Under the modern network environment, ubiquitous learning has been a popular way for people to study knowledge, exchange ideas, and share skills in the cyberspace. Existing research findings indicate that the learners’ initiative and community cohesion play vital roles in the social communities of ubiquitous learning, and therefore how to stimulate the learners’ interest and participation willingness so as to improve their enjoyable experiences in the learning process should be the primary consideration on this issue. This paper aims to explore an effective method to monitor the learners’ psychological reactions based on their behavioral features in cyberspace and therefore provide useful references for adjusting the strategies in the learning process. In doing so, this paper firstly analyzes the psychological assessment of the learners’ situations as well as their typical behavioral patterns and then discusses the relationship between the learners’ psychological reactions and their observable features in cyberspace. Finally, this paper puts forward a CyberPsychological computation method to estimate the learners’ psychological states online. Considering the diversity of learners’ habitual behaviors in the reactions to their psychological changes, a BP-GA neural network is proposed for the computation based on their personalized behavioral patterns.
In today’s society, ubiquitous learning (U-learning) has been bringing about magical changes to traditional education [
In that circumstance, the participants of ubiquitous learning have formed an ecological social community which displays many new features and is sustained by the common motivation and goal, favorable atmosphere, and emotional communication, as well as enjoyable experiences [
In the past century, learning theory has made great progress. Constructivism, situated cognition, and informal learning theories offer the guidance for ubiquitous learning by focusing on the individual’s self-construction of knowledge and aim to provide new online intelligent, virtual interactive, and seamless learning patterns for learners on the basis of such situational characteristics as location, time, and environment [
However, the environment and patterns of ubiquitous learning in a social community differ from those in classroom completely. Different social media and interactive activities between the participants have all major impacts on the cognition, emotion, and attitude in the learning process [
Therefore, considering the new features and special environment in cyberspace, a further study of ubiquitous learning theory is required to provide comprehensive and systemic guidance for improving the organizational mode, learning behaviors, and teaching skills, which involves the interdisciplinary areas of pedagogy, psychology, sociology, organizational behavior, neuroscience, and information technology. Scholars have paid attention to the above issue in the 1990s. CyberPsychology, coined by Dr. John Suler in his hypertext book “The Psychology of Cyberspace” with the first version appearing in January of 1996, launched the original conceptual framework for understanding how people react to and behave within cyberspace [
By affective computing [
This paper aims to explore an effective method to monitor and analyze the learners’ psychological reactions based on their observable behaviors in cyberspace and therefore help to conduct the smart education in ubiquitous learning. This paper is organized as follows. Section
Social community of ubiquitous learning is affected by a lot of factors associated with social, psychological, organizational, managerial, and technological aspects. Since the 1990s, a lot of scholars such as Webster and Hackley [
Recent experimental observations showed that a desirable learning atmosphere, good visual effects, pleasant voices, suitable topics and materials, and positive evaluation feedbacks were the most important factors that aroused the learners’ interest in ubiquitous learning and contributed to the pleasant emotional experiences [
The calibration for scoring records is shown as in Table
Calibration for scoring records in psychological assessment.
Variables | Scores | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Attention | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
Interest | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
Emotion | −5 | −4 | −3 | −2 | −1 | 0 | 1 | 2 | 3 | 4 | 5 |
Satisfaction | −5 | −4 | −3 | −2 | −1 | 0 | 1 | 2 | 3 | 4 | 5 |
In ubiquitous learning, the learner’s behaviors can be divided into three categories from simple to complex: operational behaviors, information exchanging behaviors, and problem-solving behaviors [
The pattern of each learner’s habitual behaviors can be expressed as formula
The action’s data are mainly acquired from the server’s log files or by some online tracking tools. The switching frequency and retention time of webpages, the locations and movements of the mouse, and the keyboard operations are all the important parameters as the features to reflect the learners’ psychological reactions. For example, if the learner has strong interest in something and is in a fairly good mood, he/she tends to stay on the interesting webpage longer, use his/her mouse and keyboard with higher frequency, answer the asked questions more quickly, and be more willing to make positive comments. On the contrary, if he/she is anxious and fretful, he/she will switch from one webpage to another frequently, move the mouse in a wide range quickly, and be more likely to give negative feedback [
Basic actions and parameters of the learner’s habitual behaviors in cyberspace.
