=Paper= {{Paper |id=Vol-2845/Paper_24 |storemode=property |title=Reflex Model of Online Learning |pdfUrl=https://ceur-ws.org/Vol-2845/Paper_24.pdf |volume=Vol-2845 |authors=Oleksii Yehorchenkov,Nataliia Yehorchenkova,Iurii Teslia |dblpUrl=https://dblp.org/rec/conf/iti2/YehorchenkovYT20 }} ==Reflex Model of Online Learning== https://ceur-ws.org/Vol-2845/Paper_24.pdf

Reflex Model of Online Learning
Oleksii Yehorchenkova, Nataliia Yehorchenkovab and Iurii Tesliac
a
Taras Shevchenko National University of Kyiv, Bohdana Havrilishina 24, Kyiv, 04116, Ukraine,
b
Kyiv National University f Construction and Architecture, Povitroflotsky ave., 31, Kyiv, 03037, Ukraine
c
National Aviation University, Kosmonavt Komarov ave.,1 Kyiv, 03058, Ukraine

Abstract
Given the increasing popularity of learning through web-services, a large number of
researches is devoted to solving the scientific problems of online learning. Such problems
include, for example, convergence and interoperability in terms of research of various forms
of convergence (organizational, technological, pedagogical, sectoral, institutional, etc.). The
article considers what is online learning and what is its current state and trends. Modern
scientific methods of online learning are shown. It is highlighted that commonly used
methods such as Big Data, intelligent data analysis, and analytical training. According to the
article, data science methods have received the highest priority in online learning studies. The
model of the reflex expert system of online learning management based on the theory of non-
force interaction is proposed and two stages in the technology of its functioning are
considered.

Keywords 1
online learning, data science, data analytics, big data, reflex model

1. Introduction

The COVID-19 coronavirus pandemic and the self-isolation regime have led to a real stir in the
field of online learning. According to the GetCourse platform, which hosts distance courses of
hundreds of online schools, sales of e-courses grew by 20% in the second half of March 2020
compared to the first half of the month. The number of users trying to organize their courses increased
by 30%. But even under normal conditions, interest in online learning is actively growing from year
to year [1].
Online learning is a system of education using information and electronic technologies. There is a
definition given by UNESCO experts: “ Online Learning - learning using the Internet and
multimedia”.
Online learning includes:
 independent work with electronic materials using a personal computer, PDA, mobile
phone, DVD-player, TV, and other electronic materials;
 receiving advice, advice, assessments from a remote (geographically) expert (teacher), the
possibility of remote interaction;
 creation of a distributed community of users (social networks) conducting common virtual
educational activities;
 timely round-the-clock delivery of e-learning materials; standards and specifications for e-
learning materials and technologies, distance learning tools;
 formation and improvement of information culture among all heads of enterprises and
divisions of the group and their mastery of modern information technologies, increasing
the efficiency of their usual activities;

IT&I-2020 Information Technology and Interactions, December 02–03, 2020, KNU Taras Shevchenko, Kyiv, Ukraine
EMAIL: alexee@ukr.net (A. 1); realnata@ukr.net (A. 2); teslyas@ukr.net (A. 3)
ORCID: 0000-0003-1390-5311 (A. 1); 0000-0001-5970-0958 (A. 2); 0000-0002-5185-6947 (A. 3)
©️ 2020 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)

251
 mastering and popularizing innovative pedagogical technologies, transferring them to
teachers;
 the ability to develop educational web resources;
 the opportunity at any time and in any place to receive modern knowledge located in any
accessible point of the world;
 the availability of higher education for persons with special psychophysical development
[2].
Current online learning research brings together pedagogical, technical, and organizational
concerns within a wider set of socio-cultural factors. Understanding issues &challenges in respect of
online learning is of significant importance to the research communities involved in online learning
and will have a significant role in forming future practices [3].
It is presented in the article definition of online learning, its advantages, and issues of it, and
modern researches of it.

