Re: About the proect of prioritizing a task list

[Robert <http://profiles.yahoo.com/robertb>]

The idea would be that there are probabilities of certain tokens
"following" other tokens, and that, taken together, there could be
probabilities that these some set of tokens would follow another set of
tokens.  (In my mind, I'm thinking the "set of tokens" would be all the
tokens in a task.)  Go through all combinations of tasks, this could find
which ones are the mostly likely to follow other ones.  (How that is
ultimately done -- i.e., whether they could simply be chained together to
come up with final list -- is unknown.)

I hope that makes sense.

 > From: Bill Bruns <http://profiles.yahoo.com/billbruns>
 > Date: Mon, 17 Jul 2017 21:48:12 +0000 (UTC)
 >
 > Robert,
 > A Markov Chain is essentially a state diagram with probabilities attached to 
 > the state transitions.It carries no information about what triggers a state 
 > change so the probabilities are applied randomly.How do you see these chains 
 > being applied here? 
 > Bill Bruns
 > 
 >  From: Robert <http://profiles.yahoo.com/robertb>
 >  To: Bill Bruns <http://profiles.yahoo.com/billbruns> 
 >  Sent: Monday, July 17, 2017 7:13 AM
 >  Subject: Re: About the proect of prioritizing a task list
 >    
 >  > From: Bill Bruns <http://profiles.yahoo.com/billbruns>
 >  > Date: Mon, 17 Jul 2017 02:57:49 +0000 (UTC)
 >  >
 >  > In case you have not seen these, here are a couple of 2 by 2 task 
 >  > matrices,one 
 >  > from Stephen Copvey and one supposedly from Eisenhower:
 >  > Covey:https://sidsavara.com/coveys-time-management-matrix-illustrated/
 >  > Eisenhower:
 >  > http://www.artofmanliness.com/2013/10/23/eisenhower-decision-matrix/
 >  > 
 >  > ========================================
 > 
 > Thanks for these.
 > 
 >  > Regarding your sample list of four tasks,it is entirely possible that
 >  > you will perform* set "sleep" on primary MP3 player first, just
 >  > because you have the MP3 Â player in your hand,while the* pay
 >  > $91.66 June 2017 Credo bill may require the tedium of finding pen,
 >  > checkbook, and stamp (if not an electronic bill).Or at least the
 >  > tedium of logging into an account.
 > 
 > True.  Nevertheless, the inefficiency of returning the MP3 player to its
 > rightful place and gathering the materials and paying the Credo bill
 > should be done first.  Obviously, not ideal w.r.t. steps or a bigger
 > situation, but the most practical for addressing what's most important.
 > 
 >  > There will be times when some task becomes extremely urgent without
 >  > changing its normal priority, for example "take the garbage out RIGHT
 >  > NOW because the truck is coming down the street!"
 > 
 > Good example.  But, if I forgot to take out the garbage because I didn't
 > execute that task, then there is a bigger problem with my todo list and
 > its system.
 > 
 > (This does bring up the possibility of "task generation".  But, let's put
 > that on the back-burner for now.)
 > 
 >  > Then there is the question of how does the list "know" when the tasks
 >  > were performed?
 > 
 > Checkmark, strike-out, removal -- doesn't matter for this project.
 > 
 >  > You may mark multiple tasks as "done" at the same
 >  > time, because you are human, so you mentally "batch up" a group of
 >  > tasks to be marked, rather than being perfectly mechanical in marking
 >  > each task immediately upon performing it.
 > 
 > True.  I, because I will know the algorithm, would mark as done tasks in
 > order of importance.  However, for someone who doesn't know the algorithm,
 > after a number of examples of completion, the algorithm will and should
 > learn what really are the most important tasks.
 > 
 >  > Part of the solution, I
 >  > think, is that the list must present a "short list" of tasks (perhaps
 >  > 3, or 5, or 10), rather than just one, as the "top".
 > 
 > In my mind, the entire task list should be ordered by importance.  If
 > those short list of tasks happen to appear at or near the top, then that's
 > fine.
 > 
 >  > Then, the human
 >  > being makes a decision of which task in the set to do next.
 > 
 > The human should always execute from top to bottom, unless they see that
 > something should be done before something else.  That's where the learning
 > (or "evolution") takes place.
 > 
 >  > I think of this as me having an AI "butler" or "good friend" that is
 >  > always at my side, and is aware of the environment around me to some
 >  > extent.  For example, "he" is aware when I take something out of the
 >  > refrigerator,and can ask, "should I put something on the shopping
 >  > list?  Or is aware of the sound of putting an object down,and can cask
 >  > "shall I make a note of what you left and where?"
 > 
 > That's undoubtedly a worthwhile long term goal, but my immediate need to
 > simply order my task list.
 > 
 >  > Of course, that is beyond the scope of the project you mentioned, but
 >  > it is my personal belief that it is where we are headed.  (Keys would
 >  > be recognized by sound. Â Location could be known and
 >  > recognized as "typical: or not.)  OK, OK, yes, this is far afield.  So,
 >  > let's just talk about arranging the list by priority, which is
 >  > recognized by the order of performing the tasks.  Even so, shall a task
 >  > be popped to the top of the list just because it was the most
 >  > recently performed?
 > 
 > It should be placed in the appropriate place based on when it was
 > completed relative to all other tasks, not simply because it was recently
 > performed.
 > 
 >  > Perhaps instead it has a priority "score" which is
 >  > incremented when it is performed.  This raises its place on the list,
 >  > but not necessarily right to the top.
 > 
 > That is a possibility.
 > 
 >  > Initially each task starts with a score of zero, and receives
 >  > "points" when it is performed.  So then it does pop to the top at
 >  > first, but as other tasks are performed, their scores rise above it.
 > 
 > Again, yes, this is a possibility.
 > 
 >  > Perhaps when I revisit this I will think in terms of probability
 >  > instead of scores.
 > 
 > Yes.  That's why the Markov Chain concept was floated.  (Alex was the one
 > that suggested that and, after playing with some other ideas, I think that
 > may be the most lucrative.)
 > 
 > Also: I wanted these tasks to be generalizable.  For example, if I pay
 > bills and all those tasks have "pay" in them, then, in general, if a new
 > billing regime is introduced (e.g., I sign up for a new service), then
 > such bills should be prioritized likewise.  Thus, making the relationships
 > "token based" (or "word based") would make the most sense.  (In the
 > future, using something like WordNet could be one way to expand the
 > generalization.  But, that would be an enhancement.)