Number | Actions | Parameters |
---|---|---|
0 | No action | The object in screen center |
1 | Mouse: click | On a button or link, on another place |
2 | Mouse: scroll | Speed, the object in screen center when stopping scrolling |
3 | Mouse: move | Speed, radius |
4 | Mouse: open a new page | Null |
5 | Mouse: change a page | Null |
6 | Mouse: close a page | Null |
7 | Mouse: store a page | Null |
8 | Keyboard: input | Number of characters |
9 | Keyboard: delete | Number of characters |
10 | Mouse and keyboard: retrieve information | Number of keywords |
11 | Mouse and keyboard: post information on BBS | Number of characters |
12 | Mouse and keyboard: send information to other people | Number of characters, number of receivers |
13 | Mouse and keyboard: chat with other people | Number of characters, number of people chatted with |
14 | Streaming media: voice communication | Acoustic feature parameters |
15 | Streaming media: video communication | Visual feature parameters |
The psychological reactions of the learners are the results of a series of neural activities dominated by the brain mechanism [
However, in the ubiquitous learning environment, the possible information we can obtain from the learners is their observable actions as well as the voice and video signals produced by their online activities under some circumstances. Considering the protection of the learners’ privacy, CyberPsychological computation cannot be based on the content analysis and semantic detection of the above information. Fortunately, technologies of psychological computation from voice and video signals have developed quickly in recent years. For example, progress has been made on the voice signal based on its acoustic features parameters such as speech speed, voice intensity, pitch frequency, LPCC (Linear Prediction Cepstrum Coefficient), and MFCC (Mel Frequency Cepstrum Coefficient) [
In our research, we mainly consider the learners’ behavioral features from their observable actions in cyberspace. The relationship between the learners’ psychological reactions and their observable features is related to the habitual patterns of the learners’ behaviors, which is dependent on a statistical study in the real cases. Figure
Actions of three learners to the same psychological reaction
From Figure
In order to monitor the learners’ psychological reactions in the learning process, we put forward a CyberPsychological computation method as shown in Figure
CyberPsychological computation method on social community of ubiquitous learning.
In the learning system, online activities such as answering questions, retrieving information, discussing problems, chatting with each other, and submitting assignments are all assigned in different functional areas on the webpage. So the layout structure and its related information of the webpage are extracted to assist in identifying the learners’ action.
Every change of the elements on the webpage, such as moving, clicking, and scrolling of the mouse or the operation of the keyboard, will trigger the corresponding JavaScript function. In order to meet the requirements of real-time data collection, we adopted the PHP language to program the above JavaScript function and process the mouse and keyboard data acquisition, which can be realized by the intelligent multiagent technology [
As we discussed before in this paper, the learners’ psychological reactions are exhibited in their diverse habitual behaviors and the CyberPsychological computation should be based on their personalized behavioral patterns. Therefore, we consider the learner’s ID and his/her historical behavior patterns in our method. The computation on the learners’ psychological reactions based on their behavioral features in the learning process can be regarded as a dynamic and nonlinear estimation problem by machine learning. So the subjective psychological assessment of learners’ situations should be given by them in the sample training.
Nonlinear Regression, Kalman Filter, Artificial Neural Network (ANN), and SVR (Support Vector Regression) have been reported as successful technologies for solving the above estimating problem [
BP (back propagation) neural network is a nonlinear function to establish the uncertain and continuous relations between input and output variables based on trained samples by machine learning. It can continuously modify the network weights and thresholds through the error back propagation algorithm (BP algorithm) and reach the target of minimizing the mean square error. However, the traditional BP neural network easily causes the nonconvergence problem or falls into a local extremum. Those defects can be overcome by combining with a genetic algorithm (GA) [
Figure
BP-GA neural network for CyberPsychological computation.
The input layer has 12 nodes which are composed of the data of course information, learner’s ID, mouse actions, and keyboard actions. The output layer has four nodes which correspond to the scores of the learner’s psychological reactions in formula (
The BP-GA neural network operates firstly with the training by a group of sample data which are tested and recorded in the real cases. After it converges to a stable state, the BP-GA neural network can be applied to the computation in the learning process.
