Why can’t learning be measured by algorithms?
Modern eLearning solutions use algorithms for a variety of purposes, including course recommendations, skill tags, completion scores, heat maps, and engagement level metrics. Anyone interested in eLearning is looking at learning in a new way. All of these methods are measurable, sortable, and optimizable. We seem to have come a long way in terms of learning. Through data-driven learning, you can increase efficiency, personalize learning, and scale up. A difficult question for L&D teams to consider is: Are we still designing learning for people, or are we designing learning for algorithms?
The learning design is optimized based on what the system rewards (i.e., system incentives), allowing for more short learning modules, more assessments (easier to measure, track, and report via the LMS), as well as smaller, bite-sized content (we call this microlearning).
Optimizing the learning experience provides great value to participants, as many learners are only interested in completing the learning experience and evaluating success, rather than building the capacity to succeed. The learning experience is never meant to be “smooth.” True learning comes from making mistakes, taking time to reflect on them, and learning from them, none of which can be measured by algorithms.
Will AI-driven personalized learning help or not?
There’s no doubt that many are using AI-based personalized learning to identify the best materials for each learner based on their previous learning experience, behavior, and role. When used effectively, it allows learners to access the right material at the right time.
In most cases, recommendations are based on a limited number of data points. What users click on, how long they view an item, or the words used to describe an item. Recommendation engines only capture what the user views. However, it does not capture what the user has learned or can apply.
Therefore, users will continuously receive recommendations that are easy and repeatable because they have advanced knowledge about those recommendations, and therefore it is not that difficult. Challenging yourself to step out of your comfort zone allows individuals to grow, and engagement drives more engagement by algorithms.
As a result, users enter a familiar learning cycle and continue to achieve success, but their actual behavior remains unchanged.
Engagement is different from learning
I can provide numerous examples to support the theory that learning has a much greater impact than engagement at the activity level. There are many examples of learners performing a task at a high level of activity, only to be unable to recall any knowledge from that performance by the next week.
For example, consider someone who has difficulty completing a simulation. This learner’s actual level of engagement may have been very low. However, because the simulations were difficult to complete, students were probably able to learn and recall the specific information that was taught very well.
If an algorithm is designed to provide maximum learning based on measurable activity, it will optimize for that activity, rather than for the vast amount of knowledge gains that a learner can achieve.
The irony is that the most effective learning methods are the least measurable compared to measurable learning methods such as reflection, peer-to-peer learning, and quiet inspiration.
Where humans still outperform machines
Algorithms can quickly find patterns in data and recognize many things at once without much effort, but applying human judgment, empathy, and understanding adds a level of value to the learning design process that cannot currently be achieved using algorithms alone.
The use of algorithms when designing learning experiences must be applied in the proper sequence and order to create an effective learning environment. Using algorithms to identify gaps in learner knowledge, adjust learning paths, and reduce administrative tasks all support the need for human judgment about what constitutes “effective learning.”
A definition of effective learning should include the following characteristics:
Create authentic experiences that express the richness of the world’s complexity. Ask a question that legitimately has more than one valid answer. Develop reflective, critical thinking and questioning skills among learners. Promote shared understanding among learners in an environment that helps foster social learning and shared meaning.
All of the above are necessary elements for effective learning, none of which are inefficient.
Designing e-learning systems using algorithms
The future direction of e-learning system design will not be achieved by abandoning algorithms or relying solely on them, but by developing systems that integrate both approaches. therefore:
We design experiences for humans first, then use algorithms to enhance them and provide support. Before developing or implementing new features or metrics for your e-learning system, you should consider three important questions: Does the feature/metric improve an individual’s ability to think or change behavior? Are we measuring how easy it is to accomplish something and what really matters? If the feature/metric is not available on the dashboard, does the feature provide value?
If the answer to all three of these questions is yes, then you have established a solid foundation for your e-learning system.
final thoughts
The bottom line is that education is more than just providing information. It helps form a learner’s identity. Learners can understand their role in society and how they contribute to the world.
Algorithms may suggest a path to follow, but where you end up is up to you, the individual learner. After all, learners don’t need an improved or more optimal learning experience. They need learning experiences that take into account how humans are designed to grow and develop.
