Piotr Wozniak, the creator of spaced repetition software (aka, the smart way to remember stuff) SuperMemo (henceforth SM) claims (and I believe him) that he was the one that really got spaced repetition going (for example, the famous graph of the serrated decay curve is allegedly actually due to him, not Ebbinghaus (who didn’t look at spaced reviews).
Though the effect was seen in earlier research, he popularized it and also added his nifty algorithm + software. Here’s a quote:
In 1984, my reasoning about memory was based on two simple intuitions that probably all students have:
if we review something twice, we remember it better. That’s pretty obvious, isn’t it? If we review it 3 times, we probably remember it even better
if we remember a set of notes, they will gradually disappear from memory, i.e. not all at once. This is easy to observe in life. Memories have different lifetimes
These two intuitions should make everyone wonder: how fast and how many notes we lose and when we should review next?
To this day, I am amazed that very few people ever bothered to measure that “optimum interval”. When I measured it myself, I was sure I would find more accurate results in books on psychology. I did not. See: Why spaced repetition research kept failing?
Impact:
It looks like basically all spaced repetition software traces back to SM, for example Anki (surveys say 30-80% of med students use it, and the anki mobile for ios is as of writing the 4th top paid app on the US Apple app store) started with SM’s algorithm, and it’s possible that Duolingo and Quizlet’s use was caused by the greater awareness SM caused or by trying to beat competitors (but this is just speculation).
Counterfactuality:
SM started in 1985 and followed an exponential growth curve since. I think the main enablers were computers[1] and a nerd disgruntled with school. I don’t think the psych research was that important here, Wozniak seems to have done the experiments before knowing the literature. From a brief check on the history of personal computing, it sounds like normal people had them since 1970, idk how common or good enough. This is some evidence against counterfactuality, but I think it’s dwarfed by the arguments in the next paragraph.
Looking forwards, **it seems plausible there’d simply not have been a Wozniak**. Most people working on education focus on schools and they either work in academia (and then not apply the research) or work in schools (which still refuse to use good research or famous well replicated results). Maybe Duolingo and Quizlet[2] (all around 2010) would’ve done it but it also sounds like they’re still inferior at it. In an alternate universe… we might just not have spaced repetition software at all. The most plausible alternate candidate I’ve heard of[3] is Math Academy (see their book-length google doc about their methods), as they seem to have put similarly[4] smart thought into edu-software drawing on good research. They started in 2016, 31 years after SuperMemo.
My takeaways:
Look for applications, and try to optimize. You can sometimes push small effects far enough to unlock new capabilities, if you’re smart about it. Do experiments. Originally it was just some intuitions about how memory worked, but experiments and optimization let you beat self help books (and the latter lets you beat academia).
Control and scale are your friends:
In this case, control via personalized curve fitting, and scale via having computers do it (tens of thousands of cards isn’t even unusual, though it sounds like early on only partial records were kept by the database
I’m reminded of an article about a paper demonstrated nuclear fission that claimed fears of big bombs were unfounded because the reaction was subcritical (and iirc the Manhattan project was already underway) - what changed was control (e.g. neutron reflectors, combining originally separated cores) and scale (“just” get more uranium).
Lastly, you’re more likely to find something where there’s massive civilizational inadequacy (like education), but you probably knew that already.
Quote: > ZX Spectrum 8-bit microcomputer. SuperMemo could not be implemented on ZX Spectrum as the computer lacked disk storage. All programs and data had to be loaded in from a cassette tape. In an overnight simulation, on Feb 22, 1986, I figured out that the buildup of knowledge in spaced repetition is nearly linear, which stands against a popular intuition that backlogs must keep increasing
Piotr Wozniak, the creator of spaced repetition software (aka, the smart way to remember stuff) SuperMemo (henceforth SM) claims (and I believe him) that he was the one that really got spaced repetition going (for example, the famous graph of the serrated decay curve is allegedly actually due to him, not Ebbinghaus (who didn’t look at spaced reviews).
Though the effect was seen in earlier research, he popularized it and also added his nifty algorithm + software. Here’s a quote:
Impact:
It looks like basically all spaced repetition software traces back to SM, for example Anki (surveys say 30-80% of med students use it, and the anki mobile for ios is as of writing the 4th top paid app on the US Apple app store) started with SM’s algorithm, and it’s possible that Duolingo and Quizlet’s use was caused by the greater awareness SM caused or by trying to beat competitors (but this is just speculation).
Counterfactuality:
SM started in 1985 and followed an exponential growth curve since. I think the main enablers were computers[1] and a nerd disgruntled with school. I don’t think the psych research was that important here, Wozniak seems to have done the experiments before knowing the literature. From a brief check on the history of personal computing, it sounds like normal people had them since 1970, idk how common or good enough. This is some evidence against counterfactuality, but I think it’s dwarfed by the arguments in the next paragraph.
Looking forwards, **it seems plausible there’d simply not have been a Wozniak**. Most people working on education focus on schools and they either work in academia (and then not apply the research) or work in schools (which still refuse to use good research or famous well replicated results). Maybe Duolingo and Quizlet[2] (all around 2010) would’ve done it but it also sounds like they’re still inferior at it. In an alternate universe… we might just not have spaced repetition software at all. The most plausible alternate candidate I’ve heard of[3] is Math Academy (see their book-length google doc about their methods), as they seem to have put similarly[4] smart thought into edu-software drawing on good research. They started in 2016, 31 years after SuperMemo.
My takeaways:
Look for applications, and try to optimize. You can sometimes push small effects far enough to unlock new capabilities, if you’re smart about it. Do experiments. Originally it was just some intuitions about how memory worked, but experiments and optimization let you beat self help books (and the latter lets you beat academia).
Control and scale are your friends:
In this case, control via personalized curve fitting, and scale via having computers do it (tens of thousands of cards isn’t even unusual, though it sounds like early on only partial records were kept by the database
I’m reminded of an article about a paper demonstrated nuclear fission that claimed fears of big bombs were unfounded because the reaction was subcritical (and iirc the Manhattan project was already underway) - what changed was control (e.g. neutron reflectors, combining originally separated cores) and scale (“just” get more uranium).
Lastly, you’re more likely to find something where there’s massive civilizational inadequacy (like education), but you probably knew that already.
Quote: > ZX Spectrum 8-bit microcomputer. SuperMemo could not be implemented on ZX Spectrum as the computer lacked disk storage. All programs and data had to be loaded in from a cassette tape. In an overnight simulation, on Feb 22, 1986, I figured out that the buildup of knowledge in spaced repetition is nearly linear, which stands against a popular intuition that backlogs must keep increasing
Maybe Khan Academy? idk if they even use spaced rep.
I haven’t actually searched, I just happened to get linked to Math Academy’s stuff somehow.
Read: competent nerd vibes.