How Online Coding Clubs Will Transform Education

The COVID-19 pandemic is forcing rapid change. Our schools are moving their classes online. And now, some of the coding clubs that were meeting face-to-face are moving to online formats. My local CoderDojo Twin Cities is also considering the move to online forms. Moving from face-to-face mentoring to an online format has challenges. The direct body language signals that mentors get from sitting in the same room with their students are muted during an online mentoring session. Parents have fears of inviting strangers to interact with their children. But I am confident with patience, persistence, and leadership we will adapt to the new online medium.

As these online coding clubs mature in their coordination and leadership, new patterns will emerge. This blog describes these patterns and groups them into six levels of capabilities and maturity for online coding clubs like CoderDojo.

It is vital for stakeholders to have a clear vision of these levels of maturity. It isn’t easy to get to the more advanced levels without progression through the early levels. These capability and maturity models are useful because they help stakeholders understand the future vision of what these coding clubs can build and the intermediate steps that organizations need to go through to get there. Each level allows organizers to prioritize their tasks and recruit volunteers with the appropriate skills for that level.

Here my prediction of levels of online coding club maturity. For each level, I have described the core activities and the skills needed to master this level before any organization can proceed to the next level. Think of them as small stepping stones to your final destination: a place where everyone is invited to learn what is important to them, at their own pace, and in a secure environment that they can afford.

Level 1: Initial Formative Online Coding Club

At this level, volunteers, mentors, parents, and students will slowly start to re-form their existing coding clubs to use the online mediums. Instead of face-to-face mentoring sessions, groups will experiment with doing mentoring via Zoom and other online tools. It will take time for parents to trust the new online coding clubs and get to know their online mentors. Only a small percentage of clubs will make it through level 1. There are challenges getting people to learn how to use online tools, and there will be a need to share content and experiences. Here are a few guidelines that are being created by the Raspberry Pi Foundation and CoderDojo:

Raspberry Pi Foundation (RPF) Online Club Guidance:

RPF Designing Online Sessions

RPF Virtual Club Guidance:

In our current CoderDojo coding clouds, students and mentors state their preferences: the first choice of a topic they want to learn (Scratch, Robots, Python), the second choice of a topic, and third choice of a topic.

At level 1, mentors will be matched to students using a mostly manual process such as using a spreadsheet to match students and mentors. For larger clubs, the time of day and the day of the week might also be aligned between students and mentors.

The primary skill for organizers is how to recruit both mentors and students and how to set up Zoom or similar online meetings with breakouts for the subject and age-specific groups. Organizers must also train mentors to build trust with parents and create guardrails for appropriate behavior. Getting laptops and wifi access to underserved communities is also critical for these early levels of online mentoring.

Level 2: Managed Coordinated

At this level, our leaders will have the confidence that going online is not only practical but for many students without transportation to events, the preferred medium for learning how to code. Instead of a focus on the logistics of setting up Zoom and training mentors about online safety for our students, the new center will be on getting the right training materials and tools set up for the students and mentors. Going online is a new skill for many of us since it involves setting up and managing online content and having teams of staff create, test, and integrate online resources for our students. At CoderDojo Twin Cities, we are starting these efforts by using GitHub Pages as a model for crowdsourcing content. I am now learning about how to use the mkdocs system to create and manage lesson plans for students. You can see a sample of my work at the CoderDojo Twin Cities Python pages here:

I would love your feedback on this site. As a challenge, can you add a new lesson plan to the site using a pull request? Many software developers can do this once they know the Markdown format of a lesson plan. For teachers, we will need some training to get them up to speed. More documentation on how to do this will be coming on the site.

During level 2, mentor-student matching might use simple matching algorithms. As these online clubs grow, the number of participants will also grow, making the matching process more complicated.

The essential skills for Level 2 volunteers are creating online content and crowdsourcing this content to have both the breadth and depth of concepts and content demanded by students. The more mentors we have creating content, the more students will be attracted to the sites through search engines. Volunteer recruiting skills are essential for Level 2 organizations to succeed.

Level 3: Defined Concept Driven Learning

The CoderDojo organization has pioneered the idea of a Concept Card, or in the terminology of CoderDojo, a “Sushi Card.” These are colorful plastic laminated cards the students can pick up when they walk into a CoderDojo session. Concept Cards have a picture and some challenges on the front of the card and answers on the back of the card. The key is to try to put only a single concept on a card. I don’t yet know precisely how concept cards will migrate to an online system. Still, I can tell you when our kids return to an online mentoring session; the mentor needs to quickly understand what concepts each student has already learned and what new concepts they are capable of learning. Reviewing a set of icons for each concept is a great way to figure out where to get started. Once we get a large number of online concept cards created in a consistent format, it is time to load them all into a searchable database. If we are at the level where some of our systems can recall what lessons a student has taken, we can take advantage of these systems.

