Keywords

1 Introduction

1.1 Challenges in Optimizing Military Training

The military training community is faced with the daunting task of providing each and every warfighter with basic, journeyman and advanced training courses – using media and methodologies that permit rapid, efficient learning and transfer of the learning to a wide range of operational tasks. For example, the navy has expressed the need for each sailor to improve his or her training efficiency and learning retention for complex naval tasking by a factor of two times or more. Therefore, there is a need to discover accelerated learning techniques through valid and reliable metrics and the comparison of human behavioral techniques that support transfer of training to the complex operational environment. Optimized training requires a notable change in the learning paradigm – a training paradigm in which the training system assists the learner in an effort to enter and maintain an optimal cognitive state for learning. Such a state should provide for enhanced skill acquisition and more rapid retrieval of information from memory, as well as the ability to sustain focused attention for longer periods of time and deeper levels of information processing (Wickens and Hollands 2000).

The problem faced is multifaceted in that accelerated learning pedagogy has not been explicated for complex training systems such as those found in the military. Further, optimal cognitive states vary across learning and performance domains, as well as within and across individuals. Even for the same task, optimal states have been shown to differ across individuals. For example, it has been shown that some individuals perform best when in higher states of arousal than others. Studies of accelerated learning are also complicated by findings by training experts that trainees often do not arrive at formal training sessions prepared to learn. Trainees may arrive in various states of arousal, fatigue, and anxiety based upon events of their personal and professional lives (e.g. marital problems, financial issues, multiple conflicting demands at work), leading to distraction and difficulty in assimilating complex training content. Further, trainees with less ambient arousal and anxiety are more likely to successfully focus and absorb the material presented. Research has shown that trainees who enter training with a negative brain state exhibit sub-par performance and learning as evidence by the measured modulation of encoding and retrieval processing within their memory systems (Margraf and Zlomuzica 2015). However, most individuals are not aware when their minds are closed to learning, and a suboptimal trainee learning state may not be apparent to an instructor.

1.2 Optimal Cognitive States for Learning

Research in the field of Neuroergonomics (the study of how the brain performs in operational environments), indicates that the type of sustained attention necessary for skill learning and safety monitoring while engaged in difficult operational tasks (Parasuraman 2000) is enhanced through techniques such as mindfulness. Self-regulation and self-management of the brain networks that underlie mindfulness lead to enhanced brain connections necessary for improving data processing and memory retrieval (Shapiro et al. 2006) and enhanced executive control over inflexible biased response (Teper and Inzlicht 2013). Thus, for deeper, accelerated learning of complex occupational skills, military personnel need to be engaged, involved, and mindful before and during the training process. The ability to improve engagement during training holds the potential to increase the transfer of training, leading to enhanced readiness and success on the across a wide range of military operational domains and tasks.

Current research also suggests that the external (learning material) and the internal (mental states) aspects of a learner should be in balance or in predictable order for optimal learning to occur (Csíkszentmihályi 2014, p.211). Csíkszentmihályi (2014) has further characterized this optimal learning state as a “flow” state - a state of heightened focus and immersion in activities such as art, play and work. Mindfulness may act to clear and awaken one’s mind, leading to reduced anxiety and improved access to attentional networks. Entering flow state may then be enhanced increasing focused attention and heightening information processing in that domain. To maintain an optimal state of learning, the training should have clear tasks, optimal challenges, and provide clear and immediate feedback. In addition, the training must match the learner’s current knowledge state and simultaneously challenge the trainee through a cooperative learning strategy such that he or she can perform the tasks alone with insight to problem solving beyond the formal instruction (Vygotsky 1987).

Thus, there is a need for a methodology and metrics to assess the best combinations of learning techniques that can be applied across various types of military training systems and a training testbed with which to assess individual and group characteristics that can accelerate the speed of learning, increase comprehension and retention, and improve transfer of training to performance on operational tasks.

