AI Meets Competency-Based Learning

The Curious Case of Learning Online

???? Why do we use the internet? 30 years ago, a question like that would have very few limited answers, especially for users who do not work in software engineering or any computer-related field. Today, there are countless purposes that we use the internet for. Take a quick look at the graph below and notice the change in the number of internet users since the 90s.

It is clear how much internet users grew in number globally from less than half a billion users in the year 2000 to almost 3.5 billion in 2016. This breakthrough has brought lots and lots of merits with it that I am sure we all are well aware of. In the past decade, the demand has obviously increased for using the Internet for educational purposes. Since then, the never-ending debate of “Which is better, learning online or F2F?” has aroused. However, this is not our concern now because there is another case that is worth discussing; did the internet make learning an easier or harder process? Most people would raise a skeptical eyebrow questioning my sanity and answer “Of course easier!” I wouldn’t disagree but such easiness comes at a price. There are actually many challenges to learning online that go beyond the ease of access to educational content, and such challenges could yield to opposite results for knowledge seekers. We will take a look at some of the challenges that face people who want to learn using the internet from a knowledge acquisition perspective.

The Internet of “Too Many” Things

???? Imagine entering your kids’ room to search for a little piece missing from one of their toys to be stunned by the mess inside. Everything is scattered all around the room, and as you start walking inside, you stumble every inch. You barely recognize toys from clothes, books, papers, and pencil colors and a zillion of other stuff everywhere. What is your chance of finding the piece that you need? That same experience happens almost every time you attempt to learn something from the internet.

This overloaded and messy space of information has made learning easier to reach but it is more like searching for a needle in a haystack. Can you believe that over?4 million blog posts are published?on the internet and over?5 billion Google searches are made every day?! [1]. Getting lost or finding dozens of barriers to focus while trying to learn online will result in losing interest in learning anything.

When you have a reference book on a specific topic, you will find a table of contents for this book, chapters, and subtopics that build up the structure of the book so that readers have an organized flow of information. If you searched that same topic on the internet, you will probably get ten times the content you found in that book; in articles, in other books by different writers, in videos, in blogs, in research papers, and perhaps in full courses. Now, the too many “Wow all these search results” will test your patience on how are you going to deal with this noise to start learning properly.

To Read or Not to Read? That’s the Question

????? ????We mentioned before that the base of internet users has greatly expanded over the last 10 years. As of June 2018, more than 55% of the world's population had access to and used the internet [2]. Indeed, not all of them use the Internet for training or learning purposes, but it is enough to mention that the worldwide market of online learning is expected to exceed 243 billion U.S. dollars by 2022 compared to 46.67 billion U.S. dollars?back in 2016 [3]. This can give you an insight into the potential of learning online and how huge the market is.

???? That old problem of knowledge accessibility never exists anymore, since resources for learning and content have become only a few clicks away. You don’t have to search for libraries or wait for book fairs or consult university professors to get materials to learn. However, the problem has become what exactly to learn, and what content should or should not be read from the myriad articles and topics online. Such an explosion of content can sometimes be frustrating for eager learners who want to cease every moment to learn something new online that is accurate and authentic. When you create a learning routine for yourself and plan to learn something new to enhance your skills in order to cope with job market demands, you start wondering “Is this a good website to learn about this topic? Is it the only one? What if there are better articles that explain the same topic in an easier way? Is this article for beginners or advanced learners?” It is hard for the learners to be sure they are learning the right thing or know if they are on the right track towards achieving a certain objective or not. The issue of not completely trusting in whatever you are reading is real and has a very bad impact.

???? How many times have you read something online, whether it was a piece of news, article, data, or statistics, and found out that there are other sources contradicting the information you read? Click-bait websites and fake content online have become a business right now. Some people just aim to have huge traffic on their websites by publishing content that is either fake, void of any real knowledge, or that is very shallow to actually learn something from it. This type of “harmful” content is increasing day by day forming lots of clouds around real and useful content out there and this, in turn, increases the dilemma of learners.

It’s About Time

???? Learning online is all about time. It is the ideal way to learn on the go; anywhere at any time and avoid wasting the time of F2F learning. This flexibility in learning has attracted more audiences to join the crew and start learning online, especially with the increasing percentage of mobile learning. As a matter of fact, 67% of people now use mobiles to learn [4]. However, if you are not enrolled in an online course with specific learning outcomes that match your needs you will have to jump from one source of content or one website to another acquiring all the knowledge and gathering the information you want. Unfortunately, this is not exactly what happens. You probably end up wasting a lot of time reading an article thinking that it will be useful or contain valuable content, but the result is like “Oh...that was a total waste of time!”

