tuxmath-devel

[Tim H. <ho...@wu...>]

Oops, forgot to CC the list.
--Tim
----------  Forwarded Message  ----------

Subject: Re: [Tuxmath-devel] GSoC project on "Personalized, adaptive game-play 
to improve learning in Tuxmath"
Date: Friday, May 20, 2011, 12:05:55 pm
From: Tim Holy <ho...@wu...>
To: Siddharth Kothari <sid...@gm...>

Hi Siddharth,

On Friday, May 20, 2011 11:04:52 am you wrote:
> > 1. When you say, "topic suggestion," are you thinking of some interactive
> > interface...
> 
> I had an interactive interface in mind, where the final decision would be
> on the player if he would like to follow the topic the game suggests or
> choose another one as per his wish.

Sounds like a good idea. If we got to the point where we "trusted" the 
results, we might want to give teachers an option to disable the choice. But I 
agree that this is the right way to start.

> > 2. What are your thoughts about per-user storage?
> 
> I will initially implement a storage structure for the Linux model, and
> than for the "school mode". Imo, this should be more valuable if I
> implement it for the "school mode", but I am not familiar with the code. I
> will try to do that :)

Again, good choices. If you want to learn more about the school mode, I did 
write up some documentation on how to use it. If you can't find it, let me know 
and I will look for it myself.

> 
> For 3.), modelling how much challenge a user is willing to take seems
> difficult. Rather than making decision on the latency in re-starting the
> game, we can base it on the progress: If the student is able to save the
> igloos, adjust the other game variables to provide more challenge; if a
> student is performing poorly, we may want to slow the speed of comets and
> likewise, adjust other game variables. But again, this approach assumes the
> amount of difficulty one is willing to take. What are your opinions?

Agreed on all counts---performance is much easier to measure, and far less 
ambiguous than trying to interpret latency. But (as you seem to understand), 
perfomance by itself doesn't define whether a student is doing "well" or 
"poorly," because there's the separate question of how high one's standards 
are. 

Maybe backing up a level: I suggested frequency-of-play simply because I 
envision that if we maximize the amount of time students play the game, we 
might also (roughly) maximize their learning rate. So the notion would be that 
the rate of ratcheting up the difficulty level could be tuned to try to maximize 
the total amount of time spent playing. But this is not trivial to get right. 

Now that I think about it, the user interface choosing the particular lesson 
solves much of the problem for you, particularly if you present the user with 
a choice of recommended lessons (maybe 3 choices, labeled something like 
"Challenge", "Middle"? (that name stinks, but can't think of something better 
now), and "Easy"). Not only does that give the user some control, but you may 
also be able to draw some conclusions if the user repeatedly chooses 
"Challenge" and use that to modify your choice of options.

> 
> > 4. ... how to decide which _type_ of problems to pose....
> 
> There is an interesting approach
> Animalwatch<http://animalwatch.arizona.edu/> (It's
> a tutor for pre-algebra mathematics) uses for this which is described
> here Using
> the student model to control problem
> difficulty<http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.38.7068
> &rep=rep1&type=pdf>. I am thinking of using a approach on similar lines,
> will post it shortly.

i'll read about it.

> 
> 5. Given this tuxmath-as-the-teacher mode of thinking, do we need to think
> 
> > somewhat about creating mechanisms to explicitly teach, rather than just
> > let
> > them practice things they've already been taught?
> 
> iiuc, this means using different teaching strategies to cater to the
> varying needs of the students (like some may like to receive more feedback
> to stay motivated, a pedagogical agent may be more of help to others).

Sorry, what I meant was "gosh, this student has completely mastered addition 
and subtraction, and is ready for more challenge. But no one has ever taught 
him/her multiplication, so we're stuck." In other words, tuxmath lets kids 
practice stuff they've been taught, but does not in and of itself teach them 
mathematics.