1.3 Adaptive Training and Augmented Cognition

Vygotsky (1987) described the concept of the zone of proximal development (ZPD) as “the distance between the actual developmental level as determined by independent problem solving and the level of potential development as determined through problem solving under adult guidance, or in collaboration with more capable peers” (p. 86). Essentially, the ZPD describes what one needs to learn with assistance of an expert, such as in a train-the-trainer paradigm, or what the learner can do without assistance. ZPD requires monitoring of in-situ performance and providing assistance or feedback from an expert trainer or a team of peers, either simulated or live. Metrics for assessing mindfulness and monitoring flow currently consist of self-report questionnaires and interviews that are non-invasive and inexpensive; however, memory bias and task interruption affect the reliability and validity of these methods. To demonstrate efficient and effective methods that increase the sailor’s learning speed, comprehension, and performance requires an ability to measure cognitive and affective states of the learner throughout the training experience. Therefore, there is a need to discover accelerated learning techniques through valid and reliable metrics and the comparison of human behavioral techniques that support transfer of training to the complex operational environment.

In addition to subjective measures of flow such as the 36-item Flow State Scale (FSS) (Jackson and Marsh 1996), several more recent research efforts have been undertaken to use psychophysiological measures to detect and characterize flow state using a wide range of sensor technologies (Berta et al. 2013). One such study demonstrated that psychophysiological data and pupil dilation characteristics were significantly different while using Facebook, an activity thought to be highly engaging over extended periods of time, as compared to induced stress and relaxation conditions on multiple linear and spectral indices of somatic activity. The psychophysiological state evoked in the Facebook condition was characterized by high positive valence and high arousal, which have been used to define Core Flow State (Mauri et al. 2011). Research has also indicated that flow experiences may combine subjectively positive elements with physiological elements associated with strainful tension and mental load based on heart rate variability and salivary cortisol (Keller et al. 2011). Similary, Nacke and Lindley (2008) demonstrated video gameplay scenarios designed for combat-oriented flow experiences demonstrated measurable high-arousal positive affect emotions, which were consistent across both subjective (questionnaire responses) and objective measures collected (electroencephalography, electrocardiography, electromyography, galvanic skin response and eye tracking).

Hou and Fidopiastis (2017) developed an adaptive intelligent tutoring methodology for accelerating learning for unmanned system operators that extended the research and development of the Defense Advanced Research Projects Agency’s (DARPA) Augmented Cognition (AugCog) program. The purpose of the AugCog training architecture was to evaluate best practices for designing interfaces that facilitated trainee learning by maintaining the trainee in the optimal state of learning (Nicholson et al. 2007; Hou and Fidopiastis 2017). The successful mapping of psychophysiological human state changes in adaptive tutoring systems has been demonstrated by the AugCog research community (St. John et al. 2003; Palmer and Kobus 2007). Specifically, cognitive and affective state changes of the learner are measurable using psychophysical metrics such as electroencephalography (EEG), heart rate variability (HRV), and electrodermal response (EDR) to identify cognitive workload (Sciarini and Nicholson 2009), engagement (Berka et al. 2007), and negative emotional states (Vartak et al. 2008). However, for proper effectiveness or training transfer to the field environment, the training system must adapt the learning content base on the skill level of the learner and optimize the learner states of engagement, performance, cognitive workload, and affect throughout the training experience (Hou and Fidopiastis 2014, 2017).

1.4 Accelerated Learning Operational Definition and Application

Accelerated Learning has been operationally defined as “The reduction of learner time required to meet learning objectives in a training event (Hoffman et al. 2010, p. 400)”. According to Hoffman et al. accelerated learning is comprised of two primary components: accelerating the learning pathway and accelerating the learning process. The former involves enabling the learner to cover and master material in less time, while the latter involves employing instructional design principles that increase learner engagement with the material. The remaining sections of this paper present a theoretical approach and initial instantiation of a method by which to apply and operationalize this definition of accelerated learning within the context of a military training task.