2. Materials and Methods

Today, among scientists there are differences in the definition of «online learning». Authors [4]
describes the differences and presents main views on «online learning»: Some prefer to distinguish
the variance by describing online learning as “wholly” online learning [5], whereas others simply
reference the technology medium or context with which it is used [6]. Others display direct
relationships between previously described modes and online learning by stating that one uses the
technology used in the other [7]. Online learning is described by most authors as access to learning
experiences via the use of some technology [8]. Authors identify online learning as a more recent
version of distance learning which improves access to educational opportunities for learners described
as both nontraditional and disenfranchised. Other authors discuss not only the accessibility of online
learning but also its connectivity, flexibility, and ability to promote varied interactions [5]. These
authors in particular not only elude to online learnings' relationship with distance learning and
traditional delivery systems but then, like [8] makes a clear statement that online learning is a newer
version or, an improved version of distance learning.
We will use the definition of authors [9] who identify online learning as a form of distance
learning or distance education, which has long been a part of the American education system, and it
has become the largest sector of distance learning in recent years.
The fact is that online learning has several advantages over face-to-face learning.
The advantages are described in the paper [4]. Here are some of that:
1. Enhancing student-to-student and faculty-to-student communication.
2. Enabling student-centered teaching approaches.
3. Providing 24/7 accessibility to course materials.
4. Providing just-in-time methods to assess and evaluate student progress [10].
Compared to the number of studies that found positive or no significant effects for student learning
outcomes in the online format, the number of studies that found mixed or negative significant effects
is much smaller, by a full order of magnitude. Some of these studies are direct contradictions of the
studies with positive results: they find that students performed worse in the online format compared to
the traditional format. Some studies’ findings are more nuanced. They find that there are negative
effects for certain groups of students and null findings for others. There are studies discussed in this
section that systematically examine the ubiquitous self-selection bias of online learning: the
endogeneity of learning environment choice. Most studies on distance or online learning do not
examine this selection bias, which some researchers posit as a culprit for the “no significant
difference” phenomenon [11].
Due to the increasing popularity of online learning and its ever-greater penetration into the sphere
of education of the whole world appear a strong interest from the scientists’ side in questions of
measurement, collection, analysis, and reporting of data about students with aim of understanding and

252
optimization of learning and conditions in which it occurs. Let's consider existing methods of online
learning.
In addition to optimization tasks, in online learning, it is necessary to use data processing methods
and learning analytics, which are used to study and build models of the online learning system.
At its core, learning analytics (LA) is the collection and analysis of usage data associated with
student learning. The purpose of LA is to observe and understand learning behaviors to enable
appropriate interventions. The reports that an LA application generates can be very helpful for
instructors (about student activities and progress), for students (feedback on their progress), and
administrators (e.g. aggregations of course and degree completion data). The main elements of LA
include (fig.1) [12]:

Figure 1: The main elements of learning analytics

Despite the enormous potential they hold to improve education, learning analytics is not without
concerns. Privacy for learners and teachers is a critical issue. While I see analytics as a means to
improve learner success, opportunities exist to use analytics to evaluate and critique the performance
of teachers. Data access and ownership are equally important issues: who should be able to see the
analysis that schools perform on learners? Other concerns relate to error-correction in analytics. If
educators rely heavily on analytics, effort should be devoted to evaluating the analytics models and
understanding in which contexts those analytics are not valid [13].
Essentially, for learning analytics to have a broad impact in education, the focus needs to move
well beyond basic analytics techniques such as those found in Google Analytics. An integrated
learning and knowledge model is required where the learning content is adaptive, prior learning is
included in assessment, and learning resources are provided in various contexts (e.g. “in class today
you studied Ancient Roman laws, two blocks from where you are now, a museum is holding a special

253
exhibit on Roman culture”). The profile of the learner, not pre-planned content, needs to drives
curriculum and learning opportunities [13].
Current online-learning research brings together pedagogical, technical, and organizational
concerns within a wider set of socio-cultural factors. These factors influence the research agenda in
the e-learning system. Understanding these broader social and cultural issues is of significant
importance to the research communities involved in e-learning and will have a significant role in
informing future practices. In consulting the research community, several research issues emerged
(fig.2) [3]:

Figure 2: Research issues of online learning

3. Model of Reflex Expert System of Online Learning Management

As was shown and proved by biologists based on the functioning of living organisms lies the
formation of reflexes to various influences. A reflex, or reflex action, is an involuntary and nearly
instantaneous movement in response to a stimulus.
At the level of simple biological objects, reflexes provide the forming of a “necessary” reaction to
the state of the environment.
At the human level, reflexes are formed not only on physical influences but also on the impact of
the socio-political sphere, on the information impacts of other people, for example, teachers at school,
colleagues, wives, etc.
Of course, these reflexes are very complex, not unambiguous, cannot be represented by a simple
model of “impact” – “reaction”. Rather, this model looks like “a lot of influences” – “the most
profitable reaction, from the position of obtaining a positive attitude”. In the non-force interaction
theory, a mathematical apparatus has been developed that allows one to describe the mechanism of
producing reflexes and the “behavior” of technical systems on this basis.

