Hi everyone,
Something just crossed my mind and I was wondering if it does make sense, which is that a higher retention will result in slower learning.
Meaning if I set my retention to 90% the intervals will be kept shorter compared to e.g. 70% to ensure the 90% retention, but this in turn would mean the Intervalls cannot grow that fast and therefore slow down learning.
Does that make sense ?
To give an example, let’s say I would be learning for a year with 70% vs 90% retention. Then I stop the learning/reviewing process for some time (e.g. a year). What would be the amount of material retained. The 70% retention resulted in higher Intervalls so the material would be stronger encoded in the long term memory compared to the 90% with shorter Intervalls being less stored in the long term memory as more often refreshed I would assume.
How much you know can be measured by adding up the probabilities of recall of all your cards that have been reviewed at least once. How much time you spend can be measured by adding up all time spent on reviews. If you then divide on by the other, you will get a measure of efficiency.
That’s what “Compute Minimum Recommended Retention” does. It tries to find a value of retention that maximizes the “stuff memorized divided by time spent” ratio. Simply put, it tries to find a value of desired retention that gives you the best bang for your buck. Remember the most while spending the least time.
Thanks for your response but I think you are talking about something different here. I am not thinking about optimizing the current amount of material I know per minute spent learning. I am rather thinking about the future amount of material I will know being affected by the retention setting.
If my understanding is correct different retention settings would have different impacts on long term retention.
So with a retention setting of 90% if I keep doing my cards everyday the true retention value will be somewhat close to this value, but maybe the retention in +100 days if i would stop doing my cards today would only be 20%, compared to 70% setting with e.g. 30% in +100 days.
Rentention is just an indicator for “how much of the material do I remember”.
Higher retention means you review more often but memory traces get stronger and you forget the information more slowly. Lower retention means less reviewing, but you might forget faster.
In theory you should be able to remember more cards for a longer period of time if your desired rentention is higher.
But: At some point more reviews = more time needlessly spend reviewing. This is where the thing that Expertium described comes into play.
So to answer your title: No, higher retention doesn’t mean slower learning. It means you review more and also remember more.
Info: You could actually use “overlearning”: this means you will review something so much, you will never in your life forget it anymore (because your forgetting curve exceeds your life span). A typical example of that is learning the alphabet – once learned as a young child, adults don’t have to regularily practise it anymore but still won’t forget the alphabet.
My Intuition about this would rather be the more the Brain has to struggle to recall something the stronger the memory traces get. And the good recall on a Short Term is only because od the frequent exposure. At least that’s what I got from a recent huberman episode, but maybe I got it wrong.
If you struggle to recall something then that basically tells your brain that there is important information that you need, but it cannot access the information. So yes, your brain will put efforts into learning that.
The cascade is something like this:
You remember things the best if…
… you need it for survival,
… it is emotionally charged,
… you repeat it very often.
With 1 being the fastest to learn and 3 the slowest. Obviously more than one of those three can be true at the same time.
The thing though is that the intervals grow over time if you use anki (and remember your cards correctly). Basically you will maintain an e.g. 0.9 retention, but you will only have to review the cards with days / months / years intervalls. That’s not really “short term” (For more info look up the forgetting curve).
So if you set a high desired retention, then you review more, and remember for longer too. Lower desired retention gives you longer intervalls but you forget more. Basically there is a sweet spot between reviewing and time investment, to get the best out of your time and learning sessions – which brings us back to what Expertium wrote above. That is the sweet spot for you, individually.
This is why FSRS (and a lot of other algorithms) make the distinction between probability of recall (R) and memory stability (S).
R=f(t, S) where t is interval length. S in FSRS is defined as “an interval that corresponds to 90% probability of recall”. S is like half-life in physics, but…uh, nine-tenths-life.
The higher S, the more time it takes for R to decay.
To give an example, let’s say I would be learning for a year with 70% vs 90% retention. Then I stop the learning/reviewing process for some time (e.g. a year). What would be the amount of material retained?
That’s an interesting question. I don’t know. I suppose @L.M.Sherlock can simulate it.