I simulated human learning in a Fortran program based on facts I'd picked up studying upper division psychology. The program modeled short-term and long-term memory. A letter would stay in short-term memory until it was displaced by others. A letter would not enter long-term memory until it had been successfully recalled from short-term memory several times. Letters could be lost from long-term memory but that was a much slower process.
I adjusted the teaching algorithm's character introduction rate and studied the shape of the expected error rate curves for my modeled student. I found that for the memory properties I'd chosen an average correct recall of 70% lead to the fastest acquisition by long term memory. Then, with real students, I simply adjusted constants in the program to get similar learning curves which I presume to be similarly optimized.
See More History.
|Last edited January 3, 2016
Return to WelcomeVisitors