learning system notes

gistfile1.txt

* each "card" has:
  * one question
  * set of answers (correct and incorrect)
  * pointers to look up answers

* forum to discuss cards

* the traditional bottom-up approach based on learning individual elements
  in isolation may sometimes not work so well.  it may be that sometimes
  it is better to try to start by teaching / learning simple combinations.  
  for example, it might be easier to learn a word or short sentence in a 
  foreign language compared to learning an individual sound (especially a 
  sound that isn't "present" in a language one already knows).  specific
  cases where this was experienced include peg specials and the hindi 
  alphabet as presented in certain places (it's not uncommon for this to
  be the case in learning any language).
  
* use of concrete examples and concepts before introduction of terminology. 
  in other words, defer discussing labels and names of things until after
  a certain amount of coverage of the target concept / example.
  start concrete, abstract later.  no more "foo bar".
  
* limit the amount of unfamiliarity per session.  too much unfamiliarity at
  once can lead to unnecessary confusion.  if something must be introduced,
  try to do it in the context of familiar elements.

* if possible, provide several examples that illustrate a point.  this may
  increase the number of people who understand at least one of the examples
  and hence the point.
  
* keep client as simple as possible -- populate db with question / answer
  pairs that are static.  can be creative about populating part -- away
  from client code.
  
* for each question / answer pair, could have list of "easier versions"
  -- these could be fetched along with initial pairs so that a user can
  request "hints" / "easier version" of a question.