Brain Function Assessment in Learning

Although the issues around emotions and learning are not new, the term affective learning has only recently been defined as the learning that relates to the learner’s interests, attitudes, and motivations. In the digital age we live though, affective learning is destined to be technology driven or at least enhanced. Having overemphasised the cognitive and relatively neglecting the affective dimension in the past, technology enhanced learning is now enforced by new neuroscience findings that confirmed that affect is complexly intertwined with thinking, and performing important functions that may guide rational behaviour, assist memory retrieval, support decision-making and enhance creativity. To cope with personalised learning experiences in such models of learners though, intelligent tutoring systems must now contain “emotion, affect and context”, in analogy to successful human tutors. However, measuring and modelling learners’ emotional and affective states remains a difficult task, especially when real-time interactions are envisaged. In this paper, the concept of affective learning is furnished with case studies where the roles of technologies, neuroscience, learning and education are interwoven. Medical education is borrowed as a domain of reference. Neuroscientific emphasis is placed in the synergy of two perspectives, namely, the detection and recording of emotions from humans and ways to facilitate their elicitation and their subsequent exploitation in the decision-making process. The paper concludes with a visionary use case towards affective facilitation of training against medical errors and decision making by intelligent, self-regulated systems that could exploit scenario based learning to augment medical minds for tomorrow’s doctors.

Human learning has been dramatically altered by the new situation that saw us climb down trees and out of the savanna to larger communities with great reliance on agriculture and more recently industry. Psychology appeared as a scientific discipline at a time that formal education for all was becoming accepted. From the very beginning psychology and education had a major influence on each other. Education is notoriously slow to change. As the ideas of developmental psychologists started influencing education policy, new paradigms for education emerge from a variety of disciplines including computer science, medicine and particularly neurosciences. Each of these disciplines has its own vocabulary and progress is often limited because there is no common framework to bring together specialists from different disciplines or to formulate common research. We provide such a framework through a generalization of key concepts of developmental psychology. In the new framework, these concepts are cloaked with what we might call the standard model of modern neuroscience. Here, we customize this framework for learning and education. Formal education is seen a continuation of a process that begins with the mother and develops in pre-school play. The main goal of this process is to maintain and continually update an internal representation of the external world in the key brain networks while keeping intact the core representation of self. The first steps in a research program using this new framework are described with some results and conclusions about future actions.

This is a position paper describing the author’s views on a potential new research direction for assessing, constructing and exploiting brain-founded models of learning of individual as well as collective humans. The recent approach – called ViewpointS – aiming to unify the Semantic and the Social Web, data mining included, by means of a simple “subjective” primitive – the viewpoint - denoting proximity among elements of the world, seems to offer a promising context of innovative empirical research in modeling human learning less constrained with respect to the previous three other ones. Within this context, a few phenomena of serendipitous learning have been simulated, showing that the process of collective construction of knowledge during free navigation may offer interesting side effects of informal, serendipitous knowledge acquisition and learning. We envision therefore an extension of the modeling functions within ViewpointS by adding measures of the emotions and mental states as acquired during experimental sessions. These brain-related components may in a first phase allow to describe and classify models in order to understand the relations among knowledge structures and mental states. Subsequently, more predictive experiments may be envisaged. These may allow to forecast the acquisition of knowledge as well as sentiment from previous events during interactions. We are convinced that useful applications may range, for instance, from Tutoring, to Health, to consensus formation in Politics at very low investment costs as the experimental set up consists of minimal extensions of the Web.

At present, there is a growing number of older population groups worldwide, which results in serious social and economic issues since some of these older people require special care. Therefore there is constant effort to maintain these older people active as long as possible. This might be done not only by pharmacological therapies, but especially by non-pharmacological approaches, among which brain training with the help of a computer seems a good solution. The purpose of this article is to explore available clinical studies implementing computer-based brain training programs as intervention tools in the prevention and delay of cognitive decline in aging, with special focus on their effectiveness. This was done by conducting a literature search in the databases Web of Science, Scopus, MEDLINE and ScienceDirect, and consequently by comparing and evaluating the findings of the selected studies. The findings show that that the exploited online brain (cognitive) training programs usually have a moderate positive effect on the delay of cognitive decline especially in the area of reasoning skills, working memory, and processing.

