GPT for Education is Plain Awesome
For every class that I teach I follow two objectives
- Maximize student learning adhering to time capacity
- Minimize time spent per assessment
These two objectives are not mutually exclusive and, on the contrary, they are strongly positively correlated.
Learning
My effect in the first objective is limited due to the human element. I collect and curate a set of learning materials. I lay out a syllabus and perform it periodically. I give constructive feedback with occasional motivation if necessary. Though,
- Student might not be motivated enough to take my course especially if not elective
- Student might not have adequate available time to study (e.g, work-life balance)
- Student’s learning style and my teaching style might not agree at all (therefore I provide all sorts of extra material)
- Student might struggle with the learning curve
- Student might think I suck at my job (if they do, that’s probably true)
Assessment
Assessment sucks. As an instructor, you need to check student work periodically. It might be vital to students but it is a repetitive and dull work. Because you simply repeat the same thing for 20-50–100 students. Some lucky professors have TAs to do the grading.
Also, in many cases, it can be automated easily. However, there are two simple edge cases that make assessment really easy.
- The student does not turn in any work. Easy 0, no time.
- The student does excellent work. A very small amount of time, easy 100.
Anything in between is just extra work for the instructor with minimal benefit on the student. I am also a firm believer of the separation of teaching and assessment.
Therefore, student shall be given opportunity to pre-assess and self-assess their own work. It is also beneficial to the student’s own learning process.
I am lucky enough to teach coding related work. Therefore using some modern tools (e.g. Quarto, GitHub) students can self assess, learn from peers and yield quality output.
Enter the GPT/LLM
LLMs became the recent hype touching almost every aspect of life. With GPT-tech (e.g. ChatGPT, Github Copilot) it is much easier to write code, ask for explanations, writing essays, paraphrasing and many other tasks.
You can ask GPT to explain a concept in a simple way, give examples, applications all sorts of stuff. You can even ask GPT to write the code for you. Naturally it will struggle to churn the output you desire but even the process is educational.
Everybody now has an assistant which they might consult at will. This is simply huge. I flattens the learning curve and reduces headwinds and speed bumps in your learning process. Plus, no more sloppy copy-pastes from Wikipedia!
I am not (that) old but before the development of internet, you were confined to a limited pool of digital resources and the physical library if you are lucky. Then, with the internet and search engines our reach expanded to a practically infinite pool of learning resources. Now, with LLMs, it is much easier to navigate the waters and fish out the information that we need.
I use LLMs for my own learning of new concepts and topics. This year, I plan to encourage students more to adapt these tools to their workflows in my course. Because learning is endless, and the quicker they grasp the basics the faster they may employ advanced skills at their work and projects.
