Author’s note: I recognize this blog post is one day late from our class schedule. Sometimes life gets in the way (I am – after all – just a human), and I felt it more important to deliver with quality rather than a rushed last-minute job. My apologies to my classmates who were expecting this up yesterday.
I have some experience already with AI-Generated Learning Pathways through a program called Degreed, which my organization promotes to our corporate learning audience as a “social learning tool”, but I thought this would be a good topic to investigate for this blog post as these sorts of tools are continuing to evolve. The concept of AI-enabled tools like this is that they analyze user data and use their algorithms to recommend potential courses or just-in-time content such as microlearning to users. In my organization we use this tool to help promote life-long learning, connected with content repositories such as LinkedIn Learning, Harvard Business Review articles, TedX (Ted Talks), and our own LMS to recommend learning content. The concept of Learning Pathways refers to how these algorithms will recommend potential courses or content which can help the learner achieve a desired level of proficiency or skill to support their learning goals. For the purpose of this discussion learning pathway can be defined as a collection of self-directed learning content which is recommended for a learner in order to support their life-long learning needs. I am examining this in the context of corporate learning.
The concept of using AI-Generated algorithms to help generate personalized learning paths is not a new concept, but there is potential for this to become a more powerful application of iterative AI technologies given new advancements and tools such as ChatGPT moving into common use. In one Mexican study from 2018 (Sanchez Nigenda et al.), researchers proposed a mathematical algorithm which could be used to generate these sorts of personalized learning pathways by comparing learner metrics to criteria attributed to the content itself, including whether or not the content was mandatory or optional for the learner, the duration of the learning activity, its “utility value” to the overall learning objectives, whether or not it was a prerequisite to other learning activities, and other valuation scores (Sanchez Nigenda et al., 2018, p. 829). This sort of approach, treating both learners and learning content as data to be categorized, treats learning as transactional instead of a human-centered joy of life. Even the authors of this study recognize that “from the professional educator perspective, not every pedagogical concept might be quantifiable.” (p. 836) There is also no way these algorithms in the AI-Generated learning paths can operate without the technology.
The tool I’m more familiar with, Degreed, is intended to operate by having the learner first identify which business skills they possess, which business skills they wish to develop, and assign a proficiency self-rating to each skill. Optionally they can ask their leader to perform a rating as well, and the system will then average the two scores together. One downfall of this approach compared to the one described above is that it asks for subjective data from humans, but doesn’t that also provide a humanistic strength by engaging the learning in what they want to learn, and what their leader can support them in learning? This approach is similar to another study, this time employing MOOCs (Massive Open Online Courses) which have an AI-driven “recommender” tool which suggests courses along a learning path based on a learner’s identified target job role (Son et al., 2021, p. 59093). I do not profess to be a data scientist nor am I an expert in designing algorithms, but it seems to me that keeping the learner (and in the case of my business’ model, the learner’s business leader) involved in the identification of job roles or skills to be developed, the potential creation of learning pathways enabled by AI tools could improve learner engagements and learning outcomes. However, by involving the human input of these tools we also run the risks of them not being used as intended (that is certainly the case in my organization where very few learners have filled out their skills profile to take advantage of the algorithm), so perhaps we won’t understand the full value of these tools until learners are more engaged in using them fully?
References
Sanchez Nigenda, R., Maya Padrón, C., Martínez-Salazar, I., & Torres-Guerrero, F. (2018). Design and evaluation of planning and mathematical models for generating learning paths. Computational Intelligence, 34(3), 821–838. https://doi.org/10.1111/coin.12134
Son, N. T., Jaafar, J., Aziz, I. A., & Anh, B. N. (2021). Meta-Heuristic Algorithms for Learning Path Recommender at MOOC. IEEE Access, 9, 59093–59107. https://ieeexplore.ieee.org/abstract/document/9399414
Andrea, thank you for sharing a thoughtful post, including the introduction to the Degreed platform. It is relevant to our current learning content, specifically referencing design thinking, and has given me a new perspective on our first group project around Leadership Up-skilling. Professionally, it has also sparked thought around our aspirations to offer curated learning content instead of constantly feeling the need to create it from scratch on our LMS, which is costly and time-consuming. It also opens possibilities to offer more customized content that appeals to a diverse community of learners. Using technology to create customized ‘learning pathways’ in our context is an exciting possibility that I will continue exploring. I have signed up for their blog, along with a short webinar to explore the difference between the LMS and LXP.
To address your comments about involving the learner in the design process, I agree that involving learners in the design process could (or should) increase engagement and motivation. Key adult learning principles include choice and a sense of self-direction (Thoms, 2001). I am curious about your company’s approach to helping employees see the platform’s benefits. What’s in it for them, and how can it help them to achieve their professional goals? In addition, if we consider Koffman’s concept of the adjacent possible (Dron & Anderson,2014), keeping humans involved in the process helps technologies to evolve into something perhaps more useful than it’s originally intended use!
I’m glad this post was useful to you Lara! I’m a firm believer of curating content whenever possible instead of feeling that everything needs to be created from scratch. There is still an important place for learning designers, educators, or instructional designers in selecting exsiting content which is useful, screening it, and organizing it into curricula which is thoughtful in it’s scaffolding/ordering as well as creating reflection questions and other activities which complement it.
As for how my organization has implemented Degreed and communicated it’s purpose to our learners, I don’t know if we would necessarily be the model to follow (although I didn’t call this out in our blog post). Although Degreed has the potential to support self-service career development learning and curated learning pathways, I’ve seen my organization has not harnessed its full potential. Instead most of our learning teams are using Degreed just to log learning activities required to support one specific leadership development program which only addresses abuot 5% of our workforce. I hope to help expand its influence to as many other learners I can, although the scope in which I can do this in my specific role is limited.
Thank you for sharing how you are using Degreed. I love the use of AI to support planned learning but wonder how reliable it actually is. If I as a learner lack self-awareness of my skills and developmental needs what I enter in the system will affect what the output is. I also wonder about how learner disabilities or other accessibility needs might be taken into consideration. Whose data is powering the AI in this situation and can it truly meet human needs? AI certainly adds complexity that we need to think deeply about to ensure effective and ethical use 🙂
Those are really great questions Leeann, and I’ve considered a few of them myself. For our organization using Degreed as a social learning tool, the learner data used for the algorithm is only the data from our own organization, but with employees in 72 different countries that’s still a significant amount of information! I do not profess to be a data scientist, but I suspect our limited use of the program makes the algorithm less effective at providing truly meaninful recommendations. For example, I am not in a leadership role currently but because most of the users are recording their leadership development training that is the majority of the recommendations I recieve for content.
One of the other points you bring up is self-awareness of the learner, and that is probably my greatest criticism for this sort of AI-powered tool which relies so much on the learner’s subjective evaluation of their own skill levels. There is the opportunity within the tool for the learner’s leader to enter their own skill evaluation and the system will evaluate the skill based on the average of those two scores, but I don’t see many individuals using that feature in my own context. One aspect of this which I do like is that the learner is able to choose their own skills where they demonstrate proficiency, as well as skills they want to develop. This opportunity for choice in the learning path could lead to greater learner engagement, since they have an active role in choosing where and how they want to develop their skills. There is also a social learning component which has the potential to build communities of practice, connecting learners with similar skills and varrying skill levels so the learning they’ve completed can be shared as curated or recommended learning content to those with other learning interests. I wish the platform had better opportunities for community connection other than direct messages to build that community of practice, and I believe a feature like discussion forums would be a great addition to this application.
I hope this helped to expand what I was thinking. Thank you again for your comments!