The need for protecting online privacy and compliance with local privacy regulations such as GPDR in Europe will limit what we can do in the short term. These requirements go far beyond what we can do with a free account on Github. An ideal solution would be a highly scalable graph database with robust vertex-based access control. Each concept would be a vertex in a dependency graph, and each student would have a vertex with a knowledge space of their learning history and their objectives. Parents would control precisely who could see what for each student’s learning history. Both students and parents could rate their experiences with individual mentors, and mentors’ history could be monitored for continuous improvement. Tools could be created to help mentors adjust their pace, presentation style on focus on specific concepts that are difficult to teach.

Level 4: Predictable Recommendation Driven

This level requires that we not only have a stable graph database of our concept cards, our students, their parents, and their experiences, but we also track the mentor and student’s recommendations on curriculums, content, quizzes, and mentors. This is where we start to take advantage of graph algorithms and machine learning to predict what courses are appropriate for mentors and students. Mentors will have to indicate what concepts they are comfortable teaching, and students will indicate what concepts they want to learn. The knowledge graph will also help mentors and students find content related to these concepts. Mentors and students will also be able to request new content, and content creator volunteers will be encouraged to create content that is in the most demand as well as suggests content to match learning objectives.

At this level, the online recommendation system will not only be able to recommend concepts and content, but it will also be able to recommend matching students to mentors and take their time-of-day and day-of-week preferences into account.

Volunteers of Level 4 organizations will need to have an understanding of how to store knowledge for each student and how to protect this data from inappropriate use. Volunteers will need to understand graph databases with fine-grain security models that comply with local student privacy regulations.

Level 5: Optimized AI-Agent Driven

At this level, the recommendation engines don’t just rank mentors, concepts, and content. They start to start to take the form of online chatbots that can have discussions about your learning objectives. The chatbots can start your sessions and answer basic questions and then bring in a mentor when needed to answer harder questions. To be effective, AI agents will need to be trained on massive databases that include taxonomies and graphs of concepts and content, but they also will work better if they can learn from chats with prior students. The more common sense graph structures they learn from, the more precise their answers will be. Systems like the Mosaic Knowledge Graph of learning events will need to be loaded into our graphs and used to train the neural agent networks.

The volunteers needed to build Level 5 coding clubs will need to understand graph databases and machine learning.

Level 6: Fully Autonomous AI-Learning Agent Clubs

This final level is where we have an extensive and well-connected knowledge graph that is used to train an online learning mentor. It would have access to all concept graphs, content, and access to billions of chats between students and mentors. The agents will understand common sense and be responsive to the particular needs of students down to recommending specific slides in a presentation, particular fragments of code, and even suggest particular discussion topics with mentors or other students (with mutual consent).

The skills for Level 6 include volunteers that have in-depth knowledge representation, graph databases, AI, machine learning, data science, NLP, reasoning, and inference.

These levels of maturity models are not unique to education. For example, self-driving cars also have six levels of autonomy. Level 0 is no autonomous driving, and level 5 is steering wheel optional. Having consistent definitions of these levels helps regulators and manufacturers agree on what should be regulated. I believe that the same regulations may also apply to online coding clubs in the future.

One example of an organization that is trying to build to the higher levels is the futurist academy. They are encouraging their members to learn how to build knowledge graphs to store learning concepts and content. Their students are already building recommendation engines using these graph databases. The Futurist Academy of learning is different from many of the mentor-limited organizations like CoderDojo. Their idea is to have students teach each other about advanced technologies. They are not waiting for their teachers to learn about emerging technologies like knowledge graphs, AI, and NLP. They are each learning one topic and teaching each other.

The COVID pandemic is not the only reason for going online and building systems that support better online learning with fewer mentors. Today students from wealthy families can afford one-on-one online mentoring. By crowdsourcing learning materials and building smarter learning management systems, we can potentially democratize learning. Students from low-income families can also play a role in the data-driven economy.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Dan McCreary

Dan McCreary

Distinguished Engineer that loves knowledge graphs, AI, and Systems Thinking. Fan of STEM, microcontrollers, robotics, DonkeyCars, and the AI Racing League.