Content uploaded by Aparna Lalingkar
Author content
All content in this area was uploaded by Aparna Lalingkar on Jul 09, 2019
Content may be subject to copyright.
Deriving Semantics of Learning Mediation
Abstract—The web is seen as a promising platform for
designing scalable educational practices across large
populations. Many of the efforts in this space use the web
primarily as an amplifier over existing models of learning that
are based on the classroom. In this paper, we propose a
pedagogic model called mediated learning where the web acts
as a platform for nurturing a learning community by
continuously mediating between knowledge need and expertise.
Mediated learning has the potential to invert the learning
pyramid by interfacing the learner with several experts as part
of a single learning experience. For supporting mediated
learning, an approach is needed that is data-intensive and
driven by social semantics. This paper outlines the proposed
pedagogic model, which comprises two primary components: a
user-end navigator component that provides a rich interface
enabling users to independently navigate through a learning
space; and a back-end community component, that performs
meaningful mediations between participants in the logical
learning space.
Keywords- mediated learning; semantics; ontology;
personalized learning; suugest system, open learning
I. INTRODUCTION
The increasing use of the web in day to day life has led to
its use for designing scalable educational models for online
education. The MOOC (Massive Online Open Course) [1]
model of learning is the idea of the classroom scaled over a
massive student population of all ages. The factory model of
the classroom can be described as a pyramid representing a
one-to-many relationship between the teacher and the
students, called the teacher-student ratio (TSR). Other
models of online learning, such as Intelligent Tutoring
System (ITS) and Adaptive Learning Environments (ALE)
[2, 3, 4], represent a one-to-one TSR. The MOOC has been
criticized for problems such as high dropout rates, lack of an
economically sustainable model, de-skilling the
professoriate, and cheating and plagiarism [5]. The major
criticism of ITS and ALEs is that they focus on creating
automated tutor agents that replace human teachers by using
cognitive science, learning science and artificial intelligence.
The proposed model enables directed, personalized
learning through a navigator which allows users to curate
their own learning experience through a web browser or a
mobile smartphone. We propose a pedagogic model based on
mediated learning [6] that is used in the proposed system,
where the web acts as a platform for nurturing a learning
community by continuously mediating between knowledge
Figure 1. Inversion of the learning pyramid: Classroom, ITS and the
Proposed pedagogy model
need and expertise through a mechanism called suggest
system. Mediated learning has the potential to invert the
learning pyramid by interfacing the learner with several
experts as part of a single learning experience. The proposed
model of learning represents a many-to-one TSR that inverts
the learning pyramid (please see figure 1). This model avoids
the problems and leverages the benefits of MOOCs, ITS and
ALEs.
In this paper, we first describe the suggest system that
makes interventions in the proposed model. Then we discuss
its deep-end architecture in the light of mediated learning,
giving an idea of how to derive the semantics of learning
mediations. This paper describes the position of the research
undertaken for the proposed system.
II. SUGGEST SYSTEM
The proposed pedagogic model has two components. The
first is the learning navigator, which helps a learner to
navigate through a logical space, giving a smooth learning
experience, and to get an overview of the learning space for
any subject. The second component is the back-end
community engine, which gives all users the opportunity to
interact and contribute to the logical space on which the
learning navigator operates. This provides a platform for
curating content using open educational resources and
resources created by the content team that are called
signature collections. The assessments created by content
team are called signature assessments. The learning space for
each subject is defined as a two-dimensional co-ordinate
system in which domains or topics are considered as X-axis
and pedagogic depth is considered as Y-axis. Each point in
this co-ordinate system is considered as a competency to
which a learning activity or a collection is attached. While
curating the content in the proposed system, a teacher can
specify a learning path. The system also suggests a learning
path based on a learner profile built by its deep-end
combined with the information obtained through the Event-
Condition-Action (ECA) rules. We define an event as any
situation which the learner faces, either within the system or
outside the system. The conditions are the information points
obtained from the learner profile and learning activity vector.
The event trigger is an element (a combination of events and
conditions) that can trigger thinking or learning. suggest
interventions (i.e. mediations) can be seen as manipulations
of pathways. The suggest system intervenes in several such
workflows on a continuous basis, based on semantics
extracted from systemic activities. Suggest algorithms are
triggered by ECA rules embedded in different parts of the
system.
