sequencing and levels

[Andrew Harrington]

Lesson Plans and Levels and Sequencing

David talked about classifying thing under various levels and interests. Here is brainstorming on a more general framework. Several of us discussed this at Pycon.

Site Atoms

Basic site building blocks are individual problems and units of tutorial/content.

The site is certainly a repository for the problems, and probably at least some tutorial/content, and maybe links for frames to external expository content.

Basic sequencing

The most straightforward way to put these building blocks together coherently is for a given instructor or teacher to integrate a lesson plan listing a sequence of expository units and interspersed problems. Many of the same problems in the general pool and some of the expository text can be referenced in many different lesson plans.

Implementation precedence - Very important: Allow teachers to provide sequences of exposition and problems. Allow students to go through sequentially and to jump the the next topic and to see the whole outline and jump around in it.

Ideally the teacher does not upload elaborate web pages for the plans, but just a text file containing a sequence of names of basic content pages, expository or problems, maybe with short plaintext comment sections, and let the system format such data dynamically and generate the html lesson plan with the proper links.

The most minimal thing for the site would be for the student to jump back to the lesson plan after each piece, and then select the next lesson piece.

More sophisticated would be to use the lesson plan to dynamically generate direct links at the bottom of content pages to the next one. Even more sophisticated would be to have several choices for links. For instance suppose the lesson plan lists the current expository section followed by several problems at the same level for drill and then several more problems of greater challenge but on the same basic topic covered by the exposition, and then more exposition: The links at the end of the current exposition or a problem after it could refer to the next problem at the same level of difficulty, the first harder problem, and to the next exposition. We would of course have to agree how this difficulty level is encoded. The simplest would be something like listing all the problems for drill at the same level on the same line....

Sequencing with more content

If there is not much content, or everyone basically develops things independently, the scheme above would be sufficient. If we aspire to having much content and cooperation and splicing, however, then we could have way too much data organized only with completely independent schemes in the lesson plans. We would find it very time consuming to manually check out and integrate with other authors' stuff.

I would like a system where authors are encouraged to add meta-data tags to their submissions, and they are encouraged to use a mostly consistent scheme. The most obvious tags are for prerequisite knowledge and learning outcomes intended (for exposition) or to be demonstrated (for problems). I volunteer to make a first pass at a taxonomy of Python/programming knowledge areas, broken down very finely, if there is positive feedback to this idea.

There are further tags/metrics that would be useful – level of creativity needed, level of integration of ideas needed (think how many steps ahead?), maybe a number from 0-100 suggesting the level of mastery indicated. This last part is subjective, but useful none the less. Any good teacher has to have such ideas in mind. An objective but not necessarily useful automatic measure would be the number of lines in the solution code.

Such meta-data would not be required – we want problem author and lesson plan authors to have a low barrier to entry. Still, if you are looking to avoid writing everything yourself, and looking more efficiently for pieces already out there, then being able to quickly read or search on such meta-data could be very useful.  (And site editors reviewing a problem might add it.)

We would always allow authors to add further meta-data. Another marker might be reading level.... Also a lesson plan could develop example situations that are used later. An earlier example would not be part of the formal Python learning taxonomy, but it could be an important prerequisite for a later piece. (Naming conventions for such special prerequisites would be good.)

Using data gathered from users

One big advantage of a server driven web site over something like Crunchy on an individual machine is the ability to continuously gather usage statistics. This data could be analyzed, and we could see that students took many tries on one problem. Maybe we need better exposition, or more problems at a lower level of creativity first.... Educators could look at this data manually and modify a canned lesson plan.

AI

With more programming and research (some time from now!), we could also use data about a users responses to dynamically change things in real time: put some AI in based on how well the student is doing, maybe guess level of competency, brightness, reading level... and have at least a suggested next problem or decide/suggest that it is time to repeat exposition or give alternate exposition or go on to a new topic, and what density and level of repetition and chunk size for the next exposition....

--
Andrew N. Harrington
 Director of Academic Programs
 Computer Science Department
 Loyola University Chicago
 512B Lewis Towers (office)
 Snail mail to Lewis Towers 416
 820 North Michigan Avenue
 Chicago, Illinois 60611

http://www.cs.luc.edu/~anh
Phone: 312-915-7982
Fax:    312-915-7998
g...@cs.luc.edu for graduate administration
u...@cs.luc.edu for undergrad administration
aha...@luc.edu as professor