The more the learner will acquire during the amount of time he/she dedicated to learning online, the more efficient and valuable the online resource is. If you decide to start a habit of dedicating 2 hours before sleep to learning something new, and 60% of those 2 hours are wasted without a real outcome, you will end up with a no-accomplishment feeling and perhaps decide not to continue with the habit. The fake content and click-bait problem that we mentioned before and that biases learners is also time-consuming and makes the learning experience boring. Since online learning should be time-saving, it is about time to think of a solution to eliminate those obstacles hindering the learning experience and make every learning moment count.

Apply the Solution to the Injury

???? When you start a new job, usually there will be an onboarding period through which one of your co-workers or the HR will introduce you to the workflow, the process involved in the new company, and if there are any important guidelines and documents to be acquainted with before you actually commence working. You will probably hear things like “Start first by reading this document, take a look at those presentations, go through the files in this folder…etc.” This chronological order is what makes the onboarding process much easier than just logging onto your computer and start exploring the company documents and files on your own. Breaking down the tasks into simpler mini-tasks will save you a lot of time, and make you follow a logical path that has been set by those with previous experience in the company to learn only the needful at this point in time without getting distracted by the huge amount of files on the company server.

???? In practice, a training model can be easily built to use the “recommendation” technique to provide personalized learning based on competencies., This manages to overcome the previously mentioned challenges for learning online; the huge amount of content, accuracy or relevancy of content, and time consumption.

Simply put, it’s like having an expert recommending to you what to do step-by-step if you need to learn about a specific topic, and getting suggestions on the optimum way to do it in a logical order and instructional sequence in addition to providing a high level of engagement with the content to guarantee a seamless enjoyable learning experience. The closest learning style to the one we are adopting is Competency-Based Learning. “This type of learning leads to better student engagement because the content is relevant to each student and tailored to their unique needs……Depending on the strategy pursued, competency-based systems also create multiple pathways to graduation”[5]. Hence, the main goal is to curate this guided learning journey for users so they only concentrate on acquiring knowledge in the amount of time they have dedicated to learning something new or developing their skills.

Bridge the Skills Gap

???? In addition to the online learning-related challenges, there is a bigger problem that is not directly related to online learning but to education in general; the skills gap. Today’s job market requires a bunch of skills that education does not supply the learners with. Bridging this skills gap will make it easier for people to find the right job armed with enough knowledge and skills to perform its different duties and responsibilities efficiently. To overcome this problem alongside offering valuable learning materials, we create our content having in mind the skills set it targets based on the job market needs. In other words, we analyze the skills from the job posts and create our content in a way to fill the gap of the missing skills and hence, learners would be qualified with the right skills to apply confidently for the job vacancies.

Creating Content that Counts (CCC)

???? To be able to create content that appeals to the target audience, satisfies their needs and greatly impacts their career life, we created a framework or a cycle for our content creation.

The 3 main stages for creating content:

1.???? AI Automation

2.???? Instructional Development

3.???? Subject Matter Expert (SME) Review

Take a look at the figure below for a quick overview of what happens in each stage.

Stage 1: AI Automation

???? ?????As I mentioned, one of our primary focuses to bridge the skills gap is to provide learners with knowledge based on real job requirements to ensure once they finish the learning journey that they are truly qualified and equipped with the right skills that the job actually demands. Therefore, our software engineers have developed an artificial intelligence system consisting of a number of engines; each has its own role. The main outcome of the Automation phase is getting results for online content that is as accurate and as focused as possible to the topic that needs to be developed in a distinctly shorter period of time compared to doing the same process using human resources.

???? The content results are also selected not only based on its relevance to the topic but also based on the skills they cover. Below is a simplified block diagram of how engines function.

These 3 filtering phases include many more sub-phases underneath, each done with another specific engine that filters content based on certain pre-identified criteria. For instance, in the 2nd filtering phase, we try to obliterate any click-bait articles, spam or news from the results. The more filters, the better and more focused the results. After that, the content is used to create something called a “Knowledge Graph”. The idea behind that knowledge graph is to match the content of articles and job posts online. This is done by analyzing several employment websites and analyzing the jobs posted and extracting job titles from those posts. There is another engine that is dedicated to extracting all the skills out there in the real world. We currently have a bank of almost 50,000 different skills. To create the knowledge graph, the engines use machine learning and natural language processing techniques to compare the requirements of each job title with articles on the web that cover the relevant topics and match them with the skills analyzed.

