Hacker News new | ask | show | jobs
by alessivs 1791 days ago
Every mentioned problem (workload, new vs. old reading task distribution, scheduling, plus retention) is exactly what incremental reading–at least, as implemented in the other prominent SRS–aims to solve (which it does, rather well). Does the Anki IR plugin not tackle them?
1 comments
jldugger 1791 days ago
> (workload, new vs. old reading task distribution, scheduling, plus retention)

I've never really understood the appeal of IR as defined elsewhere. My reading queue isn't just wikipedia dumps or some other 100 percent valuable source, often it's somewhat speculative. I really don't want to read 3 sections from random journal papers and then create cards; I need to once over a paper to even be sure the cards created are worth remembering.

Also, I bought a Kindle to read papers from more comfortable locations than a desk, with less direct light, and AFAIK, that is not directly compatible with the IR approaches that want to track reading progress. So I have no idea if these problems are solved, because I've never felt compelled to try.

link
alessivs 1790 days ago
> I need to once over a paper to even be sure the cards created are worth remembering.

In my experience, this task is still better when distributed over time. Time-distributed reading involves distributed judgement of quality.

What Incremental Reading would do for you, given enough sources and/or sufficient coverage of the subject matter, is effortlessly resurface highlights and cards whose usefulness was initially doubtful. Each resurfacing would be in a successively reshaped knowledge network that stands a better chance to help you understand its merits and role between other pieces of knowledge: between reviews, this network is forming and expanding in an initial stage, to then contract, solidify, and be pruned of noise.

When evaluating the usefulness of the card or piece of knowledge, without IR one may fall back to simply taking a snapshot of the network at t1 and t2 and judge on those premises. With IR you will have made the piece integral to the network the moment it is created, and will evolve as part of it. In the end, this piece may not be simply kept or discarded, but perhaps reassigned to complement a different source, or morphed into answering a different question. Whatever the result, it has better chance to serve your purposes (and even this capacity may be reassessed in the future).

link