I had an idea about a dynamic retention rate.
Essentially, over the first week time interval for questions, it would involve starting from the most efficient retention rate (i.e. what is the least amount of time over the long-term to learn the most amount of cards) calculated by the program. Then, it would gradually curve to the set retention rate by the user in the current “Desired Retention” over the week. So, the retention rate will be on a curve for those questions.
Goal to increase number of new questions to be completed and thus memorized over time. Should be a similar idea to current workload days for the week, where the user can manually change say the weekend to have a smaller workload. However, in this case, the modification of retention rate would be done for only those cards that have the short retrievability interval of up to a week.
I’m curious if this idea has been considered and if it might lead to better results in the long-term?
My intuition says it would because of the effects of sleep on consolidating memory.
And so, the cards that are reviewed on a shorter interval (or perhaps with shorter retrievability or less stability, or some formula that takes multiple of these factors into account) would be reviewed less times over the first week (say the formula calculates the most efficient retention rate goal as 80% and so the program would start there for cards that passed one day and gradually over that week work up to the desired retention goal of the user (say 90%)). This is fine as they would be consolidated over that week’s time by other factors such as sleeping as well. Then, as they gradually go over that week, they would go back to the 90% user desired interval (or whichever number they use). The advantage is the user can now study more cards in that same one week interval (and so could complete more new cards for example), which would increase efficiency of memorizing over time.
Other notes:
I’ve picked the numbers for retention above just as example of course.
Also, I’ve picked the one week interval as this seems like an ideal interval since it matches the current “Easy Days” function, and so I think shouldn’t be disruptive to learning in the long term. For example, if it were a month, that would be too long for the retention rate to increase up to the desired retention rate. Perhaps a shorter interval of say 5 days or something might also be another option? I’m uncertain.
One other thing that might be desirable is to only apply this to “New” cards within first week of completing them. That way, it won’t affect long-term cards as well near an exam date for example. Another idea along the same lines is having a feature to input an exam date so that the cards aren’t scheduled to 80% and go past that date, when they should be at 90% and before that date. There are likely other solutions to that same issue, but those are just two quick ones I thought of.