Design Elements in a Personal Learning Environment

In this paper I would like to address the core design elements in the development of a personal learning architecture being developed in the National Research Council's Learning and Performance Support Systems program. This program was developed and approved to address the issue of skills shortages in technical and professional industries in Canada. It is an issue that costs Canadian industry billions of dollars a year while thousands of Canadians remain unemployed. Our solution is to provide each person with a single point of access to all their skills development and training needs, individualizing their learning path, providing learning support, and supporting learning tailored to industry needs and individual performance support.

This program builds on the National Research Council's deep connection to the e-learning industry, including collaboration and commercialization across the sector. The program draws on NRC's research in other fields, such as machine learning and analytics. And NRC is free to take risks on technology that might daunt commercial providers. NRC's track record in this sector includes the leadership role it played in the eduSource network of learning object repositories, the Sifter/Filter content recommender later commercialized as Racofi, sentiment analysis in learning, the Synergic3 collaborative workflow system, and more.

NRC's Learning and Performance Support Systems program touches on all parts of Canada's learning technology, but has the most direct impact on the learning management system (LMS) sector. This is an area that includes content management systems, talent management systems, and the LMS. It also impacts content developers and e-learning distributors, including MOOC distributors and educational institutions. It also impacts end users themselves: not only students and individual learners, but also their employers.

In recent years NRC has become widely known for developing and refining the Massive Open Online Course (MOOC), including the creation of the technology behind the original Connectivism and Connected Knowledge (CCK08) MOOC offered in 2008, creating a dynamic connected application to support learning. The MOOC combined several themes which were in themselves becoming increasingly important: the idea of massively multi-user environments, the idea of using open and distributed content, the idea of fully online delivery, and the packaging of these as an online course.

The NRC-designed MOOC differs significantly from traditional courses. The most obvious difference is that the course is not located on a single platform, but is instead a web created by linking multiple sites together. The architecture of this web is intended to optimize four design principles: each member of the web operates autonomously, the web links diverse services and resources together, the web is open and supports open engagement, and the web encourages cooperative learning.

As mentioned, the first course offered using this model was CCK08, delivered through the University of Manitoba and taught by George Siemens and Stephen Downes. It attracted 2300 students and connected more than 170 distinct online resources during its 12 weeks. In the years that followed CCK would be offered three more times. Additionally, the same platform was used to deliver a course called Personal Learning Environments, Networks and Knowledge (PLENK) in 2010, as well as the 30-week course on Change.

In 2011 Stanford University offered its first MOOC, the Artificial Intelligence MOOC authored by Norvig and Thrun. It differs from the network-based connectivist MOOCC (cMOOC), though, by being centred on a single platform and focusing on content like a traditional course. The xMOOC, as this model came to be known, is characterized by limiting autonomy and diversity - all students followed the same lessons at the same pace. Although it was open, interaction flowed one-way, from professor to student.

It is also worth contrasting the pedagogy of the cMOOC from the xMOOC. Engagement is at the core of cMOOC learning. Participants aggregate resources from multiple sources, remix these in various ways, adapt and repurpose them to their own needs, and then share them. If we look at the structure of the course from this perspective, we see a network of individual learners interacting with each other and exchanging, and working with, diverse resources obtained from a variety of internet sources.

Looked at more deeply we can describe specific support requirements for each student. A student creates a resource, and makes this available to the course where it is accessed by a second student, who via this resource finds a third student's resources. From the course provider perspective, students contribute content metadata and the learning provider may create additional content, all of which is accessed and shared by course participants, who may also attend live online events or access event recordings. From the student's perspective, by contrast, the view is to a set of other students or course instructors, and via interactions with these course participants, to a wide range of resources and services across the wider internet, everything from blog posts to YouTube videos.

To support a student's involvement, therefore, technology design is based on the idea of putting at the centre of a learning network, connecting via a single environment to other participants, course resources, and myriad online services. This in turn suggests a simplified design that supports this student-centered approach with connections to learning support applications, and in particular, to resource repositories, to external cloud media storage, to learning applications and APIs, and to external graph-based analytics. These components form the core of the Learning and Performance Support Systems (LPSS) technology development proposal, which incorporates these connective elements with a personal learning record to support lifetime management of credentials, training records, and learning activities, and a personal learning assistant to manage the system.

The NRC LPSS program is a 5-year $20 million effort designed to develop these core technologies and bind them with a common platform. The program applies this technology through a series of implementation projects with commercial and technical partners, including other NRC and Government of Canada (GoC) branches. These projects are managed through a program organization that maps the technology effort to client demands and the employment outcomes described at the beginning of this paper. Program deliverables include not only the technology development, which will be implemented in corporate, institutional and government environments, but also a series of publications and white papers describing the LPSS learning network, how and why it works, and how to connect to it.

