Recent advances in artificial intelligence (AI) are promising great things for learning. The potential here is impressive, but there also exist many questions and insecurities around deploying AI technology for learning: What can AI do? Where is it best utilized? What are the limits? And particularly: What does that leave for the instructional designer and other human roles in learning, such as coaching and training?
We want to suggest that these developments are for the benefit of everyone—from organizational development strategy devised in the C-suite, via content creation/curation by instructional designers, right through to the learners, as well as coaches and trainers who work with the learners.
Learning programs can now be built automatically from existing, high-quality content and a large range of resources, in seconds and at scale. This maximizes agility for creation/curation and drastically reduces the timescales involved. AI can also deliver these programs to learners, individually personalized at a granular level. It can do this across all types of content and assessments, including VR/AR environments, allowing for effective learning at an individual’s pace. Again, AI allows us to do this at scale, while measuring progress for individual learners and groups, as well as impact of each piece of content.
How does this work? Well, at the heart of these new capabilities is the advancement of AI technologies to the stage of semantic processing. That is, it can process information and sort it into modules and competencies, completely objectively, or if required around an existing framework. For instance, a system can process a corpus of text and do several important things:
- Lay out the knowledge/content structure, in breadth and depth
- Identify content gaps, assess content variability, and assess question coverage
- Suggest questions about the content
- Answer questions about the content
- Identify learning competencies from the content, or fit the learning into an existing framework of competencies
In that sense, AI has automated the knowledge side of the story. This is an impressive time-saver! With this in place, AI can now cover a topic, scalably, via self-guided learning. Through measurement of prior existing knowledge as well as continuous measurement of learning progress, programs can fully adapt to each individual learner. This is done by identifying what the learner knows, what they don’t know, and where the boundaries of competence lie, allowing learners to learn efficiently and effectively, in their own time.
This new level of data and learner analytics can be effectively utilized to make data-informed decisions on further learning, but also inform strategy and preparation for training, workshops, and coaching. For human-led sessions and efforts, the available data allows identifying learners and planning and grouping them accordingly, thus greatly increasing the potential impact during these sessions.
Looking at the positives, this is a step forward from digital learning as we know it. The old approach produces a one-size-fits-all course, via costly and arduous human input. Time and resources are precious, and human building and tagging is therefore limited in scope. Whereas, the volume of content that can be accommodated utilizing AI is practically limitless. It makes no difference to the AI if it is given what for most learners, say on average, should be a one-hour program or a 50-hour program.
This is why organizations, L&D teams, and learners are already benefiting from AI technology in digital learning, across sectors, and its positive impact through detailed learner analytics on digital as well as non-digital efforts.
This brings us to the second concern that is often raised, about instructional design and its role in the future. If technology can handle all of the above, then one could ask: Who needs people to take videos, slides, and PDFs to tag them, put them on screen, and add a quiz? However, we need to unpack the human role in making the above work.
First, note the requirement for ‘high quality’ content. While it can be processed by the AI, it can’t be written by the AI. The original content suite has to be curated or created. Further, if the AI indicates gaps in the content, then that requires content work as well. Thus, there is a role for expertise, external to the system.
If that were truly all there is to learning, there might not be a role for designers. However, there’s more to the story. Knowledge alone rarely leads to new abilities. Too frequently, we see evidence of an old view of learning, where providing information leads to new behaviors. This is evident in bullet point presentations by experts for novices: We’re rational beings, so if presented with new information, we’ll obviously change how we do things, no? Sounds simple and convincing, except our brains don’t work that way. Particularly for complex topics, ones that we haven’t evolved to learn.
Instead, we need deep practice. This means meaningful activity, aka learning experience design (LXD). Here, the task is to create compelling contexts, believable challenges, and provide specific feedback. The VR and AR experiences that can be incorporated, for instance, need to come from somewhere! The AI can’t create them. Which brings us back to that lack of deep knowledge on the part of AI. At the moment, technology simply cannot do this.
High quality experience design is and remains the gold standard. With AI automating the process of deploying high quality content and experiences via personalized, digital programs of learning, organizations can focus their resources to create and curate, in house, effectively and efficiently.
With this new view of complementary capabilities, we predict a paradigm shift towards in-house created and curated programs, bespoke to an organization's culture, vision, clients, and products. Specialists can focus their time on the creation and curation of experiences, wasting no time on tagging and other laborious and inefficient tasks. Through the ability to manage vast content libraries with the help of AI in seconds, organizations will see much more potential as well as return on investment (ROI) from their LXD teams, and LXD teams will grow in importance as a result. L&D will be able to shift resources from knowledge content generation to review and experience design. In addition, bandwidth can be shifted to add in facilitating the informal learning that drives continual innovation to complement the necessary focus on optimal execution.
However, not only the opportunities, but also the demands on L&D teams are set to grow. The future of work will require more agile upskilling and reskilling opportunities, and the so-called Great Resignation has shown us that learning is a growing focus in organizations. Employees as well as HR teams are turning to L&D, emphasizing the expectation to be valued through growth opportunities provided as part of their package, personally as well as professionally.
Lastly, we need to recognize the times when we shouldn’t try to put information into the head, and instead put it into the world: performance support via designed job aids. Note that AI can help here too, think chatbots for example, which organizations are using inward but also outward facing, more and more successfully. Also, for in-work-flow learning, we may not even require a chatbot to pretend to be a person to give us an answer. When the AI has semantically processed vast amounts of information and content, it can simply point the enquirer to the right piece of information through a simple search, and such search functions are fast increasing in functionality and efficiency.
Again, this use of AI should not be concerning because it moves human effort to creative and unique activities that shouldn’t be automated, and certainly can’t be in the foreseeable future. It moves human endeavor from rote information presentation to engaging activity design. It achieves that desirable goal of complementing what humans do well with what technology does well, to create a whole greater than the sum of its parts.
Together, the AI part as well as the meaningful practice part open up a (hitherto impossible) opportunity to make learning and implementation of learning for performance a continuous practice in organizational development, rather than one-off training courses and certificates. The AI is a partner here, rather than a threat. The goal is to have the AI do all the boring bits, basic knowledge and practice, and rather well through personalization, which allows humans to focus on the much more fun and vitally important part of the learning journey—moving from theory to practice and performance.
If you understand what each does well, you can craft a solution that achieves more than either alone. We welcome not our robot overlords, but our technology companions that augment us in powerful ways. Here’s to a continual learning future.