The paper was presented at ICALT 2013: http://infoscience.epfl.ch/record/185963?ln=en
This paper proposes a novel approach to build and deploy learning analytics dashboards in multiple learning environments. Existing learning dashboards are barely portable: once deployed on a learning platform, it requires considerable effort to deploy the dashboard elsewhere. We suggest constructing dashboards from lightweight web applications, namely widgets. Our approach allows to port dashboards with no additional cost between learning environments that implement open specifications (OpenSocial and ActivityStreams) for data access and use widget APIs. We propose to facilitate reuse by sharing the dashboards and widgets via a centralized analytics repository.
4. Challenges
Observe, control and adjust the learning process
Consider individual capabilities and preferences
Opportunities
Remote education is mediated by a digital environment
Learners leave a ‘digital footprint’
Increasingly more traces can be recorded
How to meet the challenges?
Technology Enhanced Learning
4
Data should be used to improve learning
5. What is Learning Analytics
“LA is the measurement, collection, analysis and reporting
of data about learners in their context, for purposes of
understanding and optimizing learning and the environment
in which it occurs” – Siemens et. al.
Examples
Student drop out prediction systems
Live statistics about the learners
Individual progress vs group progress
Learning Analytics Dashboards
5
LA is becoming an essential component of a learning
system
10. What is portability?
“Software portability is the ability to run the same software on
different platforms with no or little effort”
Why is analytics portability important?
Switching the platform or using a few simultaneously
Persisting the same learning analytics environment
Portability prerequisites
Access the data in the same way on different platforms
Visually represent the UI components in the same way
Learning Analytics Portability
10
Unified data access and portable dashboard representation
11. Tightly coupled LA solutions
Developed for a specific platform, e.g. Blackboard
Platform-specific, usually proprietary APIs
Do not work on other platform
Pluggable LA solutions
Integrated into a platform via a plugin interface, e.g. Moodle
Not compatible across platforms
Standalone LA solutions
General web analytics services, e.g. Google Analytics, Woopra
Services developed specific for LA, e.g. CAM web service
Learning Analytics Architectures
11
Existing LA dashboards are barely portable
14. Open Specifications For Data Access
14
ActivityStreams unifies access to learners’ activity data
JSON
Specification
15. Open Specifications For Data Access
15
OpenSocial and ActivityStream unify data access
Social features of a LMS should be accessible as well
Relationship between users, resources and tools
OpenSocial specification
Describes common method for accessing social data
Defines a set of common APIs
OpenSocial is adopted by
Ning
MySpace
Orkut
Some TEL solutions: Sciverse, ROLE
16. Widget is a lightweight web-based application
OpenSocial widgets rendered with Apache Shindig
Learning Dashboards as Metawidgets
16
Widgets can be used to implement portable components
17. Widgets can be combined in a metawidget
Learning Dashboards as Metawidgets
17
Metawidgets provide portable dashboard representation
20. Portability is best achieved with open standards
Our solution uses ActivityStreams, OpenSocial and widgets
Already implemented in Graasp
Integration into Moodle is in progress
Can be integrated into other learning environments
Code is available on GitHub
See: http://github.com/react-epfl/
Summary
20
21. 21
Thank you for your attention!
Feedback & Question?
Andrii.Vozniuk@epfl.ch
Editor's Notes
Good morning, I’m happy to see you all here, thanks for coming. My name is Andrii Vozniuk, I’m from REACT lab of Federal Polytechnic school in Lausanne, Switzerland.
Nowadays we observe a change (a shift) happening in the way we learn. Education transforms from
having a well-defined and constrained path to more individual and flexible. Education is also becoming increasingly distributed (global) in nature. With the recent raise of MOOCs and other distance learning technologies, taking courses outside of the associated institution is becoming a common practice.
Nowadays we observe a change (a shift) happening in the way we learn. Education transforms from having a well-defined and constrained path to more individual and flexible. Education is also becoming increasingly distributed (global) in nature. With the recent raise of MOOCs and other distance learning technologies, taking courses outside of the associated institution is becoming a common practice. This shift creates new challenges that need to be addressed. It becomes much harder for teachers to observe, control and adjust the learning process. For instance, in the MOOCs it is simply impossible for a teacher to consider individual capabilities and preferences of each learner.
But together with challenges there come opportunities. These days remote learning is usually mediated by a digital media which makes it possible to observe learning process in details. This observation is usually done by tracking actions of students in a learning platform and generates their digital traces. A natural intentions is to collect this data and use it to improve learning.
[Now there is a common understanding that data should be used to improve learning]
Increasingly more digital tools are being used in a classroom, which allow to collect traces of students
Leveraging educational data
This shift creates new challenges that need to be addressed. It becomes much harder for teachers to observe, control and adjust the learning process. For instance, in the MOOCs it is simply impossible for a teacher to consider individual capabilities and preferences of each learner.
But together with challenges there come opportunities. These days remote learning is usually mediated by a digital media which makes it possible to observe learning process in details. This observation is usually done by tracking actions of students in a learning platform and recording their digital traces. A natural intentions is to collect this data and use it to improve the learning.
