Skip to content Skip to main navigation Skip to footer

Suggested approaches to Locate the Learner

Approach with Navigator

Competency Framework: In this work, we build a two-dimensional competency map that is based on generating the semantic embeddings of learning resources. Here, learning is situated in a logical space called the progression space, where each point in this space represents a competency, or a unit of learning. Competency map is a two dimensional map with horizontal axis (X-axis) representing the topics and vertical axis (Y-axis) representing the complexity or pedagogic level of the learning resources, such that when we move from left to right on X-axis or top to bottom on Y-axis, there is a progression in terms of learning. Competency map also has different learning resources mapped on to it using the semantic embeddings.

Data Analysis and Logging: Events related to all learning activities of a user (instructors, learners, content authors) are logged and run through a data processing pipeline with processors to extract and compute different measures related to competency gains, time spent, learning challenges, usage and other metrics. These metrics are further used in the computation of learner and activity vectors and in the route and reroute algorithms. The learner analytics platform also processes the data to generate reports for different stakeholders.


  • Perseverance: A measure of the total amount of time a learner spends on competencies irresp
  • Grit: A measure of the total amount of time a learner spends on competencies in spite of failure
  • Self-confidence: A measure of the learner’s assessment of his own abilities.


  • Community: Indicates the measure of the user’s contribution to the community and is computed based on the user’s interactions with the community, sharing and contributions in terms of content, responses to questions asked and other community engagement
  • Authority: Indicates a measure of the user’s credibility in a competency, topic, domain or facet and is computed based on not only proficiency, but also on community interactions of answering questions related to the competency, topic, domain or facet, remixing of user’s content and other community related activities
  • Reputation: Indicates a user’s standing in the community and is computed for a competency, topic, domain or facet based on community’s rating and sharing of user’s content and answers etc.

Evidence based proficiency(competency) modeling:

We propose that data generated during the learning process, referred to as evidence, can be used to reason about the underlying competency. We focus on activities that learners do like consuming resources like videos, text books, articles, and documents. `

The work also shows that one single average model to predict the state of competency cannot be used and we need to build separate models for different earners, which validates “The Myth of Average”

AI Techniques and Tools

Competency Framework, Data Analysis & Logging

  • Word embeddings – Word2Vec
  • Document embeddings – Paragraph2Vec
  • FastText (Word Embeddings)
  • Glove (Word Embeddings)
  • LDA (Latent Dirichlet Allocation)
  • K-means clustering

Perseverance, Grit, Self-confidence

  • Probability distributions and Statistical techniques such as normalization and standardization to represent metrics within a desirable range (0 to 1)

Evidence based proficiency modeling

  • Evidence based proficiency modeling
  • SVM
  • K nearest neighbor clustering

Citizenship metrics

  • HITS dual relation algorithm is used to compute Citizenship and Authority, i.e. Good citizens endorse good authorities and good authorities are endorsed by good citizens.
  • Bayesian Knowledge Tracing
  • Statistical Techniques such as decay functions

Was This Article Helpful?


There are no comments yet

Leave a comment

Your email address will not be published. Required fields are marked *