1. The LearningOnline Network with CAPA (LON-CAPA) Gerd Kortemeyer,Wolfgang Bauer, Deborah Kashy, Edwin Kashy, Cheryl Speier Michigan State University

2. What is LON-CAPA? LON-CAPA stands for LearningOnline Network with a Computer-Assisted Personalized Approach. developed and implemented by a group of faculty and professionals provides instructors with a common, scalable platform to assist in all aspects of teaching a course, from lecture preparation to administration of homework assignments and exams “Distributed Learning Content Management and Assessment System”

3. “Distributed” LON-CAPA is built as a geographically distributed network of constantly connected servers

4. “Distributed”: Domains The network is logically divided into domains such as “MSU”, “FSU” or “Publisher X” Domains limit the flow of user information Domains can limit access to content resources Domains limit the extent of user privileges

5. “Distributed”: Users Users can log into any machines in the network Users can access courses/resources from anywhere on the network under one username Users can have roles and associated privileges for any resources, data, and functionality on the network

6. “Distributed”: Server Types Two classes of servers: access servers and library servers Access servers: host user sessions – processing Library servers: every user and course has a home server in the network which holds all of their resources and data – storage backend

7. Access servers can have a list of spare access servers to offload sessions depending on own work load Additional round robin IP scheme possible MSU setup: “Distrubuted”: Load Balancing msu.lon-capa.org s1 Library Server s3s2 s4 Round Robbin Spares

8. “Distributed”: Dynamic Replication When a content resource is first requested, the access server finds the home library server of the author/owner, subscribes to the resource, and copies it in the background. When resource is modified on its library server, subscribed access servers are notified. Depending on last date of local access, the access servers copy the new version over, or delete local cached copy.

9. “Learning Content Management” Allows instructors to create educational materials and to share such learning resources with colleagues across departments and institutions All instructional content goes into cross- institutional shared repository and is cataloged LON-CAPA provides online tool, “Resource Assembly Tool” (RAT), to combine content into Custom Online Course Pack

10. “Learning Content Management” Pages can be constructed from fragments and other pages Sequences can be constructed from fragments, pages, and other sequences Courses point to (top-level) sequences Maps at every level are simply other content resources

11. “Learning Content Management” Graphical resource assembly tool (RAT) to construct maps Working on branching based on conditions using performance data, course data, and preferences Individualized curriculum

12. “Content Management”: Handlers Every resource in the system, both content and system programs, is called by URL – all resources can be bookmarked. Any resource can be processed on-the-fly while being delivered to the user. User calls for a resource, LON-CAPA finds appropriate chain of handlers, processes resource for desired target, and sends it out. Chain of handlers can also cover several transactions. Example: user calls for URL, but needs to login first and then pick course. All handlers can interact with session environment

13. “Content Management”: Document-driven processing Example: user calls for sample.xml sample.xml Uses my.style XML Handler triggers my.style Defines tags for targets loads User Session Environment Target Prefe- rences

14. “Content Management”: XML/MathML/HTML XML handler provides rendering of XML and HTML to targets XML, HTML, MathML, and LaTeX Math rendering capability: LaTeX can be inserted into HTML and XML documents between and tags. Identity The identity $\sin^2(\omega t)+ \cos^2(\omega t)=1$ allows us to… Math fragments are rendered into symbol fonts for target HTML, and MathML for MathML-capable browsers using a derivative of tth/ttm.

15. “Content Management”: LaTeX LaTeX files can be directly put onto the file system, and will be automatically handled as if they had one large … around them

16. “Individualized Assessment” Individualized problems: different numbers, different graphs, different options, … “Classical” online homework types: multiple choice, option response, mix-and-match, string, etc “STEM” types: numerical, multicomponent numerical, physical units, symbolic math, individualized simulations Combination of the above types Adaptive immediate feedback Multiple attempts

17. “Individualized Assessment” Example: Individualized graph, numerical answer

18. “Individualized Assessment” Example: Individualized labels, options

19. “Individualized Assessment” Example: Multicomponent numerical with individualized animation