Another engine used to enhance the quality of the results is an engine that analyzes the social shares. In other words, we get results using statistical methods indicating that a certain article about topic “X’ has been shared “X” times on social media platforms (Facebook, Medium, Reddit, Twitter,…etc.) and thus there is a good probability that this article is more relevant than another covering the same topic. Article popularity is one of the parameters that indicate the presence of valuable content, yet we do not exclusively depend on this because not everything viral means it has learning value. Other engines include: skills assessment engine that helps in creating questions based on the missing skills, and a complexity engine that identifies the complexity of the content. In fact, there are more engines than those mentioned because the more filters we have for the content and articles, the better and more accurate results we get.

???? The system also incorporates machine learning algorithms to enhance itself more over time. The more inputs we feed it, the better the recommendation quality it will produce, and the better skills matching will be done. The AI working on the recommendation is very close to that of Facebook which understands your circles of interest and by using some algorithms, it starts recommending friends or posts based on your preferences. Such a process of mining content is a fully automated process and saves a lot of time and effort.

?

Stage 2: Instructional Development

???? “Thank you technology, we’re taking it from here.” That is the slogan of the content engineering team who uses the results from the AI system to curate a learning path for the users. In our adventure trying to make the best learning experience possible, we have decided that all our content should be categorized based on its difficulty level. Take a look at our persona tree:

The content team scans the recommended results from the AI system meticulously and compares them to their findings to create a learning path based on the above tree. Topics and subtopics are designed and selected either from the Automation phase results or based on an intensive research process to identify which topics and subtopics should be covered in this persona that would empower the users with the needed skills. For each subtopic, three levels of difficulty are created; beginner, intermediate, and advanced. This should make room for everybody who seeks knowledge to benefit. If the user is still a beginner, he can start his journey from the beginning and would find appropriate content that matches his level and addresses the missing skills. On the other hand, if users have good knowledge about a topic, they would find a chance to build on their knowledge and improve themselves even more for better job opportunities or to grow their careers. As agreed before, the skills gap is one of our biggest concerns, and curating content to bridge this gap does not end with the recommendations from the AI engines, but the process continues with the content engineers. Each difficulty level or path is divided into several objectives as follows:

This sequence of content elements is designed and built based on how the learner should start his learning journey and how he should move on to the next step. Everything is guided by actionable learning objectives that allow the learners to visualize what they acquire when they finish reading the content related. The process of formulating the learning objectives is based on the learning outcomes desired and specified for each persona, and it takes into consideration the relation between the difficulty level and the type of objectives related. Bloom’s taxonomy is used to create the learning objectives in a way appropriate to the cognitive skills of users at a certain level of knowledge.

For the beginner paths, we mainly aim at providing content that matches knowledge & comprehension levels. For the intermediate paths, we take users to a higher level of thinking skills; application and analysis. Finally, for the advanced paths, we try to include content that helps learners build on synthesis and evaluation skills. Each of these levels of cognitive skills is bound by a set of guiding action verbs that give an indication of the type of content that should be present in the learning path. The result is content that is customized for each difficulty level and takes into account the writing style, language used, readability time, and complexity.

?

???? Since one of the major challenges mentioned before was time, and how online learning can be effective it succeeded in making learners utilize their time well enough to gain as much knowledge as possible, AI can help course designers create “Key Learning Cards”. These cards have two main functions:

·?????? Act as a recap on the most significant information mentioned in the content articles about a specific subtopic

·?????? A summary for those who have limited time but still feel they want to quickly learn something

???? Sometimes, what is blocking the way for you to learn online is that feeling that you will either waste a lot of time until you actually learn or that you do not really have enough time to dedicate regularly to that. Key learning cards sum up what learners need to know in a nutshell where information is chunked and written in an easy way that helps them to grasp content without having to read a lot.

