P4PP ft Moo

[TathagataChakraborti]

A while back, I helped build a prototype on ITS x Planning (called Moo), with a focus on model reconciliation between the student model and the user model. Link: https://yochan-lab.github.io/papers/files/papers/automated-planning-intelligent.pdf

The overall idea is to reconcile the student model to the system model through a series of tutorials and exams (curriculum planning) + for a given assignment give tips and hints based on the current estimated model + how the solution diverges from the solution of the reference model.

More gory details of the curriculum planning angle here (page343), the paper only has the per problem stuff: https://keep.lib.asu.edu/_flysystem/fedora/c7/206012/Chakraborti_asu_0010E_18520.pdf#page=343

Of course, it's not the same thing as P4PP but it piques my interest coz you have a source model and target model already so it's a much easier place to apply this stuff rather than attempting to estimate the student model.

Now, the target of ITSs is a bit different because it's automated and is supposed to help the user (student) get to the correct model (as opposed to solving and grading). Nevertheless, I think we can find some interesting overlap there coz you have a target model hanging around and we can do fun model space search stuff with it for tips and tricks.

Leaving expression of interest here, in case there is interest to develop this further. 😼

[haz]

This is absolutely on point. Rather than run p4pp as a straight-up grading pass, we could host it remotely and provide the feedback to the student so they iteratively improve what they're doing. Perfect for low-stakes assignments.

[haz]

I've gone and promised the students self-assessment support on all their assignments, and so will need to deliver this over the next couple of weeks :P .

[haz]

Or, rather, #2

That Moo project looks pretty awesome...it'll be quite some time before I get there...