20. Some of the History

21. 1992 CAPA Started by Ed Kashy et al. in Cyclotron Individualized assessment system for science and math Immediate feedback, multiple tries - mastery based Used paper copies of assignments and terminal input X-Windows problem editing Got Web student interface in 95

22. 1997 LectureOnline Started by Wolfgang Bauer, Walt Benenson, Gary Westfall, and Gerd Kortemeyer in Cyclotron Learning content management and individualized assessment system for science and math Sharing of content between courses Completely web-based interface

23. 1999 LON-CAPA The LearningOnline Network with CAPA Collaboration of CAPA and LectureOnline groups “The best of both worlds” Sharing of content between courses and institutions Reusability of content on different levels of granularity Distributed and Scalable

24. Who are the users (at MSU)?

25. Who uses it? Students (“It”: CAPA, LectureOnline, LON-CAPA family)

26. Who uses it? Students

27. Who uses it? Courses

28. Who uses it? Faculty

29. How is it used? Material written by faculty teaching course or “re-used” from other faculty Homework in addition to “traditional” lecture and textbook “Traditional” lecture, homework and textbook online VU courses AP courses Prelab quizzes In-class exercises

30. LON-CAPA versus BlackBoard LON-CAPA has no licensing cost, BlackBoard does :-) LON-CAPA is open-source GNU GPL, and can be adapted, enhanced, debugged, …, BlackBoard cannot :-) BlackBoard does have systems-support, lots of documentation, training material, …, LON-CAPA does not :-(

31. LON-CAPA versus BlackBoard LON-CAPA was initially developed primarily by STEM faculty for “power users” LON-CAPA makes the hard things possible, but currently does not yet make many easy things easy BlackBoard makes easy things easy, but leaves many hard things impossible

32. Faculty Perspective “My problems are computational and not multiple choice, the types of grading mechanisms built into standard instructional delivery platforms (e.g., Blackboard) would not be sufficient for me”. Professor Paul Rubin teaching MBA804

33. Faculty Perspective “Although [LON-CAPA’s] potential benefits as a distributed-network resource indexing system far outweigh what could be achieved by an off the shelf course management system (e.g., Blackboard, WebCT) it still needs considerable interface development and user testing” LBS144 Professor Janet McCray Batzli Challenge: keep hard things possible, make easy things easy

34. LON-CAPA versus BlackBoard Making the easy things easy: Recently added Chatroom Document Upload (PowerPoint, Word, etc) Online Contextual Help Significant Documentation Effort User and Administrator Mailing List Easy RedHat 7.3 Installation 3-Command-Upgrade Sequence help.lon-capa.org

35. Chronicle of Higher Education December 21, 2001 BlackBoard is affordable for now … (But) higher-education officials expect that courseware companies will raise their prices after creating a market for the software, and after seeing course-management systems become an indispensable part of the academic-computing infrastructure on most campuses. […] As the company integrates more functions into Blackboard, its executives anticipate signing many more "$200,000-, $300,000-, and $400,000-a-year relationships"

36. LON-CAPA vs. BlackBoard LON-CAPA runs on Linux (target for installs: RedHat distribution) Apache webserver mod_perl for server-side handlers JavaScript for client-side functionality MySQL as database for metadata searches LaTeX, Gnuplot MuPAD for symbolic manipulations (planned)

37. Chronicle of Higher Education December 21, 2001 (And) whatever the current faults of WebCT and Blackboard, many academic administrators say they are preferable to homegrown systems. […] Princeton developed its own course- management system, beginning in 1997. Two years later, the university decided to abandon its homegrown system in favor of Blackboard. Why?

