Editor's Review and Introduction: Cognition-Inspired Artificial Intelligence.

Cognitive science has much to contribute to the general scientific body of knowledge, but it is also a field rife with possibilities for providing background research that can be leveraged by artificial intelligence (AI) developers. In this introduction, we briefly explore the history of AI. We particularly focus on the relationship between AI and cognitive science and introduce this special issue that promotes the method of inspiring AI development with the results of cognitive science research.

Counting Pandemic Statistics Remotely Using Webcams.

OBJECTIVE: Lack of mask use during large public events might spread COVID-19. It is now possible to measure this and similar public health information using publicly available webcams. We demonstrate a rapid assessment approach for measuring mask usage at a public event. METHOD: We monitored crowds at public areas in Sturgis, SD using a live, high-definition, town-sponsored video stream to analyze the prevalence of mask wearing. We developed a rapid coding procedure for mask wearing and analyzed brief (5 to 25 min) video segments to assess mask-wearing compliance in outdoor public areas. We calculated compliance estimates and compared reliability among the human coders. RESULTS: We were able to observe and quantify public behavior on the public streets. This approach rapidly estimated public health information (e.g., 512 people observed over 25 minutes with 2.3% mask usage) available on the same day. Coders produced reliable estimates across a sample of videos for counting masked users and mask-wearing proportion. Our video data is stored in Databrary.org. CONCLUSIONS: This approach has implications for disaster responses and public health. The approach is easy to use, can provide same day results, and can provide public health stakeholders with key information on public behavior.

Cognitive Modeling of Automation Adaptation in a Time Critical Task.

This paper presents a cognitive model that simulates an adaptation process to automation in a time-critical task. The paper uses a simple tracking task (which represents vehicle operation) to reveal how the reliance on automation changes as the success probabilities of the automatic and manual mode vary. The model was developed by using a cognitive architecture, ACT-R (Adaptive Control of Thought-Rational). We also introduce two methods of reinforcement learning: the summation of rewards over time and a gating mechanism. The model performs this task through productions that manage perception and motor control. The utility values of these productions are updated based on rewards in every perception-action cycle. A run of this model simulated the overall trends of the behavioral data such as the performance (tracking accuracy), the auto use ratio, and the number of switches between the two modes, suggesting some validity of the assumptions made in our model. This work shows how combining different paradigms of cognitive modeling can lead to practical representations and solutions to automation and trust in automation.

A Computer-Based Tutor to Teach Nursing Trauma Care That Works as an Adjunct to High-Fidelity Simulation.

Educators are challenged to prepare nurses to care for low-frequency, high-stakes problems such as trauma. Computer-based tutors provide a cost-effective teaching strategy without risking patient safety. Evidence for the efficacy of this type of instruction is limited; thus, we tested the learning outcomes of a tutor on trauma care knowledge with senior nursing students. Participants were randomly assigned to either the tutor or a control condition (textbook learning). Instructional design elements incorporated into the tutor included use of multimedia content, emphasis of key points, frequent quizzing with instant feedback, and unfolding case studies to summarize key concepts. Use of the tutor led to a larger increase in trauma nursing knowledge than use of a textbook. In addition, the knowledge was retained as well as book-based learning. The effect size of the tutor, 1.15, was relatively high as well-the average for computer tutors is 0.79. Qualitative focus groups revealed that participants expressed favorable views of the tutor in comparison to textbook learning. They found it more engaging and more enjoyable and reported that it effectively organized the content. The results of this study support the efficacy of a well-designed computer-based tutor for learning key concepts of trauma nursing.

Effects of Varied Surgical Simulation Training Schedules on Motor-Skill Acquisition.

There have been many studies to evaluate the effect of training schedules on retention; however, these usually compare only 2 drastically different schedules, massed and distributed, and they have tended to look at declarative knowledge tasks. This study examined learning on a laparoscopic surgery simulator using a set of procedural or perceptual-motor tasks with some declarative elements. The study used distributed, massed, and 2 hybrid-training schedules that are neither distributed nor massed. To evaluate the training schedules, 23 participants with no previous laparoscopic experience were recruited and randomly assigned to 1 of the 4 training schedules. They performed 3 laparoscopic training tasks in eight 30-minute learning sessions. We compared how task time decreased with each schedule in a between-participants design. We found participants in all groups demonstrated a decrease in task completion time as the number of training sessions increased; however, there were no statistically significant differences in participants' improvement on task completion time between the 4 different training schedule groups, which suggested that time on task is more important for learning these tasks than the training schedule.

ACT-R: A cognitive architecture for modeling cognition.

ACT-R is a hybrid cognitive architecture. It is comprised of a set of programmable information processing mechanisms that can be used to predict and explain human behavior including cognition and interaction with the environment. We start by reviewing its history, which shapes its current form, contrasts and relates it to other architectures, and helps readers to anticipate where it is going. Based on this history, we then describe it as a theory of cognition that is realized as a computer program. After this, we briefly discuss tools for working with ACT-R, and also note several major accomplishments that have been gained by working with ACT-R in both basic and applied science, including summarizing some of the insights about human behavior. We conclude by discussing its future, which we believe will include adding emotions and physiology, increasing usability, and the use of nongenerative models. This article is categorized under: Computer Science > Artificial Intelligence Psychology > Reasoning and Decision Making Psychology > Theory and Methods.

