Bespoke Learning, AI, & ‘On The Fly’ Design.
Personalized, Individualized, Bespoke, or good old fashion Custom all imply created by the hands of a human as opposed to factory machine manufactured. As an Instructional Designer, I am feeling like I am spending as much time sussing out how to handle the rise of generative AI as actually creating ‘good learning’. At one point I was actually told to ‘just use ChatGPT’ to create a course that would be sold publicly as a learning product. Right away I was smacked in the face by the ethical use of AI.
Think back – just as a tailor would custom fit men and women’s clothing as far back as when the term bespoke originated in the 1800’s, effective face to face teachers, trainers, and coaches have cultivated a knack for tailoring education to the audience or individual.
Why? How? Because, as the TED talk commandments state, they have learned by being considerate, curious and engaged. Since they know that their audience is the hero, they have payed attention to shifting their mode of communication and, hopefully, incorporating different approaches or exercises according to how well the learner (audience) is reacting. In face to face trainings, this can be a simple as reading the expressions on people’s faces and making a shift, “on the fly”, so to speak.
Bespoke ‘On the Fly’ Design requires creativity and openness to circumstance.
Over time, lecturers, trainers, and coaches should, through reflection, feedback, and research accumulate a huge pool of clever techniques to meet any one where they are.
Ask such an educator to try to encapsulate what they do into eLearning, especially an asynchronous, self-directed course, and the practitioner almost always views the task as trying to tame an unmanageable beast. There are simply too many variables.
When I designed instruction as a horse trainer before I became a trained Instructional Designer…
…I was constantly designing lesson plans ‘on the fly’ dependent on surroundings plus the mental state of both horse and rider. No matter what the higher level ‘curriculum’ was, it was ALWAYS critical to take in the current circumstance and adjust accordingly. For this reason I developed my toolkit of practical exercises that I could use accordingly.
To help keep myself and my students grounded, I created a four level assessment system (rewards included!) and practice handbook. This was especially helpful when students were itching to jump their first fence before they were ready. Thankfully, this intrinsic motivation kept them working hard at the foundational skills required to safely make it to the next level.
The material was derived from commonly accepted safety, best practice, and show competition standards from the long established Pony Club & 4-H Manuals along with books written by notable legends such as Chef d’équipe of the US Equestrian team and others. Most, if not all, were exercises I had incorporated into my lessons at one time or the other. After I finished, I had 10 pages of practices to refer to when I needed to refresh my lesson designs. The exercise of creating this, along with a multitude of coaching experiences, provided me with a mental catalog that I would refer to in the moment during a lesson.
That is my human version of a searchable database of information, a core foundation of necessary skills & competencies surrounded by a myriad of practical ways to acquire novice, intermediate, and advanced levels of mastery.
But, for all of my experience, as a self-directed learner and ID professional, I am feeling both inspired and challenged by using generative AI. The best way I could ‘get to know it better’ is to use it, especially for subject areas where I DO have the expertise to analyze the results. I more or less consider myself to have a novice/intermediate level of experience and have read plenty about the tech in the past several years but, lately, it almost seems that AI is evolving so fast that even those in the trenches are able to keep pace. SO what to do? Conduct a personal study as a self-assigned practice exercise to learn more.
The following is a case study that may help you define the line between the benefit consulting/collaborating with AI (which we know can mistakes) vs. out of the box lessons vs. human interaction (also having been know to be wrong a time or two 🧐).
Note: It may not be necessary to read the details of each response so I built them into accordions. When you see an orange header, click to read the full response.
Case Study: ChatGPT – Lesson/Exercise suggestions for a hunter/jumper equestrian coach: Conducted on May 16th, 2024 by Patricia Stitson, Instructional Designer (ID) and Subject Matter Expert (SME).
Case Study Step 1: Introductory knowledge
Approach: Use two prompts from different perspectives understanding that the answer to the second prompt would be influence by the first.
Prompt 1: Pretend you are a hunter/jumper equestrian coach. Your student is having trouble keeping their heels down when their horse jumps over a fence. What would you do? What type of exercises could you use with the student?
Prompt 2: What would you recommend to help a student create a strong, stable, & effective lower leg?
SME Critique:
I was impressed with the high level organization and pointers that ChatGPT provided. From both the perspective of a H/J Trainer SME and an ID SME the responses reflected the different approaches set up within the prompt. The first prompt was presented as the challenge of keeping the heels down and second is asking how to develop of lower leg stability. However, it is clear that the focal point of “heels down” was established by the first.
These responses caused me to reflect back on my years as in a horseback riding instructor and as a rider. Back then, I wish I have been more aware of how easy it is to get into the rut of drilling to ‘fix’ a problem ( I noted those instructions in orange), rather than taking the more formative route of guiding the rider through building (and maintaining) strength and stability which lead to having a more effective lower leg. So, in this way, using my ID competency of reflection, I learned from working with ChatGPT that there were times I may have gotten too fixated on drilling.
From a seasoned H/J coaching point of view, the suggested training exercises were not explained well enough. If I were using these ChatGPT responses to create instruction for myself or others as a non-expert, I would not be armed with knowledge of certain subtleties nor a broader perspective on body position in general
Through this part of the case study exercise, I learned that I felt confident using ChatGPT at the beginning of the workflow to learn about some foundational techniques but then was compelled to seek out examples of human-lead instruction. ChatGPT provided the bricks but I needed an artisan to show me how to build with them.
For example, this video shows a ‘master’ coach working with two riders performing an exercise (that ChatGPT did suggest) with the goal of “Creating a Strong, Stable, & Effective Lower Leg.” Watching this, even a lay person would be educated on the ‘why and how’ whereas ChatGPT only provided the ‘what’ and would require more prompts to get down to what is explained in this 6 minute video.
Within a higher level curriculum based on what ChatGPT delivered, I could then provide a micro lesson under a module called “Cavaletti Work for Stability” based on this video or something similar.
At a glance conclusion: While using ChatGPT could expedite my process working with a SME, ultimately it would take a three way collaboration (AI, ID, & SME) to ensure efficacy.
Case Study Step 2: Custom Tailored Instruction
As a horseback riding instructor I noticed that in the video demonstration the rider’s elbows were slightly stiff. I decided to ask ChatGPT to help me address this challenge. This would further test the bot’s ability to create more ‘human’ bespoke level instruction.
Prompt 1: My rider seems to get her heels down well and holds her upper body well over cavalettis but then her elbows get stiff. 👈 Since we were within the same string I started presenting the issue as a matter of fact.
Prompt 2: Do you have an exercise that would help her keep elastic elbows while keeping her heels down and core strong specifically during the cavelletti work? 👈 Since we were within the same string, I just asked for a specific exercise to use with ‘her’.
SME Critique:
ChatGPT is missing quite a bit within both responses especially from a learner centric design strategy perspective. The ‘what’ framework again is great but, without expert knowledge, it would be difficult to know how to teach any of the exercises correctly. For example, there are a million ways to incorrectly perform a half-halt which is one of the recommended exercises.
At this level, many more questions would need to be asked to get to the how and why knowledge provided by a seasoned professional in order to create truly personalized learning, thereby making it less and less efficient and effective than working with a human SME.
More to come as far as a deeper dive.
What do you think about using AI to create self-directed study either for yourself or others?