The experiment is based on a training course of life health and medical emergency rescue. In order to exclude the effects of the instructor, we edited this course into 20 lectures which are all taught by video tutorials and operated under a designed controlling program in the learning process. Every lecture runs for 50 minutes and is divided into 5 periods (from
Learning process in each lecture of training course of life health and medical emergency rescue.
The learners are 16 social participants with more than 60 learning hours in a normal and stable environment of ubiquitous learning community. They are required to complete this course in 40 days. However, they can arrange the learning time freely in a ubiquitous learning environment. In this experiment, we assigned different activities in each period of a lecture and asked the learners to make a psychological assessment of their general situations in the final 1 minute of each period.
Table
Duration and activities in each period of a lecture.
Period | Duration | Activities |
---|---|---|
P1 | 8 | Watching video tutorials |
P2 | 12 | Watching video tutorials |
P3 | 10 | Watching video tutorials |
P4 | 10 | Completing an online individual assignment |
P5 | 10 | Discussing with other learners |
In the end of our experiment, 11 participants successfully finished this course and provided completed subjective assessments of their psychological reactions in the learning process. We extracted randomly the records of 16 lectures as the training samples for BP-GA neural network and took the rest of the 4 lectures for the test of computation results. The above process was carried out 3 times. Table
Test results of CyberPsychological computation.
Variable | Average score by assessment | Estimated result by computation | Relative error | Standard deviation |
---|---|---|---|---|
|
7.213 | 6.102 | −15.4% | 1.237 |
|
7.862 | 8.333 | 6.0% | 0.6752 |
|
4.220 | 3.281 | 22.3% | 1.401 |
|
8.581 | 7.114 | 17.1% | 1.898 |
Table
By analyzing learners’ psychological reactions and their varying characteristics in the learning process, we can explore the statistical distribution and the changing rhythms of the learning community in different periods and scenarios and find the learners’ ROI (Region of Interest). This has significant implications on the design of learning strategies such as the education scheme and teaching plans which are on-demand and more appealing and enjoyable to learners in order to provide state-of-the-art teaching models, skills, and technologies for smart education in a ubiquitous learning environment [
With the development of modern information network and mobile communication technology, personalized learning in a ubiquitous learning environment has been made possible in web space accessible by a variety of multimedia terminals anytime and anywhere. It has brought about new changes to the theories, means, and patterns of traditional education. In the U-learning environment, learners’ psychological reactions and learning experiences have significant impacts on stimulating their learning interest and improving the teaching efficiency.
This paper aims to explore an effective method to monitor the learners’ psychological reactions based on their behavioral features in cyberspace and therefore provide useful references for adjusting the strategies in the learning process. As the comprehensive result of our research, a CyberPsychological computation method based on BP-GA neural network was proposed for estimating the learners’ psychological states online. The experimental result shows that it can achieve accuracy near 78%.
Future researches should consider the dynamic psychological reactions of the learners through the studies of their physiological signals such as EEG, ECG, EDR, respiration, and skin temperature by a wearable device system and assist in obtaining a more precise psychological assessment of the learners’ situations. Besides, the influence factors of different lecture’s contents and the statistical distribution of the learning community as well as its varying characteristics in the learning process are all worthy of further explorations based on more cases and samples.
Xuan Zhou, Genghui Dai, and Shuang Huang are the joint first authors of this paper. Genghui Dai and Xuemin Sun are the joint corresponding authors.
The authors declare that there is no conflict of interests regarding the publication of this paper.
This research was supported by the 2nd Regular Meeting Project of Science and Technology Cooperation between China and Serbia (no. 2-9/2013), Shanghai Philosophy and Social Sciences Plan, China (no. 2014BGL022), Project of Scientific Research Fund of Sichuan Provincial Education Department (no. 15SB0219), and Project of Sichuan Conservatory of Music (CY2014173) and accomplished by an interdisciplinary team organized by Xuan Zhou, Genghui Dai, and Shuang Huang.