???? The IDs use the ADDIE Model (Analysis, Design, Development, Implementation, and Evaluation) as the instructional process to follow while creating the content. Moving from one phase to another does not mean what is done cannot be undone; this is perhaps an old school dealing rigidly with the ADDIE model.? By experience and due to the dynamic nature of today’s projects, some flexibility in dealing with the ADDIE model is required. Unlike the rest of the phases, the evaluation phase consists of several integrated iterations and can trigger going back sometimes to the development phase and other times even to the design phase based on feedback on content or content elements updates or enhancements based on statistical data from users’ behaviors and interactions with the content.

Stage 3: Expert Review

???? How do we make sure the developed content is accurate from a technical perspective? This happens in the third stage of our content creation process which is done through the experts (SMEs). They review the developed content by the content team and provide feedback with suggestions to improve the content (if it needs so) by adding or removing or restructuring some of the elements in the learning path. This is third-eye checking (2nd check is done by the content team for the results that came from our AI system) guarantees accurate content from a technical perspective so that we are confident enough that someone with a related experience provided his/her own insights to the work done. It is worth mentioning that the expert committee network is expanding over time to include more experts from all around the world. The process involved with experts can be summarized in the following diagram:

Just Got “Engaged”!

???? Congratulations! Succeeding to engaging your learners is part and parcel of the success of the learning journey. Learners by nature get bored easily while learning, and online learning was expected to solve this problem with all the capabilities around and multimedia aids. According to a study by The University of Warwick on MOOCs, the completion rate for most courses is below 13% [7] which is a shocking number. Learners do not complete courses for several reasons and factors; on top of which is your inability to get learners engaged enough and also because the content of the course is not relevant to what they are exactly looking for.

This problem can be content-related and can be related to the overall experience. If your content is of bad quality, boring, shallow, or irrelevant to the users, they will definitely leave your course. Sometimes, your content is good but the experience, in general, has some faults like a bad platform, complex user interface, poor presentation of information, and several other reasons. Here comes the role of engaging learners through AI that will make it possible to have a completely personalized experience that varies according to the competencies and skills that users need to learn.

Assess With No Stress

???? Assessment is an integral part of the learning process. Learners need checkpoints during their learning journey to make sure they are following up and acquiring knowledge properly. It is also another chance for learners to interact with the content instead of just jumping from one learning element to another. The checkpoints are quiz questions created by the content team at the end of each objective. They are developed in a way to assist users to learn and recap the most important information presented without the stress of how many incorrect answers they have made. Having a safe learning environment makes learners feel more comfortable and willing to continue improving without being afraid of failing.

???? In addition, there is a final assessment to evaluate the user’s performance throughout the learning path before moving forward to the next difficulty level. Unlike the formative assessment represented in the quizzes, these summative final assessments provide a more challenging opportunity with a time limit and a passing score to motivate learners to improve their level of awareness and knowledge about the subtopic.

???? A competency-based learning experience that is supported by AI ensures there is an alignment between the learning objectives and the assessment questions provided so that the process of checking knowledge is always an additional stress-free learning experience.

Questions created are linked to the skill(s) they cover. They are also to the different learning elements throughout the course. This interlinked structure makes assessing learners based on the skills or competencies acquired more accurate and leads to better outcomes since it indicates which skills have been mastered by the learners and which skills need further improvement.

Conclusion

???? In a world that has been crunched with data and content, seekers of knowledge are in dire need of a dependable guide to help them discover the right skills they need to acquire using the right content elements and the right path, and recruiters are in a dire need of something to introduce them to applicants with specific skills set needed for certain vacancies. A mix of competency-based learning and personalized learning can be a lighthouse and a savior for those who lost their way and a solution for filling the significant skills gap.

By blending artificial intelligence concepts and applications with different learning methodologies and theories, the end goal of a personalized learning experience that is both enjoyable and valuable to users has become a reality. It is a journey that any person working in the training realm, as L&D ID, or e-learning feels committed to accomplishing until they achieve a dramatic impact on the learners and redefine the concept of online learning. AI models are yet to become more and more advanced in the years to come, and surely will positively affect the training experience.

References:

1) https://hostingfacts.com/internet-facts-stats/

2) https://www.internetworldstats.com/stats.htm

3) https://www.statista.com/topics/3115/e-learning-and-digital-education/

4) https://curatti.com/e-learning-trends-2018/

5) https://www.ed.gov/oii-news/competency-based-learning-or-personalized-learning

6) https://www.teachthought.com/learning/what-is-blooms-taxonomy-a-definition-for-teachers/

7) https://warwick.ac.uk/fac/sci/dcs/people/research/csrmaj/daniel_onah_edulearn14.pdf