Also LPSS can be viewed as a stand-alone system, it is designed in a distributed and modular fashion in order to enable it to be inserted, for example, directly into work environments and corporate contexts, directly addressing human resources and training requirements. This interoperability is achieved through the personal learning assistant (PLA). Like an LMS, the PLA displays learning resources and plays interoperable learning technology (using standards such as ADL's SCORM or IMS's LTI). But it also the leading edge to much more. As mentioned above, the LPSS program is developing five core technologies, linked by the Common Framework (CF). These are the aforementioned PLA, the Resource Repository Network (RRN), Personal Cloud (PC), Competency Development and Recognition Algorithms (ACDR), and the Personal Learning Record (PLR).

Let us examine these in more detail. The first of these is the Resource Repository Network (RRN), needed to provide connectivity with external resources. This package of applications enables a user to manage and discover lists off sources and resources. In a sense, it functions like the syndicated content (RSS) readers of old, but is designed to access and manage many different forms of content, including calendar information and modern Javascript-based (JSON) descriptions of courses and programs.

A second aspect of LPSS is the Personal Cloud (PC) set of applications. These applications manage personal cloud storage services. Some of these are familiar, such as Dropbox and Google Drive, and some of these are innovative, such as personal home-hosted cloud storage using OwnCloud. But more is involved than merely storing data; resources must be secured, backed up, authenticated and synchronized. This enables LPSS to support genuine data portability, and eliminate reliance on a single provider.

As mentioned above, interoperability is achieved through the Personal Learning Assistant (PLA). In addition to displaying learning resources and running e-learning applications, the PLA is designed to 'project' LPSS capacities into multiple platforms. These include not only desktop and mobile devices, but productivity applications such as Word and PowerPoint, interactive environments such as conferencing systems and synchronous communications platforms, simulations and games, as well as tools and devices. The PLA exchanges information with these environment, enabling them to interact intelligently with the user. One example of this kind of integration is LPSS's integration with another NRC product called 2Sim, which provides virtual haptic training simulations in medical environments. By exchanging activity data (using the Experience API, or xAPI data exchange format) LPSS supports a continuous learning path using these systems.

This points to an additional set of services that can be integrated into a distributed learning application, Automated Competency Development and Recognition (ACDR). This is a set of intelligent algoritms designed to import or create competency definitions matching employment positions, to support the development of learning plans based on these competencies, to provide resource and service recommendations, and to tackle the seriously challenging task of assessing performance based on system and network interactions. It is worth noting that while LMSs and xMOOCs tout learning analytics, only a distributed personal learning network application can apply analytics using a person's complete learning and development profile, and not only the specific LMS or cMOOC.

This functionality is enabled by the Personal Learning Record (PLR), which collects learning records and credentials obtained through a lifetime and stores them in a secure locker owned by the individual and shared only with explicit permission. The PLR collects three major forms of records: learning activity and interactivity records, such as xAPI records; a person's personal portfolio of learning artifacts and evidence; and the person's full set of credentials and certifications, these verified by the issuer.

It should be noted that LPSS recognizes, and is designed to cooperate with, existing personal learning environment and personal learning records, including Europe's Responsive Open Learning Environments (ROLE) project and start-ups such as Known, Learning Locker and Mahara. Additionally, LPSS is designed to work with MOOC providers - not only NRC's gRSShopper but also Coursera and EdX. We've integrated badges in a Moodle and Mahara environment for the Privy Council Office, we're doing xAPI application profile development, and are engaged in collaborative workplace training and development. These implementation projects (as we call them) reinforce LPSS's mandate to be more than just a theoretical exercise, but to apply the technology in authentic environments, supporting individuals in a learning network and feeding this experience back into product improvement.

It may be suggested that there are any number of companies engaged in aspects of learning analytics, personal learning records, learning technologies integration, and the like. But the LPSS approach is different - by creating many small things linked together instead of one large centralized application, many tasks that were formally simple - like data storage, content distribution, authentication and analytics - become that much more difficult. Take analytics, for example - how do you do big data analysis across thousands of separate systems each with its own unique data structure? These are the hard problems NRC is trying to solve.

LPSS launched in an initial pre-alpha version October 1, 2014. Invitations may be obtained by going to http://lpss.me and filling in the short form. Users will also be asked whether they would like to participate in LPSS development research (this is not required and all personal research is subject to strict Government of Canada research ethics protocols). Functionality in this early system is limited; the first release focused on content aggregation, competency import and definition, and simple recommendation.

The next release (March 31, 2015) will feature the 'connectivist' social interaction architecture being designed through an implementation project with the Industrial Research Assistanceship program (IRAP) supporting small and medium sized enterprise. The roadmap projects two other major releases, at 6-month intervals, coupled with ongoing client-specific and industry-specific learning solutions. Technology will be transferred to partner companies beginning in 2017.