Performance Trends During Sleep Deprivation on a Tilt-Based Control Task.

BACKGROUND: Understanding human behavior under the effects of sleep deprivation allows for the mitigation of risk due to reduced performance. To further this goal, this study investigated the effects of short-term sleep deprivation using a tilt-based control device and examined whether existing user models accurately predict targeting performance. METHODS: A task in which the user tilts a surface to roll a ball into a target was developed to examine motor performance. A model was built to predict human performance for this task under various levels of sleep deprivation. Every 2 h, 10 subjects completed the task until they reached 24 h of wakefulness. Performance measurements of this task, which were based on Fitts' law, included movement time, task throughput, and time intercept. RESULTS: The model predicted significant performance decrements over the 24-h period with an increase in movement time (R2 = 0.61), a decrease in throughput (R2 = 0.57), and an increase in time intercept (R2 = 0.60). However, it was found that in experimental trials there was no significant change in movement time (R2 = 0.11), throughput (R2 = 0.15), or time intercept (R2 = 0.27). DISCUSSION: The results found were unexpected as performance decrement is frequently reported during sleep deprivation. These findings suggest a reexamination of the initial thought of sleep loss leading to a decrement in all aspects of performance.Bolkovsky JB, Ritter FE, Chon KH, Qin M. Performance trends during sleep deprivation on a tilt-based control task. Aerosp Med Hum Perform. 2018; 89(7):626-633.

IGT-Open: An open-source, computerized version of the Iowa Gambling Task.

The Iowa Gambling Task (IGT) is commonly used to understand the processes involved in decision-making. Though the task was originally run without a computer, using a computerized version of the task has become typical. These computerized versions of the IGT are useful, because they can make the task more standardized across studies and allow for the task to be used in environments where a physical version of the task may be difficult or impossible to use (e.g., while collecting brain imaging data). Though these computerized versions of the IGT have been useful for experimentation, having multiple software implementations of the task could present reliability issues. We present an open-source software version of the Iowa Gambling Task (called IGT-Open) that allows for millisecond visual presentation accuracy and is freely available to be used and modified. This software has been used to collect data from human subjects and also has been used to run model-based simulations with computational process models developed to run in the ACT-R architecture.

Adaptive Virtual Reality Training to Optimize Military Medical Skills Acquisition and Retention.

The Department of Defense has pursued the integration of virtual reality simulation into medical training and applications to fulfill the need to train 100,000 military health care personnel annually. Medical personnel transitions, both when entering an operational area and returning to the civilian theater, are characterized by the need to rapidly reacquire skills that are essential but have decayed through disuse or infrequent use. Improved efficiency in reacquiring such skills is critical to avoid the likelihood of mistakes that may result in mortality and morbidity. We focus here on a study testing a theory of how the skills required for minimally invasive surgery for military surgeons are learned and retained. Our adaptive virtual reality surgical training system will incorporate an intelligent mechanism for tracking performance that will recognize skill deficiencies and generate an optimal adaptive training schedule. Our design is modeling skill acquisition based on a skill retention theory. The complexity of appropriate training tasks is adjusted according to the level of retention and/or surgical experience. Based on preliminary work, our system will improve the capability to interactively assess the level of skills learning and decay, optimizes skill relearning across levels of surgical experience, and positively impact skill maintenance. Our system could eventually reduce mortality and morbidity by providing trainees with the reexperience they need to help make a transition between operating theaters. This article reports some data that will support adaptive tutoring of minimally invasive surgery and similar surgical skills.

Evaluating a range of learning schedules: hybrid training schedules may be as good as or better than distributed practice for some tasks.

We investigated theoretically and empirically a range of training schedules on tasks with three knowledge types: declarative, procedural, and perceptual-motor. We predicted performance for 6435 potential eight-block training schedules with ACT-R's declarative memory equations. Hybrid training schedules (schedules consisting of distributed and massed practice) were predicted to produce better performance than purely distributed or massed training schedules. The results of an empirical study (N = 40) testing four exemplar schedules indicated a more complex picture. There were no statistical differences among the groups in the declarative and procedural tasks. We also found that participants in the hybrid practice groups produced reliably better performance than ones in the distributed practice group for the perceptual-motor task--the results indicate training schedules with some spacing and some intensiveness may lead to better performance, particularly for perceptual-motor tasks, and that tasks with mixed types of knowledge might be better taught with a hybrid schedule. PRACTITIONER SUMMARY: We explored distributed and massed training schedules as well as hybrids between them with respect to three knowledge types based on theories and an empirical study. The results suggest that industrial and operator training in complex tasks need not and probably should not be done on a distributed training schedule.

Caffeine administration does not alter salivary α-amylase activity in young male daily caffeine consumers.

