Introduction to the Emerging Cognitive Science of Distributed Human-Autonomy Teams.

Teams are a fundamental aspect of life-from sports to business, to defense, to science, to education. While the cognitive sciences tend to focus on information processing within individuals, others have argued that teams are also capable of demonstrating cognitive capacities similar to humans, such as skill acquisition and forgetting (cf., Cooke, Gorman, Myers, & Duran, 2013; Fiore et al., 2010). As artificially intelligent and autonomous systems improve in their ability to learn, reason, interact, and coordinate with human teammates combined with the observation that teams can express cognitive capacities typically seen in individuals, a cognitive science of teams is emerging. Consequently, new questions are being asked about teams regarding teamness, trust, the introduction and effects of autonomous systems on teams, and how best to measure team behavior and phenomena. In this topic, four facets of human-autonomy team cognition are introduced with leaders in the field providing in-depth articles associated with one or more of the facets: (1) defining teams; (2) how trust is established, maintained, and repaired when broken; (3) autonomous systems operating as teammates; and (4) metrics for evaluating team cognition across communication, coordination, and performance.

Profiling cognitive workload in an unmanned vehicle control task with cognitive models and physiological metrics.

In the present study, we use Cognitive Metrics Profiling (CMP) to capture variance in cognitive load within a complex unmanned vehicle control task. We aim to demonstrate convergent validity with existing workload measurement methods, and to decompose workload into constituent cognitive resources to aid in diagnosing causes of workload. A cognitive model of the task was developed and examined to determine the extent to which it could predict behavioral performance, subjective workload, and validated physiological workload metrics. We also examined model activity to draw insights regarding loaded cognitive capacities. We found that composite workload from the model predicted physiological metrics, performance, and subjective workload. Moreover, the model indicates that differences in workload were driven largely by procedural, declarative, and temporal memory demands. We have found preliminary evidence of correspondence between workload predictions of a CMP model and physiological measures of workload. This suggests our approach captures interesting aspects of workload in a complex task environment and may provide a theoretical link between behavioral, physiological, and subjective metrics. This approach may provide a means to design effective workload mitigation interventions and improve decision-making about personnel tasking and automation.

Physiocognitive Modeling: Explaining the Effects of Caffeine on Fatigue.

Most computational theories of cognition lack a representation of physiology. Understanding the cognitive effects of compounds present in the environment is important for explaining and predicting changes in cognition and behavior given exposure to toxins, pharmaceuticals, or the deprivation of critical compounds like oxygen. This research integrates physiologically based pharmacokinetic (PBPK) model predictions of caffeine concentrations in blood and tissues with ACT-R's fatigue module to predict the effects of caffeine on fatigue. Mapping between the PBPK model parameters and ACT-R model parameters is informed by the neurophysiological literature and established associations between ACT-R modules and brain regions. The results from three such parameter mappings are explored to explain observed data from sleep-deprived participants performing the psychomotor vigilance test with and without caffeine. Predicted caffeine concentrations in the brain are used to modulate procedural parameters in the fatigue module to explain caffeine's effects on multiple performance metrics.

Knowledge Gaps: A Challenge for Agent-Based Automatic Task Completion.

The study of human cognition and the study of artificial intelligence (AI) have a symbiotic relationship, with advancements in one field often informing or creating new work in the other. Human cognition has many capabilities modern AI systems cannot compete with. One such capability is the detection, identification, and resolution of knowledge gaps (KGs). Using these capabilities as inspiration, we examine how to incorporate detection, identification, and resolution of KGs in artificial agents. We present a paradigm that enables research on the understanding of KGs for visual-linguistic communication. We leverage and enhance and existing KG taxonomy to identify possible KGs that can occur for visual question answer (VQA) tasks and use these findings to develop a classifier to identify questions that could be engineered to contain specific KG types for other VQA datasets. Additionally, we examine the performance of different VQA models through the lens of KGs.

Cognition-Enhanced Machine Learning for Better Predictions with Limited Data.