The prediction of corporate bankruptcy has been addressed as an increasingly important financial problem and has been extensively analyzed in the accounting literature. Over recent years, several machine learning methods have been effectively applied to build accurate predictive models for detecting business failure with remarkable results, such as neural networks (NNs) and ensemble methods. This paper investigates the effectiveness of the active learning framework to predict bankruptcy using financial data from a set of Greek firms. Active learning is an emerging subfield of machine learning exploiting a small amount of labeled data together with a large pool of unlabeled data to improve learning accuracy. From what we know so far there exists no study dealing with the implementation of active learning methodologies in the financial field. Several experiments take place in our research comparing the accuracy measures of familiar active learners and demonstrating their efficiency in contrast to representative supervised methods.

In recent years, there is a growing research interest in applying data mining techniques in education. Educational Data Mining has become an efficient tool for teachers and educational institutions trying to effectively analyze the academic behavior of students and predict their progress and performance. The main objective of this study is to classify junior high school students’ performance in the final examinations of the “Geography” module in a set of five pre-defined classes using active learning. The exploitation of a small set of labeled examples together with a large set of unlabeled ones to build efficient classifiers is the key point of the active learning framework. To the best of our knowledge, no study exist dealing with the implementation of active learning methods for predicting students’ performance. Several assessment attributes related to students’ grades in homework assignments, oral assessment, short tests and semester exams constitute the dataset, while a number of experiments are carried out demonstrating the advantage of active learning compared to familiar supervised methods, such as the Naïve Bayes classifier.

In the present study we examined the effects of working memory training on cognitive flexibility in humans. Forty healthy male participants were divided into three groups (matched for demographic variables, schizotypy, impulsivity and baseline cognitive flexibility): a) fully adapted group (participants were fully trained with an executive working memory task for six consecutive days); b) partially adapted group (participants were partially trained with an executive working memory task for six consecutive days) and c) control group (participants did not receive cognitive training). Following training, participants were examined with a second cognitive flexibility task. We found that the fully adapted group had improved cognitive flexibility (they made fewer errors and needed fewer attempts to complete the test) compared with both the partially adapted (all p values <0.005) and the control (all p values <0.05) groups, who did not differ between each other (all p values >0.2). These findings could have significant implications in the development of therapeutic approaches for the improvement of cognitive deficits in neuropsychiatric disorders.

One of the current frontier research themes in informatics relates to the extent to which computers and machines in general can become capable of learning and teaching each other. Hopes have been raised that their education could benefit from emulating mechanisms underlying learning in animal brains. An overview of these mechanisms will be briefly presented with a focus on the recently revealed fundamental role of sleep in memory consolidation and learning., Compared to brains, computers are found very much inferior when it comes to learning. Several road signs are suggested for enriching computers’ repertory in the direction of increasing their capacity to learn by becoming more brain-like. However, the prospect of achieving such goal with state of art technology appears extremely dim.

Very specifically, functional behavior assessment is a domain in developmental psychology looking at the reasons behind a child’s observed behavior. More generally, it can be considered as the search for the explanation of human and non-human actions. Towards this goal, computational cognitive neuroscience offers a new range of possibilities that contrast with the usual statistical approaches. An attempt to assess brain functionalities in learning is illustrated here through the simulation of analogical inferences. As a main result of this paper, the mapping of evolutive cognitive schemas onto neural connection structures involving two types of cognitive transfer points out to a possible discontinuity between human and non-human minds.