BACKGROUND: To follow up on a recent report from our lab [Hum Psychopharmacol 25:359-367, 2010.] we examined the effects of caffeine on salivary α-amylase (sAA) activity in response to an engaging, non-stressful task in healthy young males (age 18-30 yrs) who consumed caffeine on a daily basis. Using a placebo-controlled, double-blind, between-subjects design, 45 men received either placebo, 200 mg or 400 mg of caffeine (Vivarin®). Participants then rested for 20 minutes, and performed a 20-minute computerized air traffic controller-like task that was cognitively engaging but not stressful. Saliva samples (assayed for sAA and cortisol), blood pressure, and heart rate were taken before (baseline) and 15 minutes after the computerized task. RESULTS: Systolic and diastolic blood pressure and sAA activity increased across the laboratory session (F's > 9.20, p's < 0.05); salivary cortisol levels decreased (F = 16.17, p < 0.05). There were no main effects for caffeine administration on sAA, salivary cortisol, or cardiovascular measures, and caffeine did not interact with the task to alter these measures. CONCLUSIONS: Laboratory administered caffeine does not alter sAA activity, even when sAA activity is stimulated by participating in a cognitively engaging task. These data demonstrate that caffeine administration does not affect sAA activity, at least in healthy young men who regularly consume caffeine. Results support recent findings that basal caffeine levels in habitual caffeine users are not associated with basal sAA activity and that daily caffeine intake and diurnal sAA activity are not related.

Modeling surgical skill learning with cognitive simulation.

We used a cognitive architecture (ACT-R) to explore the procedural learning of surgical tasks and then to understand the process of perceptual motor learning and skill decay in surgical skill performance. The ACT-R cognitive model simulates declarative memory processes during motor learning. In this ongoing study, four surgical tasks (bimanual carrying, peg transfer, needle passing, and suture tying) were performed using the da Vinci© surgical system. Preliminary results revealed that an ACT-R model produced similar learning effects. Cognitive simulation can be used to demonstrate and optimize the perceptual motor learning and skill decay in surgical skill training.

Caffeine and stress alter salivary alpha-amylase activity in young men.

OBJECTIVE: We examined the effects of caffeine and a psychological stressor on salivary alpha-amylase (sAA) in healthy young males (age 18-30 years) who consumed caffeine on a daily basis. METHODS: Using a between-subjects, double-blind, placebo-controlled design, 45 participants received either 200 or 400 mg of caffeine (Vivarin) or placebo, rested for 20 min, and then performed 20 min of mental arithmetic. Saliva samples (assayed for sAA and caffeine), blood pressure, and heart rate were taken before (baseline) and 15 min after the math stressor (stress). RESULTS: Baseline sAA activity did not differ among the treatment groups; however, there was a statistically significant time by caffeine group interaction. Changes in sAA activity across the session were dependent on the amount of caffeine consumed. Following the challenge period, sAA activity among the placebo group was the lowest and sAA activity among the 400 mg treatment group was the highest. Separate repeated-measures ANOVAs conducted for each drug treatment group revealed that sAA activity increased in response to stress and caffeine (i.e., 200 and 400 mg groups) but not to stress alone (i.e., placebo group). CONCLUSIONS: Findings provide evidence for acute sAA changes in response to caffeine and stress in habitual caffeine users.

Modeling how, when, and what is learned in a simple fault-finding task.

We have developed a process model that learns in multiple ways while finding faults in a simple control panel device. The model predicts human participants' learning through its own learning. The model's performance was systematically compared to human learning data, including the time course and specific sequence of learned behaviors. These comparisons show that the model accounts very well for measures such as problem-solving strategy, the relative difficulty of faults, and average fault-finding time. More important, because the model learns and transfers its learning across problems, it also accounts for the faster problem-solving times due to learning when examined across participants, across faults, and across the series of 20 trials on an individual participant basis. The model shows how learning while problem solving can lead to more recognition-based performance, and helps explain how the shape of the learning curve can arise through learning and be modified by differential transfer. Overall, the quality of the correspondence appears to have arisen from procedural, declarative, and episodic learning all taking place within individual problem-solving episodes.

RUI: recording user input from interfaces under Windows and Mac OS X.

Event and timing logs are useful i n studying human-computer interaction,evaluating applications, and comparing input devices. Recording User Input (RUI) is a tool that records user-computer interface behavior. It is created in the .Net framework with C# for Windows and in the Carbon framework for Mac OS X. RUI runs in the background and works with software that runs under Windows or Mac OS X (10.3 Panther and later versions). We illustrate its use with a human-robot interaction interface and present two simple tests that RUI passes and that other timing software should pass: avoiding 0-msec timings and time distributions that follow a gamma (or gamma-like) distribution.

Dismal: a spreadsheet for sequential data analysis and HCI experimentation.

Dismal is a spreadsheet that works within GNU Emacs, a widely available programmable editor. Dismal has three features of particular interest to those who study behavior: (1) the ability to manipulate and align sequential data, (2) an open architecture that allows users to expand it to meet their particular needs, and (3) an instrumented and accessible interface for studies of human-computer interaction (HCI). Example uses of each of these capabilities are provided, including cognitive models that have had their sequential behavior aligned with subject's protocols, extensions useful for teaching and doing HCI design, and studies in which keystroke logs from the timing package in Dismal have been used.