#300_What I Learned After Trying to Create a Moral Education Lesson Gem 100 Times
For about two weeks, I have been going through trial and error to see if I could create a moral education lesson using a Gem.
The Gem to support independent science research took shape quickly.
The Gem to 'create an animal encyclopedia' based on learning content also worked right away.
However, a Gem to support moral education material research is proving difficult to build.
I wonder why.
While making adjustments for many hours, I realized something.
The moment a moral education lesson is bound by the Course of Study commentary, it loses its appeal.
Of course, this does not mean it should not be used as a reference.
Rather, you cannot create a moral education lesson without basing it on the Course of Study commentary.
However, basing it on something and being bound by it are different.
This was a major discovery for me.
When I try to feed the Course of Study into the knowledge base and have the AI create a lesson based on it, the AI tries to output a 'correct lesson.'
However, the words it produces are just a lesson that carefully traces the words written in the Course of Study commentary.
It is not wrong.
But the human wavering, contradictions, and essential content that cut to the core of the teaching material fade away.
In the first place, there are many steps to follow in moral education material research.
Set the objective.
Understand the reality of the children.
Grasp the plot of the teaching material.
Read the emotions and changes of the characters.
Find the moral questions contained in the material.
Think about the relationship with the target content items.
Connect it to the reality of the children.
And then, think about the questions to ask.
If I try to incorporate these into a prompt, it becomes a seven or eight-step process just for that.
While creating the Gem, I felt I understood a little why teachers say, 'Creating moral education lessons is difficult' or 'I can't think of any questions.'
So, I gave up on a Gem that creates the entire lesson plan and decided to break it down into smaller steps.
What I created this time is a Gem that supports moral education material research and question creation.
This is labeled as vol. 10, but I have made over a dozen revisions within each version.
The number of trials has easily exceeded 100.
The most difficult part was setting the criteria for judging when to move to the next step.
There is no such thing as 'finished' in material research.
Is the reading of the material sufficient?
Are the problems to be considered in class visible?
Will this question really move the children's thinking?
While checking these, I decide where to draw the line.
In the end, it is judged by the teacher's own conviction.
If I set the prompt too finely, the AI cannot reflect all the instructions and repeats the same questions.
If I set it too loosely, it moves to the next stage without deepening the thinking enough.
Too detailed, and it doesn't work.
Too loose, and thinking doesn't deepen.
I kept searching for the perfect balance of support that lies in between.
Through creating the Gem, I realized how amazing the human brain is.
At the same time, I believe that if even a part of this judgment could be reflected in AI, it would be a study that could support many teachers who struggle with creating moral education lessons.
It is still in the trial stage.
This time, I ran out of energy just creating the material research Gem.
However, come to think of it, I also feel that this material research might be the most important part of lesson planning.
Not an AI for shaping lesson plans, but an AI for reading teaching materials deeply and cultivating the teacher's own questions.
While aiming for that kind of support, I would like to continue my research that contributes to teachers' learning a little longer.