Seven Years of Spaced Repetition in the Classroom - tanagrabeast

Seven Years of Spaced Repetition Software in the Classroom

“tanagrabeast” (TB) wrote on LessWrong in 2015 and 2016 about experiments using Spaced repetition memory system in an American high school English classroom. This 2021 post follows up on that work, with the conclusion that the author doesn’t use SRS nearly so much in their classes.

Part 1 (seven-year retro)

• TB fears that if successful, his app would produce a dystopia with frustrating rote review, where “millions of children cursed my name.” Notably, he suggests that he himself would have abused it as a younger teacher by adding too many cards per day (7 is apparently far too many?)
  • “In fact, for the last four years, I’ve only used it during a two-to-three month span leading up to the state test.”
• He adds 10 cards/week, aiming for 2min of study per day. He does whole-class review, defeating adaptive scheduling. Because it’s call-and-response, there must be only one, short, correct answer.
• He questions the central frame: that a key problem with learning is long-term forgetting. He thinks forgetting is bad, but remembering is often not worth the costs, particularly when time-rival with methods which might produce more reflective insight.
  • He thinks SRS is most valuable when automaticity is important, and desirable “between 1 week and 3 months in the future.” (makes sense)
• He believes (rightly, I think) that in a classroom context, apathy is the main problem, not forgetting. SRS makes this worse because it’s boring.

Part 2 (year 3 experiment with individual SRS)

• Used Cerego for independent SRS. Believes this was worse than whole-class method, except for highly motivated students.
• Used 10-12/57min of each class for SRS. Gave a quiz once per week.
• Lots of small logistical problems, mostly having to do with emotional disconnection.

My reactions

I think TB is right to abandon whole-classroom SRS for things like word roots, and he’s right that in a classroom context, apathy/engagement are the high-order bits. Adding rote practice to an already disengaging context is not a good solution, particularly in a topic where memory is not a high-order bit.