2 Curriculum for Accelerated Learning Through Mindfulness (CALM)

2.1 CALM Theoretical Model and Testbed System

The Curriculum for Accelerated Learning through Mindfulness (CALM) theoretical model seeks to combine the two components of accelerated learning as defined by Hoffman et al. (2010): accelerating the learning pathway and accelerating the learning process. This model has been instantiated within a prototype testbed system involving a mindfulness intervention that prepares the sailor for learning and a methodology to assess the flow state of the learner throughout a contextually relevant Naval learning experience using psychophysiological measures. This model is based on the premise that a trainee who is mindful (engages in focused attention) and in the flow (meeting the learning material with appropriate cognitive challenge and skill level) will maintain optimal cognitive and affective states that prepare his brain to learn more effectively and efficiently, produce higher and longer retention of the training material, and facilitate transfer of that training to the operational environment.

Accelerating the Learning Pathway.

This component of accelerated learning involves enabling the learner to progress through and master material in less time. Within the CALM theoretical model this is achieved via a paradigm that uses an unconventional application of Item Response Theory (IRT) to drive an adaptive trainer that automatically determines the appropriate training and testing level of difficulty for a student based on his or her in-situ measured ability level. This component of the CALM testbed consists of an existing adaptive training system, the Adaptive Gaming Environment for Submarines (AGE-S), which provides computer-based Electronic Warfare (EW) Support (ES) operator proficiency instruction and assessment relevant to Electronic Support Measures (ESM).

Each training module within the AGE-S system contains a set of instructional materials, which consist of text, pictures, sound files, and short movie clips to convey the basic concepts underlying the skills to be taught. Following the instructional material, the module contains several groupings of questions. Each question consists of an EW scenario captured from an Advanced Submarine Tactical ESM Combat System (ASTECS) emulator either as a static picture, a sound file, or a movie file, plus a multiple choice question to go with the scenario. The groups of questions range from very simple, covering the most basic concepts, to intermediate difficulty, where part-task skills are tested and the number and complexity of sonar emitters is increasing, to the most difficult questions, which contain movies and complicated scenarios that closely parallel whole-task skill testing. Questions are binned into five levels of difficulty (Easiest, Easy, Medium, Hard, Hardest). The adaptive engine measures a student’s performance on the questions and determines whether to provide questions of the same, easier, or harder level of difficulty. Each time a question is answered incorrectly, process feedback is provided in the form of a brief video clip containing step by step instructions, similar to a worked example, detailing the way in which the trainee should have completed the exercise to arrive at the right answer. When a question is answered correctly, the system indicates that the right answer was provided, and the next question is immediately presented, with no additional feedback.

As such, adept students will move quickly through the material, without pausing for additional instruction or feedback, progressing to increasingly harder questions and completing the module in less time, while less proficient students will take longer to progress through the course material, but will receive the extra instruction they need along the way to successfully complete the module.

Accelerating the Learning Process.

This component of accelerated learning involves employing instructional design principles that increase learner engagement with the material. Within the CALM theoretical model, during training content development and validation, this is achieved by using psychophysiological measures to objectively assess cognitive engagement across multiple individuals while interacting with instructional and assessment materials having varying levels of difficulty. This component of the CALM prototype system consists of a Testbed for Intelligent Tracking and Real-time Assessment of Trainee Engagement (TITRATE), comprising a physiological sensor suite and validated analytic algorithms for evaluating a subject’s cognitive state in real-time while completing AGE-S assessment questions across all five levels of difficulty. These algorithms emphasize detection of optimal levels of engagement for a particular individual using a comparison to baseline and comparison to performance outcomes methodology. The current TITRATE prototype system uses the B-Alerttm X10 (Advanced Brain Monitoring, Carlsbad, CA) mobile sensor system, a commercial off the shelf (COTS) technology that provides nine channels of high-quality EEG, plus one optional channel for Electrocardiography (ECG), Electromyography (EMG), or Electrooculography (EOG). The current prototype also uses a previously developed and validated engagement metric, which uses discriminant function analysis (DFA) methods to derive a four-class quadratic DFA model to distinguish high engagement, low engagement, distraction, and sleep onset classifiers.

