Outcomes Mapping

Tags: Content maps, adaptive learning, recommender systems

Created using Xoces.js.

What is outcomes mapping?

Learning outcomes specify what learners will know or be able to do as a result of a learning activity. Related terms you might hear are learning objectives, learning goals and intended learning outcomes. While there are subtle differences among these, they all fit our outcomes mapping framework discussed here.

An example of a learning outcome from an object-oriented programming course: Discuss the advantages of composition vs. inheritance.

In most current practices, learning outcomes are presented in a list — probably in the subject syllabus:

Learning outcomes in a list

Putting outcomes in a list form is easy, but it presents various problems for learners, course authors and anyone trying to do analysis:

  1. A list implies a step-by-step progression, starting from one to the next, which might not at all be the case.
  2. A list fails to capture the relationships among outcomes. What if Outcome 5 requires Outcome 3 but not Outcome 2?
  3. A list doesn't let you represent the relationships across courses. What if Outcome 2 on Charles' list requires Outcome 6 on René's list?
  4. You can't do sophisticated analysis on a list.

This is where mapping comes in.

When we map learning outcomes, we specify relationships between learning outcomes. For example, we can specify: Learning Outcome 2 requires Learning Outcome 1. This relationship connects the two learning outcomes:

Drawing a relationship between learning outcomes

Relationships can be specified across all learning outcomes. It might be that a content expert does this. Or we can use automated or semi-automated methods, using machine learning techniques, to help us infer relationships. We can also specify groupings of outcomes — for example, groupings along traditional course boundaries, or groupings of smaller sets of outcomes in modules, or however we find it valuable to structure our data set. Then our model includes relationships among outcomes within a module or a subject grouping, as well as relationships across different groupings.

Drawing relationships in this way gives us a network structure. The ontology of the network structure is shown here:

Learning outcomes ontology

So why would we want to have this network? It turns out that having this network structure is useful for many reasons. We can query over the network to answer questions that would otherwise be very hard in a traditional, table-based structure.

After mapping, the resulting mapped data set can be visualized in different ways. For example, it can be visualized as a chord diagram (as shown in the video above):

Learning outcomes chord diagram

Or as a network map (we used Rhumbl, a free network visualization tool, to create this):

Learning outcomes Rhumbl map

Clearly, the same data set can give rise to many different visualizations — this is one of the advantages of modeling and mapping.

Applications for Adaptive Learning

One of the most exciting technologies on the rise is that of adaptive learning systems. We won't go into adaptive systems in much detail, but at the core of every adaptive system lies the mapped data set of entities and relationships across entities. Mapping granular learning outcomes is an essential component of building this data set.