38. Chronicle of Higher Education December 21, 2001 "The problem with homegrown systems is you typically have one or two staff people who know how the system works, and those people might leave or decide to do something else," Mr. Goldstein [Director of Academic Services, Princeton] says. Problem: “one or two staff people” Solution: Collaboration with other institutions Workshops Conferences Investment in documentation, etc

39. Summary:Why do it? Solid, scalable platform Easily expandable Customizable The hard things are done Control over codebase Standard open-source tools and libraries Independence No licensing fee for both application and operating system Forming academic consortium “Linux” model of sustainability

40. The Project “Investigation of a Model for Online Resource Creation and Sharing in Education Settings” Uses LON-CAPA as model system Five year project, funded by the National Science Foundation Information Technology Research Program

41. The Project: Research Goals Open source code development concept for Instructional Management System Pooling online educational resources between departments, colleges, universities and schools Business models for “trading” such resources on small level of granularity Metrics for resource quality and effectiveness Curriculum adaptivity to learner types and remediation needs

42. The Project: Pilot Users, Boards Pilot users: 6 high schools, 2 community colleges, 1 four year college, 12 universities Business advisory board: McGraw-Hill, Saunders, Wiley, Freeman/Worth, Dell, RedHat, Apex Learning, university representatives Three member Evaluation Board Four high school teachers in Research Experience for Teachers (RET) supplement

43. The Project: National STEM Digital Library Supplement to cross-integrate LON-CAPA with NSF National STEM Digital Library NSDL available to LON-CAPA as content domain LON-CAPA resource pool available in NSDL as associated library LON-CAPA can be used as course delivery tool for NSDL content

44. The Project: Status Started October 2000 First user meetings with 50 faculty from other universities at MSU Second user meeting at Florida State University with 56 faculty from 22 institutions Next user meeting at Truckee Meadows Community College, Reno Programmer Workshops at MSU in Summer

45. Status: Installations High Schools Mio AuSable High School - Mio, Michigan Charlotte High School - Charlotte, Michigan Fowlerville High School - Fowlerville. Michigan Theodore Roosevelt High School - Wyandotte, Michigan Community Colleges Westshore Community College - Scottsville, Michigan Truckee Meadows Community College - Reno, Nevada

46. Status: Installations Universities (department level) Ohio University at Athens - Athens, Ohio SUNY Stony Brook - Stony Brook, New York Florida State University - Tallahassee, Florida University of Massachusetts at Amherst - Amherst, Massachusetts Michigan State University, various departments - East Lansing, Michigan Outside United States Simon Fraser University - Vancouver, Canada University of Halle - Halle, Germany University of Oldenberg - Oldenberg, Germany

47. Installations in Progress University of North Carolina at Greensboro - Greensboro, North Carolina University of Michigan, ITCS - Ann Arbor, MI Eotvos Lorand University - Budapest, Hungary Dakar University - Dakar, Senegal George Washington University - Washington, D.C. University of South Carolina -Columbia, South Carolina

48. On the web... www.lon-capa.org korte@lon-capa.org WANTED Collaborators

49. Network Architecture: Distribution LON-CAPA is build as a geographically distributed network of constantly connected servers

50. Network Architecture: Domains The network is logically divided into domains such as “MSU”, “FSU” or “Publisher X” Domains limit the flow of user information Domains can limit access to content resources Domains limit the extent of user privileges

51. The Tool: Coding Environment LON-CAPA runs on Linux (target for installs: RedHat distribution) Apache webserver mod_perl for server-side handlers JavaScript for client-side functionality MySQL as database for metadata searches MuPAD for symbolic manipulations (planned)

52. LON-CAPA vs. BlackBoard

53. Network Architecture: Server Types Two classes of servers: access servers and library servers Access servers: host user sessions – processing Library servers: every user and course has a home server in the network which holds all of their resources and data – storage backend

54. Access servers can have a list of spare access servers to offload sessions depending on own work load Additional round robbin IP scheme possible MSU setup: Network Architecture: Load Balancing msu.lon-capa.org s1 Library Server s3s2 s4 Round Robbin Spares

55. Network Architecture: Dynamic Replication When a content resource is first requested, the access server finds the home library server of the author/owner, subscribes to the resource, and copies it in the background. When resource is modified on its library server, subscribed access servers are notified. Depending on last date of local access, the access servers copy the new version over, or delete local cached copy.