254
By assumption, the interaction of neurons is based on the same laws and is realized by the
developed model of non-force interaction. And this found its realization in reflex intelligent systems
capable of accumulating information about the functioning environment and forming an adequate
response (reflexes) to everything that happens in this environment.
This is a system for evaluating investment proposals in development; natural language access to
databases; assessment of the influence of harmful substances in the water resources of the region on
the health of the population; forecasting the results of sports events. Their main advantage is the
simplicity of development and the effectiveness of solving a variety of intellectual problems.
These systems are represented in Figure 3.

Figure 3: Reflex systems

Let us review the voice control systems.
The question arises whether it is possible to present the process of voice interaction using the
models that are developed in the non-force interaction theory?
Traditionally, speech recognition systems are based on the principle: “spoken language” 
“representation of speech as a set of linguistic constructions”  “speech understanding”.
Based on the non-force interaction theory the other model of natural language recognition can be
suggested: “spoken language”  “estimation of non-force (informational) impact on the reaction” 
“reaction (understanding or behavior)” = REFLEX [14].
The principal scheme of the reflex voice control system is presented in figure 4.

255
Input signal

Voice Activity Phoneme module
Detection

Signal with voice

Phoneme Acoustic
recognizer model

Phonemes

Reflex module
State of Reaction
the system recognizer Reflex DB

Reaction

Reaction execution on
device

Figure 4: The principle scheme of the reflex voice control system

In essence, the reflex voice control system consists of two modules. The first module is a
phoneme module. It receives sound signals from the microphone, through the voice activity detection
algorithm detects voice in the input signal, and then transforms the input signal into a phoneme
sequence and sends these phonemes to the reflex module. It is possible to use different phoneme
modules, for example, CMUSphinx or google speech recognition API, or other systems. The second
module is the reflex module. In the input of the module is the sequence of phonemes and in the output
of the module is a reaction. And then the voice control system executes the reaction on the device, for
example, mute sound. The reflex module uses the reflex database and state of the system to determine
the reaction. The module does not include grammar rules or a language dictionary.

256
Because dictionaries and the analysis of the words (grammar) are not needed there is no need to
create a mining model text, such systems are relatively traditional:
• easier;
• cheaper;
• does not require restrictions on the language;
• use a universal algorithm that can be used to create many recognition devices and not just speech;
• easy to learn;
• good to extract any useful information of the speech stream;
• conversely, do not react to the noise, the speech that does not contain any commands, chatting;
• angling for useful information in a stream of free speech;
• allow the course to add new commands;
• respond appropriately to the surrounding speech and voice background (ie, extraneous
information not relevant to the management unit);
• tolerant to the errors in the speech of the speaker;
• allow you to reprogram the behavior of the voice;
• learn from their mistakes;
• perform fixation, analysis, and correction of recognition errors;
• perform speaker identification;
• adjust to the style and habits of the speaker [15].
As can be seen from these results, it possible to speak of the importance of spreading the ideas of
the above in the scientific world. And about the possibility of applying this experience to create online
training systems.
The authors suggest a method for constructing a reflex expert system for managing individual
online training. Such a system will reflect the knowledge of trainees in some software and information
environment, which provides the forming of necessary reactions. Let us introduce several definitions
Definition 1. Informed agent (IA), the model of the trainee in the information base of the reflex
expert system of individual online learning management (RESL) in the context of courses, capable of
forming the RESL reaction.
The informed agent is an interrelated collection of data and knowledge reflecting the real or
abstract trainee, as a result of the interaction of elements of which the RESL reaction is formed.
Definition 2. RESL reaction is a course unit (topic, lecture) that must be studied by the informed
agent.
Write
I  {I x }, j  1, q, (1)
where Ix – informed agent; І – RESL informed agents; q – number of informed agents.
R  {rk }, k  1, w, (2)
where rk – RESL reaction; R – possible RESL reactions; w – number of possible RESL reactions.
I x  I rk  R : ( I x , rk )  true, (3)
where ( I , rk ) – predicate that defines reaction of the informed agent
x

Definition 3. Under the management of online learning, we will understand the automatic
definition of RESL course units (topics, lectures) that need to be studied by the trainee in the course.
Now solve the problem of how to determine the reaction of an informed agent? To do this, we
need to formally submit an informed agent. We assume that an informed agent consists of three data
and knowledge areas:
I x  I1x  I 2x  I 3x , (4)
x
where I 1x – test results (data); I 2x – interview results (knowledge); I 3 – list of elements of the
form of training and / or course units that need to be studied (RESL reaction).
I 3x  {rkl }, l  1, m j , (5)