The fields of machine learning (ML) and cognitive science have developed complementary approaches to computationally modeling human behavior. ML's primary concern is maximizing prediction accuracy; cognitive science's primary concern is explaining the underlying mechanisms. Cross-talk between these disciplines is limited, likely because the tasks and goals usually differ. The domain of e-learning and knowledge acquisition constitutes a fruitful intersection for the two fields' methodologies to be integrated because accurately tracking learning and forgetting over time and predicting future performance based on learning histories are central to developing effective, personalized learning tools. Here, we show how a state-of-the-art ML model can be enhanced by incorporating insights from a cognitive model of human memory. This was done by exploiting the predictive performance equation's (PPE) narrow but highly specialized domain knowledge with regard to the temporal dynamics of learning and forgetting. Specifically, the PPE was used to engineer timing-related input features for a gradient-boosted decision trees (GBDT) model. The resulting PPE-enhanced GBDT outperformed the default GBDT, especially under conditions in which limited data were available for training. Results suggest that integrating cognitive and ML models could be particularly productive if the available data are too high-dimensional to be explained by a cognitive model but not sufficiently large to effectively train a modern ML algorithm. Here, the cognitive model's insights pertaining to only one aspect of the data were enough to jump-start the ML model's ability to make predictions-a finding that holds promise for future explorations.

Editors' Introduction: Best Papers from the 18th International Conference on Cognitive Modeling.

The International Conference on Cognitive Modeling brings together researchers from around the world whose main goal is to build computational systems that reflect the internal processes of the mind. In this issue, we present the four best representative papers on this work from our 18th meeting, ICCM 2020, which was also the first meeting to be held virtually. Two of these papers develop novel techniques for building larger and more complex models using Reinforcement Learning and Learning By Instruction, respectively. The other two show how cognitive models connect to neuroscience, drawing on details of the hippocampus and cerebellum to constrain and explain the cognitive processes involved in memory and conditioning.

How Packaging of Information in Conversation Is Impacted by Communication Medium and Restrictions.

In team-based tasks, successful communication and mutual understanding are essential to facilitate team coordination and performance. It is well-established that an important component of human conversation (whether in speech, text, or any medium) is the maintenance of common ground. Maintaining common ground has a number of associated processes in which conversational participants engage. Many of these processes are lacking in current synthetic teammates, and it is unknown to what extent this lack of capabilities affects their ability to contribute during team-based tasks. We focused our research on how teams package information within a conversation, by which we mean specifically (1) whether information is explicitly mentioned or implied, and (2) how multiple pieces of information are ordered both within single communications and across multiple communications. We re-analyzed data collected from a simulated remotely-piloted aerial system (RPAS) task in which team members had to specify speed, altitude, and radius restrictions. The data came from three experiments: the "speech" experiment, the "text" experiment, and the "evaluation" experiment (which had a condition that included a synthetic teammate). We asked first whether teams settled on a specific routine for communicating the speed, altitude, and radius restrictions, and whether this process was different if the teams communicated in speech compared to text. We then asked how receiving special communication instructions in the evaluation experiment impacted the way the human teammates package information. We found that teams communicating in either speech or text tended to use a particular order for mentioning the speed, altitude, and radius. Different teams also chose different orders from one another. The teams in the evaluation experiment, however, showed unnaturally little variability in their information ordering and were also more likely to explicitly mention all restrictions even when they did not apply. Teams in the speech and text experiments were more likely to leave unnecessary restrictions unmentioned, and were also more likely to convey the restrictions across multiple communications. The option to converge on different packaging routines may have contributed to improved performance in the text experiment compared some of the conditions in the evaluation experiment.

Gorham Disease Limited to the Left Upper Extremity Without Hand Involvement.

In this case report, we present a young female patient with a history of Gorham disease, who sustained pathologic fractures of the left radius and ulna after a low-impact fall. Massive osteolysis of the left forearm and wrist was noted on plain radiographs. The patient had had 8 previous left upper-extremity fractures without evidence of disease in any other area of the body.

Visual Working Memory Resources Are Best Characterized as Dynamic, Quantifiable Mnemonic Traces.

Visual working memory (VWM) is a construct hypothesized to store a small amount of accurate perceptual information that can be brought to bear on a task. Much research concerns the construct's capacity and the precision of the information stored. Two prominent theories of VWM representation have emerged: slot-based and continuous-resource mechanisms. Prior modeling work suggests that a continuous resource that varies over trials with variable capacity and a potential to make localization errors best accounts for the empirical data. Questions remain regarding the variability in VWM capacity and precision. Using a novel eye-tracking paradigm, we demonstrate that VWM facilitates search and exhibits effects of fixation frequency and recency, particularly for prior targets. Whereas slot-based memory models cannot account for the human data, a novel continuous-resource model does capture the behavioral and eye tracking data, and identifies the relevant resource as item activation.

Model flexibility analysis.

