The idea is to collaboratively build a graph-based map of the knowledge space, designed to be navigated seamlessly, in a fully personalized fashion, with a bias towards curiosity-driven exploration and the flexibility to accommodate any topic. This additionally enables the platform to serve as a personal learning management system and to become a social network.  

A knowledge map 

We want to envision knowledge spatially. When you think about it, isn't learning like setting out on a journey to unknown places? We ask ourselves:  

  • "Where am I? Where do I come from?" (What do I know?
  • "Where to go?" (What do I want to learn? What are my objectives?
  • "How to get there?" (Where do I find the best resources? What approach should I take?
  • "What's the best way to get there for me?" 
    Take the shortest path if you're in a rush, the scenic route if you want to check out all the most famous approaches or theories, or a gentle slope if you don't want to sprain your brain. 
    Conceiving of knowledge as space also allows for less linearity and more flexibility. You should be free to explore, wander around, get lost, travel back, change paths, and go off the beaten track. In other words, let curiosity guide you. 
    Self-growth indeed often comes from information we are not explicitly looking for. This is why we want to endorse a bias towards exploration and organic growth rather than goal-driven navigation. Acquiring knowledge and skills thus becomes a consequence of the learning process rather than an objective in itself. However, for this to work, the learning experience has to be seamless and rewarding. 

A tailored learning experience 

Every learner has a unique set of experiences, abilities, interests, and goals. Providing a seamless learning experience requires a personalized navigation system. 

Based on what you know 

We understand new concepts by building on top of concepts we have already assimilated. From a precise idea of what you know, it is possible to deduce what you don’t know and therefore to provide the next steps to build a clear and optimized path to success.  
Suppose you're an experienced software engineer and you want to learn a new programming language. You don't want to go through the same process as someone who knows very little about coding. Ideally, you want to know how the new language compares to languages you already know so that you just need to learn what is different.  
Integrating prior knowledge is also mandatory for quality continuous learning. Within a certain domain, recommendations should always be relevant, whether the user is a complete beginner, an expert who wants to stay up-to-date, or any level in-between. In other words, recommending content in your zone of proximal development. 

Based on how you learn best 

We all have strengths and weaknesses, different abilities that need to be taken into account in order to provide a rewarding, seamless experience, and avoid frustrations. Instead of following existing models to categorize learners, we aim to let data speak for itself. Machine learning in particular can be leveraged to cluster learning profiles in an unsupervised fashion.  
In practice, this amounts to serving someone with the content that fits them best: preferred media type or pedagogical approach for instance, or even content that includes examples and use cases that they can easily relate to. 
Learning can be extremely enjoyable, but too often we get stuck and become frustrated, or, get bored. To prevent this, we want to implement solutions to switch between alternative pedagogical approaches with one click, so that you can find a different way to tackle the same concept immediately. Perhaps a more visual or practical approach for instance. 
Finally, when sharing knowledge, a teacher is expressing something that they have internalized, with the intention of having someone else internalize it as well. This requires a lot of empathy, but ideally also commonality in the way of thinking. Teachers are inherently subjective, and we aim to embrace this subjectivity by connecting them with the right audience. That is to say, connecting people who understand each other.  

Based on your motivations and needs 

Understanding something effortlessly is pointless if you drop out before reaching your goal. When learning with a specific objective in mind, you want to: 

  • learn precisely what is needed, down to the level of understanding required for each concept. You want to avoid repeating things and assimilating unnecessary information. Our aim is to ensure that you are in control of your time and of what you want to learn. 
  • find the right balance between efficiency and enjoyability. Learning a new language is a classic example in that regard. 
    Your curiosity should however never be held back. Going for your latest obsession or changing course on a whim are great ways to plan a route of which you won't get tired. To this end, we want it to be easy and natural to cross bridges between domains.  
    If you are passionate about microeconomics, maybe you get interested in game theory and from there, move on to behavioral biology. Or you end up wandering through some of the implications of microeconomics on elections before setting off to investigate the workings of satire during the Trump presidency.  
    Self-directed learning, which has the learner in charge of defining the learning process, does not have to be a lonely undertaking. Making the right choices for yourself often implies getting help from external sources, in particular to become more aware of what the all the possible options are. Mapedia also aims to empower self-directed learners this way.  

