayed test. We analyze which features best explain individual performance.

Improving Human-Machine Cooperative Classification Via Cognitive Theories of Similarity.

Acquiring perceptual expertise is slow and effortful. However, untrained novices can accurately make difficult classification decisions (e.g., skin-lesion diagnosis) by reformulating the task as similarity judgment. Given a query image and a set of reference images, individuals are asked to select the best matching reference. When references are suitably chosen, the procedure yields an implicit classification of the query image. To optimize reference selection, we develop and evaluate a predictive model of similarity-based choice. The model builds on existing psychological literature and accommodates stochastic, dynamic shifts of attention among visual feature dimensions. We perform a series of human experiments with two stimulus types (rectangles, faces) and nine classification tasks to validate the model and to demonstrate the model's potential to boost performance. Our system achieves high accuracy for participants who are naive as to the classification task, even when the classification task switches from trial to trial.

Reductions in carotid chemoreceptor activity with low-dose dopamine improves baroreflex control of heart rate during hypoxia in humans.

The purpose of the present investigation was to examine the contribution of the carotid body chemoreceptors to changes in baroreflex control of heart rate with exposure to hypoxia. We hypothesized spontaneous cardiac baroreflex sensitivity (scBRS) would be reduced with hypoxia and this effect would be blunted when carotid chemoreceptor activity was reduced with low-dose dopamine. Fifteen healthy adults (11 M/4 F) completed two visits randomized to intravenous dopamine or placebo (saline). On each visit, subjects were exposed to 5-min normoxia (~99% SpO2), followed by 5-min hypoxia (~84% SpO2). Blood pressure (intra-arterial catheter) and heart rate (ECG) were measured continuously and scBRS was assessed by spectrum and sequence methodologies. scBRS was reduced with hypoxia (P < 0.01). Using the spectrum analysis approach, the fall in scBRS with hypoxia was attenuated with infusion of low-dose dopamine (P < 0.01). The decrease in baroreflex sensitivity to rising pressures (scBRS "up-up") was also attenuated with low-dose dopamine (P < 0.05). However, dopamine did not attenuate the decrease in baroreflex sensitivity to falling pressures (scBRS "down-down"; P > 0.05). Present findings are consistent with a reduction in scBRS with systemic hypoxia. Furthermore, we show this effect is partially mediated by the carotid body chemoreceptors, given the fall in scBRS is attenuated when activity of the chemoreceptors is reduced with low-dose dopamine. However, the improvement in scBRS with dopamine appears to be specific to rising blood pressures. These results may have important implications for impairments in baroreflex function common in disease states of acute and/or chronic hypoxemia, as well as the experimental use of dopamine to assess such changes.

Using Highlighting to Train Attentional Expertise.

Acquiring expertise in complex visual tasks is time consuming. To facilitate the efficient training of novices on where to look in these tasks, we propose an attentional highlighting paradigm. Highlighting involves dynamically modulating the saliency of a visual image to guide attention along the fixation path of a domain expert who had previously viewed the same image. In Experiment 1, we trained naive subjects via attentional highlighting on a fingerprint-matching task. Before and after training, we asked subjects to freely inspect images containing pairs of prints and determine whether the prints matched. Fixation sequences were automatically scored for the degree of expertise exhibited using a Bayesian discriminative model of novice and expert gaze behavior. Highlighted training causes gaze behavior to become more expert-like not only on the trained images but also on transfer images, indicating generalization of learning. In Experiment 2, to control for the possibility that the increase in expertise is due to mere exposure, we trained subjects via highlighting of fixation sequences from novices, not experts, and observed no transition toward expertise. In Experiment 3, to determine the specificity of the training effect, we trained subjects with expert fixation sequences from images other than the one being viewed, which preserves coarse-scale statistics of expert gaze but provides no information about fine-grain features. Observing at least a partial transition toward expertise, we obtain only weak evidence that the highlighting procedure facilitates the learning of critical local features. We discuss possible improvements to the highlighting procedure.

Interindividual variability in the dose-specific effect of dopamine on carotid chemoreceptor sensitivity to hypoxia.

