One of my classes, Trends & Issues in eLearning, recently discussed adaptive learning.
I was a bit worried that our assigned viewing/reading might not give a fully accurate picture of adaptive learning, given that the authors seemed to be in a position where they would benefit from putting rose-colored glasses on their audience. (For example, the author of “What Is Adaptive Learning Anyway?” works for – and is posting on the website of – a company positioned to make money off the uncritical embrace of adaptive learning technology. If there are shortcomings, they aren’t likely to be talked about here.)
To balance this out, I ended up reading an article cited by another student (Johanes, P., & Lagerstrom, L., 2017) as well as this EdSurge article, which helped me better understand both pros and cons.
Reflecting on what I’ve read, here are some potential benefits of adaptive learning:
- It is learner-centric, with the express goal of helping the learner learn.
- It provides input and feedback that might be difficult to obtain in other formats, such as real-time, detailed data for teachers and course designers.
- It allows for real-time improvement and adjustment, which can parlay the aforementioned feedback into things like effective course design evaluation.
- For certain audiences, such as adult learners, this medium has the potential to create tailored courses to a wide range of learners, addressing specific needs in a way that make required training feel more useful and less tedious.
As with any learning approach or tool, there are also some drawbacks to adaptive learning which are worth unpacking a little:
A focus on the individual learner to the detriment of the group; a focus on explicit learning to the detriment of implicit learning.
Learners are siloed off to their own, independent experience, which can create some negative impacts.
I think this is more of an issue for K-12 and university learners. There is something to be said for the social component of learning, that feeling of “engagement” that comes from human interaction, and the contextual comprehension that comes from interacting with fellow students who have differing levels of expertise than you do.
This may not be as much of an issue for adult learners, since so much of adult learning is on-the-job, just-in-time, or some other knowledge delivery system that typically does not have a huge impact on socialization and sense of community.
The typical issues that crop up anytime algorithms are involved: data privacy, algorithmic bias, pattern bias, and black box obfuscation of code.
(Note: for an in-depth discussion of this I highly recommend C. O’Neil’s Weapons of Math Destruction, which is a fascinating read about algorithms in society at large.)
To break this down a little more:
- Transparency/availability/security of data = Who owns it? What happens if it is breached? Can it be sold? How might this impact students?
- Pattern bias = We see the patterns that we want/expect to see.
- Algorithmic bias = This is when computer systems create unfair outcomes that privilege certain groups of users over others, which can result in “impacts ranging from inadvertent privacy violations to reinforcing social biases of race, gender, sexuality, and ethnicity.” [Wikipedia]. O’Neil examines this at length in her book and sums it up nicely in a single core question: “whether we’ve eliminated human bias or simply camouflaged it with technology.” In adaptive learning, this could manifest in coding or design choices that move students into different tracks depending on traits not related to their actual competency (e.g. cultural context, implicit bias, internalized oppression. While this is an issue across education – see this article for an example – algorithms can obfuscate the issue, what this article refers to as “garbage in, gospel out”.
Student classifications and trajectories have the potential to not only silo students but create negative loops.
Johanes & Lagerstrom describe this possible issue: “Let us assume that an algorithm classifies students based on their performance as low, medium, or high achievers in a particular subject. What happens based on that classification? Are students locked into certain learning trajectories based on it? If so, then who deals with the edge or outlier cases?”
In other words, if a student gets categorized into the lowest group, does the adaptive learning program create opportunities for them to move out of that group? Is the student aware that they are in the lowest group, and if so does that negatively impact their engagement and hinder their progress? If they have a learning disability that might contribute to this low classification, who might notice and provide appropriate assistance – the algorithm? An instructor?
Again, these are issues that exist across education, but adding an algorithm to the mix creates an additional level of complexity that instructors must be aware of if they choose to incorporate this technology.
To be clear, the reason I’ve spent more time talking about downsides than upsides isn’t because I think adaptive learning is a bad idea – I actually quite like it! I simply wanted to fill in some of the opposing argument to this week’s generally laudatory readings/viewings.
The better you know both the strengths and weaknesses of a technology, the more effectively you can implement it.
Overall, I think adaptive learning has enormous potential, particularly for corporate learners. For younger learners (K-12 and university-level), adaptive learning may be better suited to being one tool in the toolbox instead of the entire system – its instructional advantages correspond to disadvantages in student socialization and sense of community, which are especially important for engagement, implicit knowledge, and well-rounded personal growth.
Johanes, P., & Lagerstrom, L. (2017). Adaptive learning: The premise, promise, and pitfalls. In Proceedings of the 124th ASEE Annual Conference and Exposition. Presented at the American Society for Engineering Education (ASEE) Annual Conference and Exposition, Columbus, OH.