Examining how naturalistic sleep loss modulates performance: interactions of brain networks and social networks
Sleep loss is a well-known, but not well-understood, source of physiological change linked to performance impairment. These performance fluctuations are thought to reflect variability in the underlying physiology that modulates the efficacy and/or efficiency of brain network dynamics that support task performance, but previous research has found significant variability between individuals in their susceptibility to sleep loss. Some of this variability has been accounted for by trait differences measured through questionnaires, but there is burgeoning evidence that a person’s brain network connectivity (both structural and functional) can account for variability between individuals. Furthermore, a person’s social connections and their social network structure has also been shown to account for health outcomes and susceptibility to illness, so we are exploring whether social networks can also mediate impacts of sleep loss on performance. We also have a working hypothesis that social- and emotion-based tasks may be differentially impacted by sleep loss based on overlap in some of the underlying neural mechanisms.
In our work, we are currently analyzing a longitudinal study of naturalistic sleep loss over 16 consecutive weeks where participants continue their normal daily activities and sleep habits, but we measure their sleep history objectively with a wrist actigraph watch and subjectively with standardized sleep diary questions. In this dataset, we measured performance across six different tasks and a resting state task every two weeks, and we collected a host of physiological measurements, including simultaneous measurement of brain activity using common neuroimaging methods (fMRI/EEG), anatomical brain scans for both gray matter and white matter (MPRAGE/dMRI), eyetracking, heartrate, and blood/saliva. Ongoing analyses examine what timescale of sleep history captures state-dependent effects on performance as well as numerous analyses on how this relationship is modulated by individual differences in brain networks, social networks, and physical activity.
Collaborations are desired on (1) this dataset, (2) similar datasets that new collaborative partners may have collected, or (3) a new 16 week longitudinal study that is currently being designed. This research supports ARL’s ERA in Human-Agent Teaming and the ERA in Discovery on Complexity and Emergence.
Developing Novel Group Performance Metrics Using Network Analysis
Communication is an essential component of organizational functionality. Task-oriented groups rely on communication to disseminate necessary information and achieve coordinated task execution. The structure of these communication networks is driven by individual attributes, group attributes, and network-level features such as reciprocity and preferential attachment. From centralized hierarchies to decentralized teams, communication network structure shapes group performance, although no single structural configuration is ideal in all circumstances.
In our research, we determine how the determinants of network structure impact performance in real-world task environments of groups ranging in size from about a dozen to several hundred. We utilize data from training environments to gather metrics of group performance—such as preparation, adaptation, and efficacy—and model the group’s communication network during those events. We model network structure as a function of individual attributes such as past experience, rank, and role and as a function of group attributes such as cohesion and within-group mixing. Comparing within and across groups over these training exercises, we examine the association between group performance and coefficients of identical network models. Our goal is to identify associations between our model results and group outcomes (e.g. do we observe higher performance in networks with more communication across teams than in networks where we observe more communication within teams?).
Collaborators are desired (1) to develop related research questions and work collaboratively using one of our several existing data sets from Army training exercises or (2) to aid in the development of substantively or methodologically motivated approaches for developing novel, network-based performance metrics (particularly for dynamic networks). This research supports ARL’s ERA in Human-Agent Teaming and the ERA in Accelerated Learning for a Ready and Responsive Force.
Dr. Sean M. Fitzhugh, email@example.com, (410) 278-5940
Pervasive biological sensors and ever-improving algorithms have contributed to the proliferation of psychophysiological data that is leading to an ever more quantified self. A limitation of many of these approaches is that singular peripheral physiological measures, such as heart rate, do not map uniquely to putative changes in distinct psychological processes. To address these shortcomings, this research adopts the biopsychosocial model of challenge and threat in order to infer affective motivational states from multivariate physiological states. In particular, peripheral cardiovascular measurement incorporates electrocardiographic, impedance cardiographic, and continuous blood pressure monitoring technologies to infer superordinate, affectively-valenced motivational states. Physiological sensors are placed on participants as they play games, perform cognitive loading and stressful behaviors, and engage in interpersonal tasks with other people or virtual humans. Potential applications of this basic research can inform the design of physiologically aware training interfaces as well as virtual humans and environments, as well as wearable peripheral physiological sensors for the Warfighter. The project seeks collaborators to develop (1) blood flow sensors that are robust to movement artifacts and (2) mappings and representations for how interfaces and virtual humans should respond to changes in psychophysiological states.
