Home → Community Archive →

Quantified self: measuring how well I learn

• Edit

thomas.tempe

01 Feb, 2017 02:37 PM

Hi all,
I've been wondering for quite some time what was the impact of my learning environment on my memorization performance.

To answer that question, I've written the quantified self add-on which exports a log file showing together the date&time on which you learned, and your answer on the following review.

I'm starting this thread in the hope of discussing my findings with other data geeks with large decks and a passion for optimization.

My first results are not too encouraging: I've looked at whether my memorization was as effective during my morning commute (riding a car, in an environment I find quite distracting, with ~40% longer average review times). Based on ~8000 data points, it's slightly better than otherwise (2% positive correlation, which is probably just noise).

This leads me to two questions:
* will anything I do (aerobic exercise, plentiful sleep, smart drugs, right time of day, food intake...) yield a measurable improvement? * how do I determine that my measured improvement is statistically significant?

1. 1 Posted by Vit on 04 Feb, 2017 10:49 PM

   

   Edit December 12 2017:
   Can you build a chart to discover the best value for the Graduating interval ?

   • Edit

2. 2 Posted by ZapBeeb on 05 Feb, 2017 12:04 PM

   

   I think that at least an exercise regimen and a correct sleep will infact improve memorization.
   As for other factors, including precise dietary advices and eventual use of supplements, there are probably right and wrong things to do, but who knows what they are.

   Regarding supplements i often thought of trying some for a given period (unmarked capsules with placebo, supplement1, supplement2... to be taken for a month or so each, not together of course, and then after checking the data), but i never came to do it myself.
   On examine.com there are some that i'd like to try, though.

   • Edit

3. 3 Posted by Eric on 26 Oct, 2017 11:10 PM

   

   Hi Thomas.
   Here are some links you'll probably like to read:

   aerobic exercise:
   http://www.gwern.net/Treadmill#treadmill-effect-on-spaced-repetition-performance-randomized-experiment

   smart drugs:
   http://www.gwern.net/LSD-microdosing

   right time of day and plentiful sleep:
   http://www.gwern.net/Spaced-repetition#when-to-review

   nevertheless he himself did a thorough experiment which might show otherwise:
   https://groups.google.com/forum/#!topic/mnemosyne-proj-users/8yZz9BbqKl4

   • Edit

4. 4 Posted by thomas.tempe on 30 Oct, 2017 01:49 PM

   

   Wow!

   This Gwen is truly incredible.

   I just spent my last 10
   hours of free time going through his website (not that that would be
   enough).

   Very instructive, thanks for the links!

   Thomas

   On
   2017-10-27 07:10, Eric wrote:

   • Edit

5. 5 Posted by Dainius Sileika on 26 Feb, 2020 05:22 PM

   

   Hello!

   I just came across this addon, and I was wondering if you might update it for Anki 2.1? I'm an MSc Psych student who's doing a small project on the difference between cloze and regular cards, and it would be a huge help to be able to export this data, instead of copying it by hand from every card.

   Thanks!

   D

   • Edit

6. 6 Posted by ijg on 29 Feb, 2020 01:27 PM

   

   @Dainius Sileika: Your project sounds interesting. So I just updated the add-on for 2.1 and uploaded it as Quantified Self add-on - export your review log (fork for Anki 2.1). Check if it works for you and let me know.

   You might not need the add-on: Anki's database is sqlite and can be opened without Anki with tools like "DB Browser for SQLite". This allows you to save a table as csv. Then you "just" need to convert the review times from unix time and maybe combine different database tables ...

    

   Thanks to thomas.tempe for creating this add-on and releasing it as free and open source software. Once there's an official update I can remove mine.

   • Edit

7. 7 Posted by Dainius Sileika on 29 Feb, 2020 01:40 PM

   

   Thank you! It looks to be working, although I've already gone in by hand
   and tried to extract info from many of the cards by hand.

   Small sample size this time, it's a "practice" experiment, but especially
   if I end up doing a bigger experiment, this tool will be essential.

   Will let you know, and thanks again!

   -D

   • Edit

8. 8 Posted by thomas.tempe on 01 Mar, 2020 02:21 AM

   

   Hi Dainius, ijg,

   I'm happy that this was useful to you.

   I don't have any plans to further develop or maintain it. Thanks to ijg for your port :)

   Yours sincerely,

   Thomas TEMPE

   • Edit

Comments are closed, but you can start a new discussion.