Modeling the knowledge space 

The data structure we went for to model knowledge is a graph (for more details, checkout Approach). It allows for enough flexibility to support the kind of tailored experience that Mapedia will offer, in addition to being particularly adapted to model space. 
The data model design must be done iteratively and collaboratively, but we envision it to follow these basic principles: 

  • Generic modeling of knowledge - various domains (programming, biology, sociology, etc.) should be modeled in the same way. As discussed later, this provides many advantages. 
  • Precise yet abstract modeling - it should go all the way down to the level of individual concepts and the relationships between them, such as dependencies or similarities. 
    We understand that some tradeoffs have to be made in order to avoid the full complexity of the knowledge space while providing optimal value.  
    This modeling is our knowledge map. It gives you visibility over your itinerary, shows you what you know, what you don't know, and where to go next. From this model, we can deconstruct knowledge to reconstruct it in a personalized way for the learner. 

Collaborative, social, community-based 

Building this type of modeling represents a lot of work. The process could be partially automated through machine learning but also needs to represent the current social status quo. Concepts and their relationships or degrees of knowledge are ideas defined and shared on a societal level. The modeling must therefore reflect the current consensus (or the lack of one).  
What skills are you expected to have as a junior software engineer? This varies widely depending on the location or the company. Reflecting those divergences has to be done collaboratively.  
A precise modeling of knowledge allows Mapedia to support small and specific pieces of content. For example, an explanation for one individual concept, an exercise to assess a learner's understanding of one limited topic, or a task to practice one particular skill. This content can then be dynamically arranged to accurately answer a user's needs. Doing so vastly lowers the entry barrier for user-generated content and bypasses the inflexible structure of traditional courses, although it does create challenges in composing coherent learning paths. 
Working collaboratively is how we think these challenges should be addressed. This implies facilitating and encouraging feedback and quality assessment, using tools such as a review system, analytics about learners' content consumption (time spent, dropout rate or % of correct answers to tests for instance), and enabling interaction with and among learners (questions, live chats, forums).  
We embrace no preconceived ideas about what good learning material is and aim to push for creativity in designing new pedagogical approaches. The objective is to improve the content collaboratively based on feedback, community response, and data.  
There should also be room to express appreciation for someone's hard work. Even if we will only support free content on the platform, you should be able to thank, support content creators, and give them the recognition they deserve. You will eventually be able to create a public profile, where contributions and levels of expertise would be displayed. Ideally, this could even become an alternative to paid certifications, leveraging the fact that in order to teach something effectively it must first be fully mastered.  
Finally, building relationships among learners is also a great way to make the learning experience more engaging and ultimately more successful. We aim to develop self-managed communities around domains and support learning with peers. As an example, this could mean that you would be able to easily find people with a similar background to yours, learning the same things, struggling with the same concept, or interested in working on the same educational projects.  

A generic platform 

The platform aims to support almost any kind of domain that can be taught online and any kind of content that teaches it
Most people do not stay in one siloed field forever. We all have various interests that evolve over time. By targeting a wide range of domains and use cases through a generic modeling of knowledge, you can come back to Mapedia even as your priorities shift. In other words, it can become a free, personal learning management system that has no strings attached to the institution you are attending at one point in your life (school, university, work). Long-term use would also provide an ever-improving experience as ubiquitous data is continuously acquired about the user's learning profile, across various domains.  
Additionally, a generic product can take advantage of highly reusable features to build a flexible architecture. In turn, this facilitates the integration of community-developed modules supporting different types of content creation.  
Such a high level of abstraction is required in order to satisfy enough use cases and create a go-to solution for learning. Ideally, Mapedia would become the "entry point" for the user to learn anything. 
If you're curious to see how we plan to make it happen, follow us to Approach.

© 2023