(right now it's just sm2)
| Name | Inputs | Outputs |
|---|---|---|
| SM2 | quality, repetitions, previous ease factor, previous interval |
interval, repetitions, ease factor |
- Random chance re-encounter
- Leitner box
- Luhmann's Zettelkasten https://zettelkasten.de/posts/zettelkasten-improves-thinking-writing/
- surprise and serendipity
- relatedness
- capturing things that you didn't realize or expect (should be a real extension)
- loose filing, interconnectedness - you have to forget about them so you can find them again.
- Admonishes against sorting and categorizing, favoring natural clusters around frequently noted topics
- 'for every note, an id'
- direct connections
- managing a graph-of-linked-notes
- fixed interval
- Pimsleur
- naive exponential
- mnemosyne - (basically sm2) https://github.com/mnemosyne-proj/mnemosyne/blob/master/mnemosyne/libmnemosyne/schedulers/SM2_mnemosyne.py
- neural network
- fibonacci, hilariously
- https://www.quora.com/Whats-the-best-spaced-repetition-schedule
- answerer actually does lay out a lot of the other possible sequences
- ebbinghaus forgetting curve
- R = e ^ (-t/s), where R is recall, s is 'strength' of memory, t is time, e is euler's number.
- s gets higher after subsequent review
- original formula, for 'savings' as a percentage (with t in minutes): 100 * (1.84 / (log10(t)^1.25 + 1.84). Savings was defined as 'time saved from original learning time'
- should show a lower bound for information for facts we are trying to remember (meaningless, disconnected)
- nice formula elaboration https://psychology.stackexchange.com/a/5201
- another stackexchange post suggests an alternate formula: R = a + (1-a) * b * (1 + t)^-beta
- More mathematical models of memory: https://psychology.stackexchange.com/questions/5772/what-are-the-mathematical-models-of-memory?rq=1
- sm-15 https://github.com/slaypni/SM-15
- halflife-regression from duolingo: https://github.com/duolingo/halflife-regression
- 'memorize', an optimized algorithm for spaced repetition: http://learning.mpi-sws.org/memorize/ and https://github.com/Networks-Learning/memorize
- self-authored cards vs. other-authored cards
- learning to author cards well
- what kinds of inputs and data are easy / useful / possible
- kinds of cards, kinds of capture, kinds of interfaces
- curation vs. authoring
- SRS for many users
- overall system variables: fact introduction rate, card revision, time spent per day, overall error rate
- card feedback on prompt quality
- systems that promote usage frequency
- systems that are resilient to schedule-fall-apart
- context factors that affect retention in the short and long term (memory cues, prompts, and the environment)
- prior knowledge
- degree of importance
- individual capability and context
- general awareness / linkedness
- review rate
- frequency of retrieval
- meaningfulness, stress, sleep
- general mnemonic representation skills and techniques
- Interference from other knowledge / distractors / similar (but wrong) concepts
- Coherence with 'correct' prior knowledge
- Cognates and false cognates
- multiple presentations in multiple forms (castles in the clounds / bootstrapping / skyhooks)
- fake social media feed as a srs presentation vehicle
- encoding errors are more likely when related facts are forgotten or misremembered
- SRS is important because it helps the average recall strength for a particular fact
- massed might be important if the subsequent 'fact network' is being built immediately
- There are strong graph effects of related cards
- Some facts don't seem to suffer from forgetting. Why? Is it per person/fact pair, or are there generalities here?
- Remembering and getting it right feels good! What level of 'right' and 'wrong' should we aim for for getting the habits firmly in place?
- For 'natural' SRS systems (note reviewing, social media, environmental cues), how do we recognize that the world will provide us with chances to remember
- how can we capture those in a computer-data driven srs?
- how can we set those kinds of systems up for optimality?
- how does SRS relate to the Getting Shit Done / inbox zero / trusting your own capture mechanisms / limiting work in progress ideas?
- what are the neural substrates of remembering and forgetting? https://psychology.stackexchange.com/questions/9595/what-are-the-neural-substrates-of-retrieval-induced-forgetting
- are there real RIFo effects?
- Nice discussion of some of the factors in memory models: https://www.reddit.com/r/Anki/comments/alhjgv/anyone_know_of_any_nonsm2_nonleitnerbased_srs/
- Mess: https://www.ribbonfarm.com/2017/01/05/tendrils-of-mess-in-our-brains/