Knowledge tracing aims to quantify how well students master the knowledge (tags) being tutored by analyzing their learning activities (e.g., coursework interaction data). It plays an important role in intelligent tutoring systems. In this paper, we cast knowledge tracing as a performance-prediction problem, which predicts the performances of students on exercises labeled by multiple knowledge tags, and propose to tackle this problem using Deep Learning techniques. We applied several Recurrent Neural Network architectures to model complex representations of student knowledge and predict future performances of students. Our experimental results demonstrate that the neural network architecture based on stacked Long Short Term Memory and residual connections give superior predictions on the future performances of learners. To model how a student answered a question that contains multiple knowledge tags, we explored three different variants to map knowledge states to prediction.

The current study aims to measure the goal orientations motivation in different scenes of a video-game. The evaluation of player experience was done with both subjective measures through questionnaire and objective measures through brain wave activity (electroencephalography - EEG). We used GameFlow questionnaire to characterize the player’s mastery goal in playing video game (Master or Performant). In terms of brain activity, we used the Frontal alpha asymmetry (FAA) to assess the player approach/withdrawal behavior within a game scene. Using game scene’s design goal (defined by OCC variables) and player personality traits (using Big Five questionnaire), the resulting machine learning model predicts players’ motivational goal orientations in order to adapt the game. In this study, we address player’s motivation in game scenes by analyzing player’s profile, his situation in scene and affective physiological data.

Humans’ cognitive and affective states are constantly subject to regular and sudden changes. The origins of these changes are multiple and unpredictable. Virtual Reality (VR) game environments could represent an immersive unconstrained experimental context in which game designers could control a wide range of parameters that act on these states. In this paper, we propose to track and adapt to individuals’ frustration and excitement levels in real time while interacting with a VR environment. We developed “AmbuRun”, a VR game designed to modify the speed and the difficulty in real time. A neural agent was created to control these parameters within the game using an intervention strategy that was intended to induce appropriate modifications of the players ‘excitement and frustration level. An experimental study involving 20 participants was conducted to evaluate our neurofeedback approach. Results showed that intelligent control through neurofeedback of speed and difficulty affected excitement and frustration before and after the agent action.

Decision-making can be regarded as a cognitive process integrated in our interaction with the environment. This interaction comprises of a plethora of sensory or mental stimuli. Among them, visual awareness and semantic recognition during decision-making tasks are of main importance. Nowadays, disaster education becomes a part of the curricula to foster a more resilient population, and relevant research emerges. The purpose of this exploratory study was to investigate visual awareness and semantic recognition during a visual decision-making task concerning earthquakes, by measuring brain activity and especially event-related potentials. The task consisted of digital images, representing useful and non-useful items constituting a survival kit in case of an earthquake. The subjects, seven adult males, had to distinguish between those useful and non-useful items. A late positive component (P300) and an early posterior negativity (N200) were studied since they are the most prominent components for categorization tasks. Our results suggested that participants distinguished the useful items in a series of non-useful stimuli based on their semantic content. These preliminary results indicate that these stimuli could be integrated in an educational digital environment concerning disaster preparedness.

Real-time neurofeedback plays an increasing role in today’s clinical and basic neuroscience research. In this work, we present a real-time sleep EEG spindles detection algorithm fast enough to be used for real time acoustic feedback stimulation. We further highlight the architecture of a system that implements the algorithm and its experimental evaluation. This system can handle EEG data acquired by various means (i.e. conventional EEG systems, wireless sensors) and a response time of a few msecs has been achieved. The presented algorithm is dynamically adaptive and has accuracy similar to other well-known non real-time algorithms. Comparison and evaluation was performed using EEG data from an open database.