56. Resources: Handlers Every resource in the system, both content and system programs, is called by URL – all resources can be bookmarked. Any resource can be processed on-the-fly while being delivered to the user. User calls for a resource, LON-CAPA finds appropriate chain of handlers, processes resource for desired target, and sends it out. Chain of handlers can also cover several transactions. Example: user calls for URL, but needs to login first and then pick course. All handlers can interact with session environment

57. Resources: Document- driven processing Example: user calls for sample.xml sample.xml Uses my.style XML Handler triggers my.style Defines tags for targets loads User Session Environment Target Prefe- rences

58. Resources: XML/MathML/HTML XML handler provides rendering of XML and HTML to targets XML, HTML, MathML, and LaTeX Math rendering capability: LaTeX can be inserted into HTML and XML documents between and tags. Identity The identity $\sin^2(\omega t)+ \cos^2(\omega t)=1$ allows us to… Math fragments are rendered into symbol fonts for target HTML, and MathML for MathML-capable browsers using a derivative of tth/ttm.

59. Resources: LaTeX LaTeX files can be directly put onto the file system, and will be automatically handled as if they had one large … around them

60. Resources: Assessments “Personalized” by course, user, and instance. Assessments are XML documents. Converters for both CAPA and LectureOnline problems exist (functionality superset). Web-based form editor for these XML documents will be provided similar to LectureOnline. Code segments within numerical type problems are written in Perl. Other problem types need no coding.

61. Resources: Publication Authors have a construction space (private) to edit their resources Authors can publish their resources into the public resource space Triggers several actions: generation of a new version (keeping all old versions), collection of metadata for cataloging, notification of all subscribed access servers. Construction Space Publication Step Resource Space Metadata

62. Resources: Metadata Metadata is collected for every resource at publication time Publication handler identifies possible keywords, gets default data

63. Resources: Granularity LON-CAPA RAT allows to dynamically combine content resources at several levels of granularity by reference Lowest level: fragments. One GIF, one paragraph of text, one applet, one problem Next level: Pages Next level: Sequences of pages Top-level: Maps for Courses (sequences)

64. Resources: Adaptivity Graphical resource assembly tool (RAT) to construct maps Allows for branching based on conditions using performance data, course data, and preferences Individualized curriculum

65. Data Structure: Users Courses are assigned to users, not vice versa Users keep their login, data, preferences, etc, throughout complete tenure Users assume a set of (temporary) roles over their tenure

66. Data Structure: Privileges Each role has a set of privileges Privileges on system, domain, and course level Allows for course- coordinator defined custom roles

67. Data Structure: Courses Internally, courses are handled like users without login privileges Each course has a unique ID, no semester transition Course points to its top-level resource map Deadlines, spreadsheets, etc, can be associated with the course Sections are part of course Homework and quiz performance data is associated with user by course

68. Data Structure: Authentication Course lists in CSV, tab-separated, and space separated format can be uploaded (planned: XML) It is checked if the user already exists; if not, new user is established on one of the library servers of the domain Role of student in that course is added Users are authenticated by their home server. Authentication can currently be internal, UNIX, or Kerberos. Authentication modular, other modes can be added All authentication information is 56-bit (DES) encrypted between client and server, and 128-bit (IDEA) between servers. No certificates necessary.

69. Data Structure: Resource parameters Deadlines, open dates, publish-answer dates, maximum number of attempts, weight of problem parts, etc, can be independently set for courses, sections and individual students Can cover all resources, one map, or only one resource Resources publish their parameters in metadata, no predefined set of parameters

70. Data Structure: Locking, Offline To use LON-CAPA for exams, locking mechanisms will be established: access to resources, all communication can be locked (planned, infrastructure in place) Role of Exam Proctor in system To facilitate paper based exams in offline mode, individualized exam sheets can be printed and “checked out” – bar code on top of page to “check in” grades for that exam (planned, infrastructure in place)

71. Data Structure: Assessment data Only raw data is stored All attempts are stored Data is stored in non-randomized form to allow for item analysis LON-CAPA provides spreadsheet functionality with sheets at assessment, user, and course level (exporting into each other) to calculate grades and statistics

72. On the web... www.lon-capa.org korte@lon-capa.org