Here is an example of an event: Leena, a 9th Grade
student, takes a photo of a flower at Lalbagh botanical
garden, and her phone GPS sends co-ordinates of her
location to the system. Leena’s profile information, such as
she has mastered flora identification; she has learned
independently Botany-I with 60% proficiency; she has
studied the history of Karnataka with 70% proficiency as a
part of her classroom learning; she prefers videos and
images, could be considered as conditions. Combinations of
any given event with given conditions could be considered
an event trigger that can initiate one or more suggest
interventions. The suggest trigger may intervene not just in
the student’s learning pathway, but also in the workflows of
other pertinent stakeholders like the teacher, course content
creator, etc. A system of Notifiers and Listeners manages
ECA rules by subscribing to different forms of event
notifications and checking the relevant condition to finally
call a pertinent suggest method. The suggestion of learning
pathways, i.e. the mediation offered, will be determined by a
thorough understanding of the learner – including their
progress, performance, proficiency, preferences, portfolio,
markers, history, and goals – and of the learning activity –
including its relevance, engagement and efficacy – providing
a consistent learning experience throughout.
Basic prerequisites for achieving this vision include the
capability to locate the learner, curate potential learning
activities, and make appropriate suggestions based on their
context. The suggest sub-system leverages key elements
from learning science, cognitive science, neuroscience and
data science to design the most precise learning pathway for
every learner. This requires multivariate optimization across
different learning pathways based on learning, cognitive and
data sciences. From a theory perspective, Learning Navigator
blends elements of Piaget’s constructivism [7] and
Vygotsky’s zone of proximal development (ZPD) [8] with
the cognitive science results on elements of surprise,
suspense and conflict that produce cognitive engagement [9];
with research into the factors that persuade learners to stay
on task and develop grit and perseverance; and with a data
science [10] approach for finding coherent learning
sequences. All learning activities are annotated with data and
metadata. These activities will be scored for the specific
learner, and the top-ranking one will be suggested as the next
activity in their learning pathway.
Based on the mastery of competency model, the ZPD
will be computed and a learning route will be suggested. The
learning route will be modified by finding the most coherent
sequence of the learning activities. This sequence of
activities will be ranked by considering learning principles
from cognitive and neuro-science, i.e. surprise, suspense, or
the positive addictiveness of a resource. The ranked learning
pathways will be the mediation offered by the system. The
ECA rules will provide another mediation of learning
resources: either a learning pathway or an individual learning
resource. For communication between the actual suggestions
- i.e. learning pathways or single learning resource - and the
event triggers, various ontological models would be used.
Figure 2 depicts the logic behind the suggest system.
Figure 2. Suggest system Logic
III. SEMANTIC PROCESSING FRAMEWORK (SPF)
The suggest system of the proposed system provides
semantic processing and intervention into learning pathways
that are managed by the learning navigator. All data related
to the student activities (classroom study or independent
learning) are integrated in the back-end. All analytics and
interventions of the suggest system is based on semantic
rather than systemic considerations: that is, they are meant to
provide a rich cognitive experience to the user, rather than
optimize the operational performance of the system. As a
result, suggestions are often provided based on materialized
Figure 3. The 4-layer cake architecture of SPF
semantic associations across different entities that are
computed by the Semantic Processing Framework (SPF) –
sometimes also called the deep-end of the overall system.
SPF operations involve generating various kinds of semantic
associations by reasoning over data coming from learning
resources and learning activities. This involves extracting
semantic associations between disparate elements from
datasets that are large in volume, unstructured or loosely
structured, of different varieties, representing potentially
high-velocity events, and even of questionable veracity (the
archetypal V’s of Big Data). Figure 3 depicts the overall
architecture of the SPF. This is in the form of a 4-layer cake
that is in turn divided into two broad realms, called
perception and cognition realms, respectively. Perception
involves identification of (perceiving) semantic entities and
associations from raw data. Cognition, on the other hand,
involves making sense of the perception and acting on the
derived sense. This involves elements like interpreting the
perception, explaining the observation, analyzing the
situation, computing implications, designing strategies, etc.
The perception realm is concerned with recognizing
different objects of interest from the raw data, as well as
relationships and dependencies across them. This realm
comprises the Data and Analytics layers. The Data layer
reads the raw data from different data sources, such as
activity data from learning activities, resource data from the
resource repository, and system logs. This layer comprises a
set of applications that perform data cleaning,
canonicalization, ETL (Extract-Transform-Load) operations,
view materialization, etc. The Analytics layer implements
algorithms to convert raw data into semantic tokens. It
contains algorithms for Named Entity Recognition (NER),
classification of resources, identification of association rules
and other interesting association patterns, computing
aggregates like activity vectors and learner vectors, etc.