A good fit of model predictions to empirical data are often used as an argument for model validity. However, if the model is flexible enough to fit a large proportion of potential empirical outcomes, finding a good fit becomes less meaningful. We propose a method for estimating the proportion of potential empirical outcomes that the model can fit: Model Flexibility Analysis (MFA). MFA aids model evaluation by providing a metric for gauging the persuasiveness of a given fit. We demonstrate that MFA can be more informative than merely discounting the fit by the number of free parameters in the model, and show how the number of free parameters does not necessarily correlate with the flexibility of the model. Additionally, we contrast MFA with other flexibility assessment techniques, including Parameter Space Partitioning, Model Mimicry, Minimum Description Length, and Prior Predictive Evaluation. Finally, we provide examples of how MFA can help to inform modeling results and discuss a variety of issues relating to the use of MFA in model validation. (PsycINFO Database Record

SAwSu: an integrated model of associative and reinforcement learning.

Successfully explaining and replicating the complexity and generality of human and animal learning will require the integration of a variety of learning mechanisms. Here, we introduce a computational model which integrates associative learning (AL) and reinforcement learning (RL). We contrast the integrated model with standalone AL and RL models in three simulation studies. First, a synthetic grid-navigation task is employed to highlight performance advantages for the integrated model in an environment where the reward structure is both diverse and dynamic. The second and third simulations contrast the performances of the three models in behavioral experiments, demonstrating advantages for the integrated model in accounting for behavioral data.

Interactive team cognition.

Cognition in work teams has been predominantly understood and explained in terms of shared cognition with a focus on the similarity of static knowledge structures across individual team members. Inspired by the current zeitgeist in cognitive science, as well as by empirical data and pragmatic concerns, we offer an alternative theory of team cognition. Interactive Team Cognition (ITC) theory posits that (1) team cognition is an activity, not a property or a product; (2) team cognition should be measured and studied at the team level; and (3) team cognition is inextricably tied to context. There are implications of ITC for theory building, modeling, measurement, and applications that make teams more effective performers.

Visual scan adaptation during repeated visual search.

There is no consensus as to how to characterize eye fixations during visual search. On the one hand, J. M. Wolfe, G. A. Alvarez, and T. S. Horowitz (2000) have described them as a haphazard sequence of fixations. On the other hand is research that shows systematic repetition of visual patterns when freely viewing a scene (T. Foulsham & G. Underwood, 2008; D. Noton & L. W. Stark, 1971a). Two experiments are reported that demonstrate the repetition and adaptation of visual scans during visual search, supporting an adaptive scanning hypothesis. When trials were repeated in a simple search task, visual scan similarity and search efficiency increased. These increments in similarity and efficiency demonstrate the systematic and adaptive nature of visual scans to the characteristics of the visual environment during search.

ProtoMatch: a tool for analyzing high-density, sequential eye gaze and cursor protocols.

ProtoMatch is a software tool for integrating and analyzing fixed-location and movement eye gaze and cursor data. It provides a comprehensive collection of protocol analysis tools that support sequential data analyses for eye fixations and scanpaths as well as for cursor "fixations" (dwells at one location) and "cursorpaths" (movements between locations). ProtoMatch is modularized software that integrates both eye gaze and cursor protocols into a unified stream of data and provides an assortment of filters and analyses. ProtoMatch subsumes basic analyses (i.e., fixation duration, number of fixations, etc.) and introduces a method of objectively computing the similarity between scanpaths or cursorpaths using sequence alignment. The combination of filters, basic analyses, and sequence alignment in ProtoMatch provides researchers with a versatile system for performing both confirmatory and exploratory sequential data analyses (Sanderson & Fisher, 1994).

Carotid distensibility characterized via the isometric exercise pressor response.

Distensibility of the large elastic arteries is a key index for cardiovascular health. Distensibility, usually estimated from resting values in humans, is not a static characteristic but a negative curvilinear function of pressure. We hypothesized that differences in vascular function with gender and age may only be recognized if distensibility is quantified over a range of pressures. We used isometric handgrip exercise to induce progressive increases in pressures and carotid diameters, thereby enhancing the characterization of distensibility. In 30 volunteers, evenly distributed by gender and age across the third to fifth decades of life, we derived pulsatile distensibility slopes as a function of arterial pressure for a dynamic distensibility index and compared it with a traditional static index at a reference pressure of 95 mmHg. We also assessed intima-media thickness (IMT). We found that women had greater distensibility slopes within each decade, despite comparable IMT. Furthermore, declines in distensibility slope with increasing age were correlated to increased IMT. The static distensibility index failed to show gender-related differences in distensibility but did show age-related differences. Our results indicate that gender- and age-related differences can be manifest even in young, healthy adults and may only be identified with techniques that assess carotid distensibility across a range of pressures.