--Tim


> But
> given that the domain here is elementary Mathematics, I find it hard to
> think of an explicit teaching mechanism. Do you have some suggestions?
> 
> Thanks,
> Siddharth
> 
> On Fri, May 20, 2011 at 7:29 PM, Tim Holy <ho...@wu...> wrote:
> > Hi Siddarth,
> > 
> > I like it very much. Some concrete feedback, mostly a set of questions
> > asking
> > for more detail:
> > 
> > 1. When you say, "topic suggestion," are you thinking of some interactive
> > interface from which the user chooses a task? Or is the "suggestion" an
> > internal one, and the game will execute whatever parameters are
> > suggested?
> > 
> > 2. What are your thoughts about per-user storage? Are you going to employ
> > the
> > Linux model (store in the user's home directory), or are you going to use
> > the
> > infrastructure in the "school mode"? Or are you planning on not saving
> > data across sessions at all? (Without saving, it will be hard to test
> > and debug, so
> > I don't recommend the latter.)
> > 
> > 3. I have the sense that kids differ quite a lot in terms of how much
> > negative-
> > outcome they can tolerate: some will plunge right back in after they lose
> > all
> > their igloos, others will be bothered if even a single igloo gets hit by
> > a comet. Those that enjoy the tension will appreciate being challenged,
> > those that do not will probably feel more comfortable if they are
> > consistently performing at 95% or higher.
> > It would be really interesting if it were possible to model not just the
> > student's mastery, but also what level of failure/challenge is acceptable
> > to
> > the student. Naturally, this seems hard. One option would be to ask for
> > feedback from the student, and a second might be to pay attention to the
> > latency in re-starting a new game of tuxmath. If it is a few seconds, the
> > kid
> > seems ready for more. If it's a few months, perhaps the kid got scared
> > off. But
> > obviously it's questionable how reliable such information will be.
> > 
> > 4. In your section on "modifying the game variables," certainly the speed
> > is
> > an important parameter. But a more interesting---and more conceptually
> > challenging---issue is how to decide which _type_ of problems to pose. Is
> > the
> > kid ready for addition facts up to 10, or should we stick with 0 to 5?
> > How can
> > you tell when the kid is ready for multiplication?
> > 
> > 5. Given this tuxmath-as-the-teacher mode of thinking, do we need to
> > think somewhat about creating mechanisms to explicitly teach, rather
> > than just let
> > them practice things they've already been taught?
> > 
> > Again, it looks like lots of fun ahead. I'm glad you're tackling this.
> > 
> > Best,
> > --Tim
> > 
> > On Friday, May 20, 2011 07:43:52 am Siddharth Kothari wrote:
> > > Hi everyone,
> > > 
> > > I made a blog (gscbbn.wordpress.com) which I will keep updating the
> > 
> > details
> > 
> > > with regards to the project.
> > > 
> > > I would appreciate a feedback. The approach and deliverables are
> > 
> > tentative,
> > 
> > > let me know if there is something I should incorporate.
> > > 
> > > Thanks,
> > > Siddharth
> > > 
> > > On Wed, Apr 27, 2011 at 10:11 PM, Tim Holy <ho...@wu...> wrote:
> > > > On Wednesday, April 27, 2011 08:01:22 am Siddharth Kothari wrote:
> > > > > Hi everyone,
> > > > > 
> > > > > My project proposal on "Personalized, adaptive game-play to improve
> > > > > learning in Tuxmath" has been accepted in GSoC
> > > > > (link<
> > > > 
> > > > http://www.google-melange.com/gsoc/project/google/gsoc2011/sidi/14001
> > > > 
> > > > > >). Bart Massey from PSU will be my mentor who would be helping me
> > > > > >in the
> > > > > 
> > > > > design of the Bayesian network for knowledge representation and
> > 
> > making
> > 
> > > > > inferences based on it. I look forward to the Tuxmath community
> > 
> > members
> > 
> > > > to
> > > > 
> > > > > have discussions regarding the implementation details, and some
> > > > 
> > > > occasional
> > > > 
> > > > > code-related help. I will start exploring the code-base in about a
> > 
> > week
> > 
> > > > > once my semester exams are over.
> > > > 
> > > > That's great! Congrats!
> > > > 
> > > > --Tim
> > 
> > -------------------------------------------------------------------------
> > 
> > > > ----- WhatsUp Gold - Download Free Network Management Software
> > > > The most intuitive, comprehensive, and cost-effective network
> > > > management toolset available today.  Delivers lowest initial
> > > > acquisition cost and overall TCO of any competing solution.
> > > > http://p.sf.net/sfu/whatsupgold-sd
> > > > _______________________________________________
> > > > Tuxmath-devel mailing list
> > > > Tux...@li...
> > > > https://lists.sourceforge.net/lists/listinfo/tuxmath-devel

-----------------------------------------