[Now there is a common understanding that data should be used to improve learning]
Increasingly more digital tools are being used in a classroom, which allow to collect traces of students
Leveraging educational data
That’s exactly the goal of recently emerging field of learning analytics. So what is learning analytics? According to Siements et. al. LA is “the measurement, collection, analysis and reporting of data about learners in their context, for purposes of understanding and optimizing learning and the environment in which it occurs”.
There is a strict definition of the LA on the slide but what it says is that the goal is to make use of the data to improve learning. Such an improvement could possibly be an early prediction of a drop out, or just informing the teacher regarding progress of a class, or just showing to student their performance compared to the distribution of a class to make them more aware of their results.
Numerous studies show that integrating learning analytics into the learning process could improve the outcome. But existing popular online learning management systems (LMSs) and personalized learning environments (PLEs) lack or provide limited LA dashboards.
Let’s have a look how LA component is integrated into existing learning systems.
TODO: LA component often comes in a form of a learning dashboard – show few examples
Self-monitoring for learners
Awareness for teachers and administration
It works
And indeed it is being integrated in one way or another into existing learning systems. For example, here we see Blackboard, a popular system. It has a learning component. Here we show an example of a component to increase student’s awareness and motivate them by showing their progress relatively to the progress of classmates.
Another example is a Moodog plugin for Moodle which presents visualization of student activities based on Moodle logs.
TODO: Replace with Purdue signals
That’s exactly the goal of recently emerging field of learning analytics. So what is learning analytics? According to Siements et. al. LA is “the measurement, collection, analysis and reporting of data about learners in their context, for purposes of understanding and optimizing learning and the environment in which it occurs”.
There is a strict definition of the LA on the slide but what it says is that the goal is to make use of the data to improve learning. Such an improvement could possibly be an early prediction of a drop out, or just informing the teacher regarding progress of a class, or just showing to student their performance compared to the distribution of a class to make them more aware of their results.
Numerous studies show that integrating learning analytics into the learning process could improve the outcome. But existing popular online learning management systems (LMSs) and personalized learning environments (PLEs) lack or provide limited LA dashboards.
Let’s have a look how LA component is integrated into existing learning systems.
TODO: LA component often comes in a form of a learning dashboard – show few examples
Self-monitoring for learners
Awareness for teachers and administration
TODO: higher resolution
And the third example that I would like to show is Google Analytics, which is often used for online courses to monitor student behaviour. The common between the demonstrated approaches and actually the common way to deploy analytics is to build an analytic dashboard.
Summary: Existing learning dashboards are barely portable: once deployed on a learning platform, it requires considerable effort to deploy the dashboard elsewhere.
Some configuration might still be required
Cost is considerably lower than building from scratch
Why portability?
Portability offers freedom
Openness leads to better adoption
understand the concept of portable learning dashboards, we first define software portability in general. Software portability is the ability to run the same software on different platforms with no or little effort. Some configuration of LA tools on a new platform might still be required but its cost is considerably lower than developing the tools from scratch.
Our goal is to achieve portability be implementing learning dashboards as external pluggable components and using well-defined interfaces with a learning platform.
TODO: should better explain why portability is important. Now it’s hard to understand why we are working on this problem.
Learning system perspectives:
Structural View
Dynamic View
TODO: Add three images with blocks to show different types of architectures
Our goal
Implement LA dashboard as a pluggable component
Use standardized and open interfaces with the platform
Hence learning systems are extended with plugins or augment
Summary: Existing learning dashboards are barely portable: once deployed on a learning platform, it requires considerable effort to deploy the dashboard elsewhere.
Ecosystem lock-in
Data is locked
Analysis tools are locked
In order for LA tools to function, a learning platform must provide access to its stored data. Such data represents dynamic (behavioural) and static (structural) views of the platform.
In order for LA tools to function, a learning platform must provide access to its stored data. Such data represents dynamic (behavioural) and static (structural) views of the platform.
In order for LA tools to function, a learning platform must provide access to its stored data. Such data represents dynamic (behavioural) and static (structural) views of the platform.
TODO: Visualize widgets and metawidgets
TODO: Visualize widgets and metawidgets
Classical three tiers
Built upon open standards
Can work for existing systems
Customizable
TODO: Ask Evgeny regarding support of OpenSocial in Moodle
To sum up, we are strong believers that the openness and portability of learning analytics tools will increase their adoption. We also believe that to make a solution (architecture) adopted it is required to build it upon open standards. Existing learning dashboards don’t promote are barely portable: once deployed on a learning platform, it requires considerable effort to deploy the dashboard elsewhere.
Hence in this talk we proposed a novel approach to build and deploy learning analytics dashboards in multiple learning environments. Our approach allows to port dashboards with no additional cost between learning environments that implement open specifications (OpenSocial and ActivityStreams) for data access and use widget APIs.
We suggest constructing dashboards from lightweight web applications, namely widgets.
We propose to facilitate reuse by sharing the dashboards and widgets via a centralized analytics repository.
TODO: what the systems that wants to make a use of it will need to do?