Closing the Loop.

Additionally, the CALM theoretical model advances current training paradigms by integrating a mindfulness module (MindMod) that prepares the trainee for learning or assists the trainee in returning to an optimized cognitive state for learning when engagement levels become too low. The MindMod component of the CALM prototype system consists of a series of mindfulness exercises, including guided meditation, which can be selected based on trainee preference, and which will be objectively assessed for efficacy under future research.

Finally, the envisioned CALM testbed system will include an Adaptive Driver to Augment Performance and Training (ADAPT) Application Programming Interface (API) that enables real-time correlation of training performance to cognitive state metrics and subsequent adaptation of training content (e.g., complexity/difficulty, modality, scaffolding) in order to maintain an optimal and accelerated state of learning.

2.2 CALM System Proof of Concept Functionality Testing

Proof of concept functionality testing was conducted in order to verify the ability of the prototype testbed system to (1) collect clean and complete physiological sensor data sets; (2) accurately synchronize physiological data timestamps to task related timestamps, task events, and relevant contextual information (e.g., start and stop of each question, question level of difficulty, answer accuracy); (3) assess the sensitivity of the TITRATE sensors and algorithms with respect to detection of potential changes in user engagement while completing the AGE-S module questions at varying levels of difficulty; (4) assess the sensitivity of the TITRATE sensors and algorithms with respect to detection of potential differences in cognitive state between a regular meditator and a novice meditator; and (5) assess the sensitivity of the TITRATE sensors and algorithms with respect to detection of potential changes in cognitive state before and after a brief meditation session. Functionality testing data were collected for two individuals, one who has maintained a dedicated meditation practice over many years and one having no prior meditation experience. Both individuals, having had prior training and experience in using the AGE-S training system, completed a total of 100 AGE-S questions, across all five levels of difficulty, as well as a five-minute meditation session half-way through the testing session. The 100 questions were broken into 10 blocks, each with a specified level of difficulty, ranging from Easiest to Hardest. As shown in Fig. 1, each participant completed 10 Easiest, 10 Easy, 10 Medium, 10 Hard, and 10 Hardest questions, followed by the five-minute meditation (indicated by the green arrow as question #51). Both participants then completed 10 Hardest, 10 Hard, 10 Medium, 10 Easy, and 10 Easiest questions. EEG and ECG were collected for both participants throughout the testing and meditation sessions.

Fig. 1.
figure 1

Mean engagement state for the meditator and non-meditator across all 100 questions and the five-minute meditation session (represented as Question #51). (Color figure online)

Full size image

Figure 1 provides average estimates of the highest probability EEG-based engagement state for the meditator (represented in blue) and the non-meditator (represented in red) across 100 AGE-S questions answered the five-minute meditation session (indicated with a green arrow and represented as Question #51). The engagement algorithm estimates the mostly likely state of the user, selected from Sleep Onset (SO), Distraction (DIS), Low Engagement (LE), or High Engagement (HE) on a second by second basis. The algorithm indicates the most likely state for each second with a specified value assigned to each state as follows: 0.1 (SO), 0.3 (DISS), 0.6 (LE), or 0.9 (HE). Given that each question took longer than one second to answer, and the times to complete each question varied, multiple state estimations were recorded for each question, and the estimated states may have changed over the course of each question. Initial analyses first evaluated the average estimated state over the course of each question and the five-minute meditation session. As such, the plotted points in Fig. 1 do not fall strictly at the specified level of a specific engagement state, but rather indicate an average level across the four potential states for each question and during meditation.

In addition to displaying the engagement probability for each question answered, as well as during the five-minute meditation, Fig. 1 also indicates whether each question was answered correctly or incorrectly. The circular symbol on the line graph for each question is filled in if the question was answered correctly; if the symbol is empty, this indicates that the question was answered incorrectly.