257
where m j – number of RESL reaction in the informed agent Ix
I 1x  I 2x  { y j }, j  1, s, (6)
where s – number of combinations of trainee responses contained in the results of the
interview/questioning and testing; yj – the combination of the trainee's answers contained in the
results of the interview/questioning and testing.
The combination of answers is understood as the combination of answers of the trainee. For
example, if there were 3 questions in the test (or interview) that were answered:
1) A.
2) B.
3) A.
Then the informed agent will contain a combination of answers:
1A; 2B; 3A; 1A2B; 1A3A; 2B3A; 1A2B3A.
For informed agents who have the same test results and interview reactions should be the same.
I x , I s  I : I1x  I1s , I 2x  I 2s  I 3x  I 3s (7)
But in general, if the course units that need to be studied by different informed agents are the
same, the test results and / or the interview results may not be the same.
I x , I s  I : I 3x  I 3s  I1x  I1s  I 2x  I 2s (8)
So, if some informed agent must determine a comfortable form of training and course units that
need to be studied, it is required in the set of all informed agents to find one that has the same content
in the context of test and interview/questioning results. In general, no such informed agent will be
found. After all, it is quite difficult to find two trainees who know the same discipline. Then, to
determine the RESL reaction, we use the results obtained in the theory of non-force interaction.
We assume that the reactions are formed in the informed agent as a result of the non-force
(informational) influence of the trainee’s knowledge, embodied in the model of the informed agent.
More precisely, in the part of this model that corresponds to the results of testing and
questioning/interview
I 3x  f ( I 1x , I 2x ) (9)
Then using the mathematical apparatus of the theory of non-force interaction, it is possible to
calculate the magnitude of the influence of the state of the trainee, embodied in the results of testing
and interviews on the RESL reactions. And choose the reactions to which the effects are most
significant.
There are two stages in the technology the RESL operation.
1. Stage of training.
1.1. Information agents test results and / or interview results are put into the RESL.
1.2. For each information agent, the teacher indicates which course unit he should study (the
reactions of the RESL are indicated). As a result, with a significant number of information
agents, unconditional and conditional probability of RESL reactions are calculated:
n0x (10)
p0x  ,
m

where n 0x – the number of informed agents for which the reaction Rx was chosen; m – number of
informed agents; p0x – unconditional probability of choosing reaction Rx
n xj (11)
p  w
x
j , j  1, s,
n
k 1
k
j

258
where n xj – how many times the Rx reaction was chosen, provided that the a combination of
answers yj was found in the results of the questioning/interview and testing; n kj – how many
times the Rk reaction was chosen, provided that the a combination of answers yj was found in
the results of the questioning/interview and testing; s – number of different combinations of
answers.
The conditional probability of an Rx reaction occurring provided that a combination of answers yj
was found in the results of the survey/interview or testing.
p xj  p( Rx / y j ) (12)
2. Stage of management.
2.1. From the known probabilities of the RESL reactions, certainty (13) and the awareness (14) of
RESL are calculated regarding these reactions.
𝑝𝑗𝑥 1 − 𝑝𝑗𝑥 (13)
0,5 ⋅ √ 𝑥 + − 2, 𝑝𝑗𝑥 ≥ 0,5
1 − 𝑝𝑗 𝑝𝑗𝑥
𝑑𝑗𝑥 = , 0 < 𝑝𝑗𝑥 < 1, 𝑗 = 0, 𝑠;
𝑝𝑗𝑥 1 − 𝑝𝑗𝑥
−0,5 ⋅ √ + − 2, 𝑝𝑗𝑥 < 0,5
1 − 𝑝𝑗𝑥 𝑝𝑗𝑥
{

1 (14)
i xj  , 0  p xj  1, j  0, s,
2 p xj  (1  p xj )
where d jx , j  1, s – certainty of an Rx reaction, provided that a combination y j was found in the
answers ( d 0x – certainty of reaction in the absence of action on an informed agent).
2.2. The calculation of the total increment of certainty, for all non-force actions on an informed
agent,
𝛥𝑑 𝑥 = ∑𝑠𝑗=1(𝑑𝑗𝑥 ⋅ 𝑖0𝑥 − 𝑑0𝑥 ⋅ 𝑖𝑗𝑥 ) = 𝑖0𝑥 ∑𝑠𝑗=1 𝑑𝑗𝑥 − 𝑑0𝑥 ∑𝑠𝑗=1 𝑖𝑗𝑥 (15)
where d
x
– a total increment of certainty of the informed agent, given by the influence of
all combinations of responses I1x  I 2x .
2.3. The calculation of the increment of awareness of the informed agent (from formula 14)
i x  d   1 ,
x 2 (16)
x
where i – increment of awareness of the informed agent.
2.4. Calculation, according to the formula (13), of the new Rx reaction certainty, which is created
by all combinations of answers I 1x  I 2x
d x  d x  i0x  d 0x  i x , (17)