The output of the Analytics layer forms the input to the
Cognition realm of the SPF. The Cognition realm in turn
comprises two layers, called the Semantics and Mediation
layers respectively. The objective of the Semantics layer is to
interpret the various patterns that were mined by the
Analytics layer. The Semantics layer includes topic
modeling, computing activity and learner siblings and
computation of narrative arc. Insights obtained from the
Semantics layer are used by the Mediation layer to design
semantic interventions, i.e. suggestions to the system. The
Mediation layer would be using an ontology bank. The SPF
is inspired by these concepts from cognitive science. There
are several theories about the interplay between semantic and
episodic memory systems, addressing phenomena and
intermediate structures involved in the distilling of general
semantic knowledge from episodic experiences, and the role
of general semantic knowledge in cognitively driven actions
in specific episodic contexts [11, 12, 13].
IV. CONCLUDING REMARKS AND FUTURE WORK
In this paper we have proposed a suggest system and the
deep-end logic behind it for deriving semantics to provide
mediated learning. The SPF is modeled as a cognitive being
that is continuously observing the proposed system
comprised of learners, resources and learning activities. The
proposed learning navigator actually locates a learner and
navigates her towards her learning goal by offering
personalized scientifically designed learning pathways. The
proposed platform considers any situation as a learning
situation and models it into a scientific learning experience,
and hence can make a learner fall in love with learning. The
proposed system leverages the power of technology to
amplify the learning ability of a learner, and also provides
an experience of being part of a learning community of
teachers, experts and peers. The implementation of these
ideas is planned in the future research work.
REFERENCES
[1] L. Breslow, D. E. Pritchard, J. DeBoer, G. S. Stump, A. D. Ho, and D.
T. Seaton, “Studying learning in the worldwide classroom: Research
into edX’s first MOOC”, Research & Practice in Assessment, vol. 8,
2013.
[2] J. R. Anderson, C. F. Boyle, and B. J. Reiser, “Intelligent tutoring
systems”, Science (Washington) vol. 228 (4698), 1985, pp. 456–462.
[3] A. Paramythis and S. Loidl-Reisinger, “Adaptive learning environments
and e-learning standards”, In 2nd European conference on e-learning,
Vol. 1, 2003, pp.369–379.
[4] A. C. Graesser, M. W. Conley, and A. Olney, "Intelligent tutoring
systems", In APA educational psychology handbook, vol. 3, 2012.
[5] G. Siemens, “Massive open online courses: Innovation in education.
Open educational resources”, Innovation, research and practice, vol.
5, 2013, pp. 5-15.
[6] A. Kozulin, “Psychological tools and mediated learning”. In
Vygotsky’s educational theory in cultural context, 2003, pp. 15-38.
[7] B. J. Wadsworth, “Piaget's theory of cognitive and affective
development: Foundations of constructivism”. Longman Publishing,
1996.
[8] P. E. Doolittle, “Vygotsky's Zone of Proximal Development as a
Theoretical Foundation for Cooperative Learning”, Journal on
Excellence in College Teaching, vol. 8(1), 1997, pp.83-103.
[9] A. Ortony, G. Clore and A. Collins, “The Cognitive Structure of
Emotions”, Cambridge Uni. Press, New York, 1988.
[10] A. R. Rachakonda, S. Srinivasa, S. Kulkarni, and M. S. Srinivasan, “A
Generic Framework and Methodology for Extracting Semantics from
Co-occurrences”, Data Knowl. Eng., 92, 2014, pp. 39–59.
[11] D. L. Greenberg and M. Verfaellie, “Interdependence of episodic and
semantic memory: evidence from neuropsychology”. Journal of the
International Neuropsychological society, vol.16 (5), 2010, pp. 748–
753.
[12] L. Renoult, P. SR Davidson, D. J. Palombo, M. Moscovitch, and B.
Levine, “Personal semantics: at the crossroads of semantic and
episodic memory”, Trends in cognitive sciences, vol. 16 (11), 2012,
pp. 550-558.
[13] M. L-H V˜o and J.. M Wolfe, “The interplay of episodic and semantic
memory in guiding repeated search in scenes”, Cognition, vol.
126(2), 2013, pp. 198–212.