The regular meditator (shown in blue), performed very well on the first 50 questions, answering only one incorrectly in the Easiest block, one incorrectly in the Easy block, two incorrectly in the Medium difficulty block, two incorrectly in the Hard block, and two incorrectly in the Hardest block. Upon beginning the second block of Hardest difficulty questions (starting with Question #52), immediately following the five-minute meditation, the regular meditator answered three questions incorrectly. In the subsequent Hard difficulty question block, he answered four incorrectly, followed by just one incorrect answer in the Medium difficulty question block, one incorrectly in the Easy block, and no incorrect answers in the final Easiest block of questions.

The non-meditator performed well on the first block of Easiest questions, getting only one incorrect, and on the Easy questions, getting two incorrect. However, he began to struggle more with the Medium questions, getting three incorrect, and then really struggled with the Hard and Hardest difficulty question blocks, answering seven and eight incorrect, respectively. Following the meditation session, the non-meditator continued to struggle, answering nine of the second block of Hardest questions incorrectly and seven of the Hard questions incorrectly. He then began to recover in performance during the Medium question block, answering only two incorrectly, and then maintained good performance throughout the remainder of the questions, answering just two Easy questions incorrectly, and none of the final block of Easiest questions incorrectly.

While, on average, both individuals remained somewhere between Low Engagement and High Engagement throughout the testing and meditation session, the meditator maintained consistently higher engagement levels than the non-meditator throughout all the question blocks, which may have contributed to better performance. For some questions, the meditator’s average engagement value was 0.9, indicating that he was in a state of High Engagement the entire time he was answering that question (e.g., Questions #21, #65, #90, and #97). Notably, all four of those questions were answered correctly.

Interestingly, the meditator displayed markedly lower engagement levels during the five-minute meditation session (indicated as Question #51). This result is to be expected with a highly practiced meditator, who is likely able to go into very deep states of consciousness very quickly; those skilled in meditative practices are extremely good at disengaging from active thought processes and unintended distracting thoughts, even within a short meditation session. During longer sessions, these deeper states are very similar to sleep onset from an engagement perspective, and while experienced meditators can get into these states fairly quickly, practiced meditators typically bring themselves out of those states slowly at the end of a meditation session. In this case, immediately following the short (five-minute) meditation, the regular meditator returned to pre-meditation engagement level, but performed poorly much more poorly on the Hardest block of questions as compared to his performance on the Hardest block preceding meditation. This may be an indication that he went into an extremely low engagement state during meditation and was still coming out of that state when attempting to answer the questions in the next block. This is to be expected with a highly practiced meditator.

The non-meditator not only exhibited lower engagement levels overall, for some questions, his average engagement state fell below the threshold for Low Engagement of 0.6, indicating that during at least part of the time while answering those questions, his highest probability cognitive state was Distraction (e.g., Questions #47, #81, #83, #88, and #89). The non-meditator’s average engagement level during the meditation was markedly higher than the regular meditators, remaining in an average state of Low Engagement throughout the meditation exercise. This is typical for individuals inexperienced in meditation, who may find it very difficult to clear their minds, reducing focused attention and dismissing distracting thoughts.

In order to better represent the highest probability engagement state (i.e., the state that was most prevalent for each question), the mode value is presented in Fig. 2 for each question, as well as the mode value for the five-minute meditation session (again shown as question #51).

Fig. 2.
figure 2

Mode cognitive state (level of engagement) for the experienced meditator and non-meditator across all 100 questions and the five-minute meditation session (represented as Question #51).

Full size image

Figure 2 provides a clearer picture of the cognitive state levels for both the meditator and non-meditator across the entire data collection session. The regular meditator was in a state of High Engagement for 95 of the 100 questions, dropping to Low Engagement for the remaining five questions, but only dropping below Low Engagement during meditation. Conversely, the non-meditator was only in a state of High Engagement for 62 of the 100 questions, dropping to a state of Low Engagement for 37 questions; and furthermore, the non-meditator did not drop below Low Engagement during meditation, but did drop down into the Distraction state on Question #80.