where d x – new Rx reaction certainty.
2.5. Calculation by the formula (14) of the new awareness of the Rx reaction, which is given by
knowledge of the impact of all combinations of answers I1x  I 2x

ix  d   1 ,
x
2 (18)

where i x – a new awareness of the informed agent.
2.6. From formula (8), we estimate the probabilities of the reaction Rx
dx (19)
p x  p ( R x / I 1x I 2x )  0,5  ,
2i x

259
 
where p x  p Rx / I1x  I 2x – estimation of the Rx reaction probability with actions I1x  I 2x .
2.7. Selection of reactions for which the probability estimation is greater than the threshold value
R x  R : p x  p min, , (20)
where pmin – the minimum acceptable value of the reaction probability estimation.

4. Conclusion

According to the study, methods of data science got the highest priority in online learning
research. In these areas, lots of development has to be made. As technology is growing day by day,
learner interest in online learning increases very rapidly. Now researcher has begun to work on
designing new online learning models that can work according to learners' interests and preferences. It
is a perspective direction of further researches.

5. References
[1] E-learning: fashion trend or education of the future? URL: https://sike.ru/e-learning-modnyj-
trend-ili-obrazovanie-budushchego
[2] E-learning. URL: https://ru.wikipedia.org/wiki/e-learning
[3] Hemant Rana, Manohar Lal, Rajiv: E-learning: Issues and Challenges, 2014.
URL:https://www.researchgate.net/publication/272863519_E-learning_I ssues_and_Challenges
[4] Joi L. Moore, Camille Dickson-Deane, Krista Galyen. “e-Learning, online learning, and distance
learning environments: Are they the same?”. Internet and Higher Education, 14(2011):129–135,
2011.
[5] Oblinger D.G., Oblinger J.L.: Educating the net generation, EDUCAUSE, 2005.
URL:http://net.educause.edu/ir/library/pdf/pub7101.pdf
[6] Lowenthal, P., Wilson, B. G., Parrish, P. “Context matters: A description and typology of the
online learning landscape”. AECT International Convention, Louisville, KY. Presented at the
2009 AECT International Convention, Louisville, KY, 2009.
[7] Rekkedal, T., Qvist-Eriksen, S., Keegan, D., Súilleabháin, G.Ó., Coughlan, R., Fritsch, H.,et al.
“Internet based e-learning, pedagogy and support systems”. Norway: NKI Distance Education,
2003.
[8] Benson, A. “Using online learning to meet workforce demand: A case study of stakeholder
influence”. Quarterly Review of Distance Education, 3(4) (2002): 443−452.
[9] Evans, J., Haase, I. “Online business education in the twenty‐first century: an analysis of
potential target markets”. Internet Research, 11(3) (2001): 246–260.
doi.org/10.1108/10662240110396432
[10] Educational Benefits of Online Learning. URL:
http://blackboardsupport.calpoly.edu/content/faculty/handouts/Ben_Online.pdf
[11] Tuan Nguyen. “The Effectiveness of Online Learning: Beyond No Significant Difference and
Future Horizons”. MERLOT Journal of Online Learning and Teaching, Vol. 11, No. 2 (2015):
309−319.
[12] Malcolm Brown: Learning Analytics: The Coming Third Wave, 2011. URL:
https://net.educause.edu/ir/library/pdf/ELIB1101.pdf
[13] Audrey Watters: How data and analytics can improve education, 2011. URL:
https://www.oreilly.com/ideas/education-data-analytics-learning
[14] Teslia I., Popovych N., Pylypenko V. and Chornyi O. “The Non-Force Interaction Theory for
Reflex System Creation with Application to TV Voice Control”. In Proceedings of the 6th
International Conference on Agents and Artificial Intelligence, 2014: 288-296. doi:
10.5220/0004754702880296
[15] GSOC 2017 Project Overview - Computer Assisted Learning Including Speaking Skills, 2017.
URL: https://cmusphinx.github.io

260