Dr. Peter Khooshabeh, firstname.lastname@example.org, (310) 574-7818
Real World Behavior
Understanding long-term participant engagement with data collection systems in the real world
Participants commonly fail to provide the necessary degree of engagement with data collection systems in long-term, real-world studies. Critically, such studies tend to have small sample sizes, so missing data or loss of participants to drop out can have an immensely negative effect on statistical power. Furthermore, low engagement can create a selection process whereby the observed sample may over-represent certain participants—those who are willing or able to engage—which threatens the validity of the experiment. Thus, research focused on long-term, real-world data collection has a pressing need to address the issue of poor participant engagement. The long-term goal of our research program is to meet this need by design and test adaptive interventions that improve engagement with data collection systems. Critically, existing theoretical accounts of when and why participants fail to engage with data collection systems are too vague to guide an adaptive intervention. Thus, the immediate goal of our research program will be to develop detailed computational models of participant engagement. We welcome collaborators who wish to share data and analytical techniques or potentially test adaptive data collection systems in their own long-term, real-world studies.
Dr. Evan Carter, email@example.com, (410) 278-7747
Quantification of Visual Perception in Naturalistic Tasks
Many experiments, specifically those using EEG to study visual perception have been conducted in highly controlled laboratory settings, requiring participants to constrain their eyes to a fixed location. A major limitation of this approach is that controlled stimulus presentation to a prescribed location provides an inadequate representation of how we process visual information in our day-to-day interactions. Numerous changes in the brain, specifically in the visual system, occur prior and subsequent to each eye movement. These changes cannot be adequately represented in typical fixed eye position paradigms. Therefore, in order to best understand visual perception in real world tasks, neurophysiological experiments in particular must leverage rather than limit eye movements. This can be accomplished by using eye-tracking data to determine the onset and offset of saccades thereby providing unique events for time-locking in the EEG data and creating saccade and fixation-locked potentials (S/FRPs). In addition to mobile eye tracking, we use the FRP methodology to isolate processes involved in visual perception that are absent in fixation-constrained passive viewing in order to provide a more holistic understanding of visual search in real-world scenarios.
Our current research project encompasses three experiments. Experiment 1 uses controlled visual stimuli to characterize the relationship between high dynamic range luminance, spatial frequency and saccade magnitude on the fixation-related brain response. The second experiment examines the temporal profile of the cortical enhancement/suppression following a saccadic eye movement. Experiment 3 is designed to establish how bottom-up and top-down biases affect attentional allocation in real world search tasks. We seek collaborative research partners with expertise in: 1) mobile eye tracking data acquisition and analysis and 2) visual search behavior for multiple camouflaged targets.
Dr. Anthony Ries, firstname.lastname@example.org, (410) 278-0915
Visual search and recognition under real-world dynamics
Machine learning algorithms for visual recognition and human-agent teaming are typically trained on laboratory stimuli with limited dynamic range (e.g. ~100:1 contrast ratio, for targets and distractors with similar likelihoods of appearance). In the real-world, search algorithms must be able to predict naturalistic behavior under high-dynamic range inputs (e.g. up to 1,000,000:1 luminance contrast ratio) and cannot miss rare events (e.g. in baggage screening). We are conducting visual search experiments to characterize the behavioral and neurophysiological mechanisms underlying normalization to real-world dynamics. We are seeking collaborators with behavioral, neurophysiological, or imaging data or computational / machine learning expertise that can help us to advance and broaden the real-world generalizability of our model of visual recognition, with the broader goal of improving the real-world robustness of intelligent agents and autonomous robots.
Dr. Chou Hung, email@example.com, (410) 278-9255
Effects of Cognitive Stress on Physical and Cognitive performance
Our goal is to understand the interactive effects of physical and cognitive stress on individual Soldiers and small teams. More specifically, this area of research is focused primarily on the effects of cognitive stress on physical performance and the performance of operational tasks. Recently completed projects have addressed how prior cognitive fatigue influences subsequent performance in a full body discontinuous exercise task (gross-motor activity) and response inhibition in marksmanship performance (fine-motor activity). Future investigations will continue to explore how Warfighter performance is influenced by the interaction between cognitive and physical fatigue on a continuum ranging from a controlled laboratory setting to more ecologically valid scenarios. To accomplish this goal, we will continue to utilize our state-of-the-art 3000 square foot biomechanics laboratory, obstacle course equipped with timing gates, 2.5 mile Wi-Fi enabled cross country course, and our live fire marksmanship range. We welcome collaborators with expertise in autonomic research and interest in understanding the interaction between physical and cognitive performance.
Dr. James R. Head, firstname.lastname@example.org, (410) 278-8574