3 Conclusions

Taken together, these data results indicate very clear differences between the two individuals, with the regular meditator maintaining higher engagement overall and better performance overall, particularly for the harder difficulty questions before the meditation session. It is possible that the non-meditator performed more poorly due to lower engagement, but it is also possible that the meditator simply has greater knowledge of the subject matter and the AGE-S question types. Most interestingly, the two participants demonstrated very different engagement levels during the meditation session, with the meditator exhibiting much lower engagement while completing the meditation session. As such, it appears that the meditator was better able to maintain higher levels of focused attention over a long period of time while answering questions, but was also able to quickly drop into a state of lower engagement while meditating, possibly due in part to his regular meditation practice.

While investigators (Kabat-Zinn 1990, 1998; Shapiro and Schwartz 2000; Teasdale 1999; Segal et al. 2002), provide descriptions of what mindfulness may be, the field lacked an operational definition from which to create testable hypotheses and determine the utility of such a construct in everyday life (Bishop 2002). Mindfulness, for the purpose of this effort, is operationally defined as a basic human ability to regulate the focus of attention toward current actions and events, without influence of personal affective states, such as when experiencing anxiety or arousal (Bishop et al. 2004). A mindful brain state potentially translates into learning through the breaking of brain habits that keep the brain inflexible to learning new concepts and strategies important for performing tasks in dynamically changing operational environments, such as military contexts (Langer 2000). Langer (2000) also suggests that current curricula are setup for mindless learning through repetition and single exposure. The brain is working in mindless mode when it (1) operates out of an inflexible habit and (2) when it biases to a particular perspective. In each case, the brain is not allowed to extend learning when the context of application changes.

Inherent in the mindless mode of brain operation during training is the lack of control or self-regulation of focused attention (i.e. engagement) by the trainee. Research in the operational environment that utilizes brain state measures such as electroencephalography (EEG) indicate that mindfulness improves the capability of the brain to engage in the type of focused and sustained attention necessary for skill learning and safety monitoring (Parasuraman 2000). Other benefits of improved self-regulation and self-management of the brain networks within the training context include enhanced neural connections necessary for improving data processing and memory retrieval (Shapiro et al. 2006) and enhanced executive control over inflexible biased response (Teper and Inzlicht 2013). The ability to improve engagement during training holds the potential to deepen learning such that the transfer of skills training to the military context is quicker and more sustained, leading to enhanced readiness and success on the battlefield.

There are potentially two components of mindfulness relevant for studying the effects of mindfulness on training complex military skills: (1) acquiring skills in the self-regulation of attention and (2) adopting an acceptance toward life’s experiences (Bishop et al. 2004). The concept of “staying in the moment” relies upon the brain’s ability to process the immediate experience without distraction. Distraction and boredom are cognitive states that research suggests inhibits or preoccupies the attentional network, and therefore negatively impacts learning (Eastwood et al. 2012). A person who is more in control of their focused attention can potentially shift attention quickly back to the immediate task if the brain becomes distracted. There are more brain resources to process the training material. Additionally, the attitude one has toward the tasking of the current moment can affect how deeply one learns. For example, curricula that do not foster perspective taking (passive) and are repetitive (boring) can create a context where trainees cannot maintain attentional focus or switch from where the brain wanders back to the task (Posner 1980). Consequently, the training information is not deeply process such that learning takes longer and transfer of training is improbable (Langer and Moldoveanu 2000). Following a flow state methodology where the learning material and the internal mental states of the trainee are kept in balance during training may be a better predictor of accelerated learning and training effectiveness than other constructs such as cognitive load (Fidopiastis 2011). Future research is needed to formalize and validate this theoretical model within an experimental paradigm that allows for further exploration of the effects of various instructional design and adaptive training techniques, as well as various mindfulness interventions, on both performance and learner engagement.