Mox Box

LON-CAPA is a learning management system (LMS) that excels particularly in its quizzing engine.  Locally on UI's campus, rather than trying hard to prevent students from collaborating on homework problems, some departments use L-C because every student sees a computer-generated variation of the same problem.  Thus, sharing an answer (like "5") is meaningless.  Rather, students can share the "process" by which they arrived at their own answer, effectively meaning students are helping to teach one another. My department is looking to develop our LON-CAPA capability so as to provide support resources to campus.  So, this week, two of my colleagues and I are attending a LON-CAPA Workshop at MSU (even if it is a bit painful to be surrounded by so much green!).  Today's session is presented by Stuart, one of the key folks behind LON-CAPA, though the two original authors are also present in the room.   Presentation notes:

• Marriage of CAPA (created in 1992) and LON in about 2000
• 108 US institutions use L-C; international usage covers every continent except Australia and Antarctica; supports localization
• About 50-50 Universities and high schools; some publishers are starting to develop L-C resources
• At its base, a shared cross-institutional resource library, which makes it unique from other LMSs/CMSs
• 450,000 shared resources available (approximately 200,000 of which are randomization-based problems)
• Dynamically builds navigation between resources
• Each resource has dynamic metadata when it's used in a course, including
  • Number of accesses
  • Preceding and subsequent resources were used in the course (context of the resource)
  • Courses it is used in
  • Number of tries to successfully complete it and degree of difficulty
• K12 has a dedicated subset repository called "TheDump"
• Course assembly (course content menu) shows resources and metadata (due date, discussion activity, etc.)
• "What's New" page acts as an instructor's dashboard, shows problems to be hand-graded, active discussions, difficult problems based upon stats, problems with errors, etc.
• Even more with statistics: can view individual student performance and class aggregate data (responses chosen upon each attempt, etc.)
• Contextualized discussions--essentially a discussion board dedicated to each "problem" or question in L-C
• Student collaboration via mail, discussion boards (both attached to each problem and more generic ones), synchronous chat, portfolios (students upload their own files and set access), and blogs.
• Groups: can provide each group their own chat, discussion boards, shared portfolio, etc.
• With a history in physics (woo-hoo!) and math, is very strong at rendering math (via <m></m> tag containing TeX markup)
• Integrated with Maxima, a computer algebra system, useful in evaluating formulas, constructed with DragMath Visual Editor
• Can import from other LMSes--showed example of exporting Blackboard (via IMS) into L-C.  Supports older versions of Blackboard, WebCT, WebCT Vista, and older Angel systems; looking to add more recent LMSes including Moodle.  Can also import certain testbank formats.
• Can also clone a course, which takes all course data except student data.
• Supports iClicker and iClicker2 systems via upload of iClicker data

An interesting side-conversation about cheating also ensued, suggesting that even with computer-randomized problems, there are limitations:

• What if someone publishes a problem that has been reengineered to accept the numeric values presented by L-C and spit out the right answer (very few questions have been reverse-engineered in this way)
• Then the discussion jumped to research that shows cheating does not lead to learning, thus can be caught on the back-side via exam scores

More to come, later, if appropriate.

Presented by Jill Joline Myers, Western Illinois University and Leaunda S. Hemphill, University of Illinois at Urbana-Champaign

Educational examples of metadata:

• ID3 tags in MP3s
• Word document authorship information
• Revision history in wikis, Word
• Page view/access date/time 

Patterns in metadata are highly indicitative of an individual--e.g. telephone calling patterns are largely identical for a person, even if the person changes phone numbers.

Metadata is usually automatically generated by the application.

Delete does not necessarily mean gone.

Types of metadata:

• Substantive (types of content changes made)
• System-based (info created by the user)
• Embedded (appears in files, like graphics, links, etc.)

Exposure of metadata may violate confidentiality.

Several examples of the risks associated with metadata, namely in the realm of crime, e.g. a photo sent out via Twitter from a smartphone which included precise lat/long of where the pic was taken, making predators aware of the precise location of a potential target.

"The benefits of metadata far outweigh the risks of metadata.  However, we must become educated about metadata and ensure any risk-inducing metadata is not shared publicly.