Feedback constitutes a fundamental aspect of educational systems that has a substantial impact on students learning and can shape their mental models. The delivery of appropriate feedback in terms of time and content is crucial for facilitating students’ knowledge construction and comprehension. In this paper, we examine the complex nature and the efficiency of feedback in the context of a virtual reality educational environment. More specifically, we study the effect that different types of feedback such as feedback with visualized animations of procedures, can have on students learning and knowledge construction in a virtual reality educational environment for learning blind and heuristic search algorithms. An experimental study was designed where participating students were engaged with learning activities and solved exercises in different feedback conditions. Results from the study indicate that visual types of feedback can have a substantial impact on students’ learning, assisting them in better understanding the functionality of the process studied with respect to performance and mistakes.

Relaxation techniques can relieve us from stress, anxiety, pain and maladies. Many researchers succeed in relaxing subjects by various methods. However, few concerned about the ability of relaxation. Hence, the main goal of this study is to help people relax faster. We developed a virtual reality neurofeedback relaxation training system, called Virtual Sophrologist, which 1) immerses users in fantastic environments by a Virtual Reality headset, 2) guides users to follow the Sophrology instructions by a female voice, and 3) displays feedback in real time, which are translated from the Meditation Score collected by EEG. To evaluate this system, we recruited 6 subjects to participate in our 8-session relaxation training and collected their subjective data (by self-report) and objective data (by EEG) to measure from psychological level and to calculate the Time Interval to Relaxation that they took to reach the maximum Meditation Score. The results show 1) decreases in Anxiety and Depression Score from the psychological level, 2) a decrease in Time Interval to Relaxation and 3) an increase in the maximum Meditation Score. Therefore, our system will be useful as a training tool for users who need or want to relax fast and deep whenever they need.

The hippocampus is a brain region crucially involved in various cognitive functions including learning and memory processes. The hippocampal functions are performed by specific computations of its intrinsic neural circuitry in combination with interaction of the hippocampus with other brain regions. Therefore, the hippocampus has been conceived as a key network to studying and understanding the fundamental neural computations that supports higher brain functions. The hippocampus-involving functions are segregated along the longitudinal axis of the hippocampus. Importantly, it has been recently revealed that the local hippocampal circuit presents significant specializations between the two opposite poles or segments of the structure, namely between the dorsal and the ventral hippocampus suggesting that distinct neural processing may support the different functions performed by the hippocampus segments. The signal processing by neural networks crucially involves synaptic computations. In this study, we examined the synaptic dynamics of the dorsal and ventral synapses under conditions of different activation frequencies. We found that under consecutive activation the dorsal synapses display strong facilitation at a wide range of frequencies (1-40 Hz) while in ventral synapses the facilitation is restricted only to low activation frequency (1 Hz) and it lasted very shortly during activation. Thus, ventral synapses are mostly depressing. This evidence suggests that the dorsal hippocampal synaptic circuit presents wide-band filtering characteristics while the ventral are depressing low-pass synapses. The differing synaptic properties of the dorsal and the ventral hippocampus may underlie the higher ability for long-term plasticity of the dorsal hippocampus and the initiation of basic endogenous network oscillation in the ventral hippocampus.

Detecting the real-time human emotions became recently one important issue in Artificial Intelligent (AI). Numbers of research on emotional facial expressions, the effect of emotion on heart rate, eye movement and the evolution of emotions with the time show the interest of this topic. This paper presents a method for observing the human’s emotional evolutions (sequence of emotions) based on brain activities in its different parts. The Emotiv EPOC headset collects the data of Electroencephalograms (EEG) of the participant to calculate the arousal and valence. After training the system with headset output data, the noise and brain’s data other than emotional information will be cut out according to two levels of filtering. Finally, mapping the result (arousal and valence) with two dimensions circumplex space model presents the real-time emotional evolutions of the participant. Real-time emotional evolutions show all the picks of positive and negative feelings, moreover, analyzing the EEG data will allow recognizing the general emotions, which are the strongest routine senses of the participant. Comparing the unexpected reaction, the time taken by general emotion and feelings in picks, give us a tool to observe the health situation of the people and on the other hand, is an instrument to measure the mood of the people against a video game, news and advertising.