Human studies use varying levels of low-dose (1-4 µg·kg(-1)·min(-1)) dopamine to examine peripheral chemosensitivity, based on its known ability to blunt carotid body responsiveness to hypoxia. However, the effect of dopamine on the ventilatory responses to hypoxia is highly variable between individuals. Thus we sought to determine 1) the dose response relationship between dopamine and peripheral chemosensitivity as assessed by the ventilatory response to hypoxia in a cohort of healthy adults, and 2) potential confounding cardiovascular responses at variable low doses of dopamine. Young, healthy adults (n = 30, age = 32 ± 1, 24 male/6 female) were given intravenous (iv) saline and a range of iv dopamine doses (1-4 µg·kg(-1)·min(-1)) prior to and throughout five hypoxic ventilatory response (HVR) tests. Subjects initially received iv saline, and after each HVR the dopamine infusion rate was increased by 1 µg·kg(-1)·min(-1). Tidal volume, respiratory rate, heart rate, blood pressure, and oxygen saturation were continuously measured. Dopamine significantly reduced HVR at all doses (P < 0.05). When subjects were divided into high (n = 13) and low (n = 17) baseline chemosensitivity, dopamine infusion (when assessed by dose) reduced HVR in the high group only (P < 0.01), with no effect of dopamine on HVR in the low group (P > 0.05). Dopamine infusion also resulted in a reduction in blood pressure (3 µg·kg(-1)·min(-1)) and total peripheral resistance (1-4 µg·kg(-1)·min(-1)), driven primarily by subjects with low baseline chemosensitivity. In conclusion, we did not find a single dose of dopamine that elicited a nadir HVR in all subjects. Additionally, potential confounding cardiovascular responses occur with dopamine infusion, which may limit its usage.

Effect of bilateral carotid body resection on cardiac baroreflex control of blood pressure during hypoglycemia.

Hypoglycemia results in a reduction in cardiac baroreflex sensitivity and a shift in the baroreflex working range to higher heart rates. This effect is mediated, in part, by the carotid chemoreceptors. Therefore, we hypothesized hypoglycemia-mediated changes in baroreflex control of heart rate would be blunted in carotid body-resected patients when compared with healthy controls. Five patients with bilateral carotid body resection for glomus tumors and 10 healthy controls completed a 180-minute hyperinsulinemic, hypoglycemic (≈3.3 mmol/L) clamp. Changes in heart rate, blood pressure, and spontaneous cardiac baroreflex sensitivity were assessed. Baseline baroreflex sensitivity was not different between groups (P>0.05). Hypoglycemia resulted in a reduction in baroreflex sensitivity in both the groups (main effect of time, P<0.01) and responses were lower in resected patients when compared with controls (main effect of group, P<0.05). Hypoglycemia resulted in large reductions in systolic (-17±7 mm Hg) and mean (-14±5 mm Hg) blood pressure in resected patients that were not observed in controls (interaction of group and time, P<0.05). Despite lower blood pressures, increases in heart rate with hypoglycemia were blunted in resected patients (interaction of group and time, P<0.01). Major novel findings from this study demonstrate that intact carotid chemoreceptors are essential for increasing heart rate and maintaining arterial blood pressure during hypoglycemia in humans. These data support a contribution of the carotid chemoreceptors to blood pressure control and highlight the potential widespread effects of carotid body resection in humans.

Retrieval practice over the long term: should spacing be expanding or equal-interval?

If multiple opportunities are available to review to-be-learned material, should a review occur soon after initial study and recur at progressively expanding intervals, or should the reviews occur at equal intervals? Landauer and Bjork (1978) argued for the superiority of expanding intervals, whereas more recent research has often failed to find any advantage. However, these prior studies have generally compared expanding versus equal-interval training within a single session, and have assessed effects only upon a single final test. We argue that a more generally important goal would be to maintain high average performance over a considerable period of training. For the learning of foreign vocabulary spread over four weeks, we found that expanding retrieval practice (i.e., sessions separated by increasing numbers of days) produced recall equivalent to that from equal-interval practice on a final test given eight weeks after training. However, the expanding schedule yielded much higher average recallability over the whole training period.

Maximizing students' retention via spaced review: practical guidance from computational models of memory.

During each school semester, students face an onslaught of material to be learned. Students work hard to achieve initial mastery of the material, but when they move on, the newly learned facts, concepts, and skills degrade in memory. Although both students and educators appreciate that review can help stabilize learning, time constraints result in a trade-off between acquiring new knowledge and preserving old knowledge. To use time efficiently, when should review take place? Experimental studies have shown benefits to long-term retention with spaced study, but little practical advice is available to students and educators about the optimal spacing of study. The dearth of advice is due to the challenge of conducting experimental studies of learning in educational settings, especially where material is introduced in blocks over the time frame of a semester. In this study, we turn to two established models of memory-ACT-R and MCM-to conduct simulation studies exploring the impact of study schedule on long-term retention. Based on the premise of a fixed time each week to review, converging evidence from the two models suggests that an optimal review schedule obtains significant benefits over haphazard (suboptimal) review schedules. Furthermore, we identify two scheduling heuristics that obtain near optimal review performance: (a) review the material from µ-weeks back, and (b) review material whose predicted memory strength is closest to a particular threshold. The former has implications for classroom instruction and the latter for the design of digital tutors.