GTAAP: An Online System For Managing & Assigning Graduate Teaching Assistants to Academic Tasks

1. GTAAP: An Online System For Managing & Assigning Graduate Teaching Assistants to Academic Tasks Ryan Lim Constraint Systems Laboratory Department of Computer Science & Engineering University of Nebraska-Lincoln Supported by the Department of Computer Science & Engineering and NSF CAREER award #0133568 Lim: MS Project defense

2. Outline • Introduction • Task, motivations, challenges • Constraint model • Previous work • Our contributions • System implementation • Conclusions and future work Lim: MS Project defense

3. Computational task • Hiring & managing GTAs as instructors + graders • Given • A set of courses • A set of Graduate Teaching Assistants (GTAs) • A set of constraints that specify allowable assignments • Find a consistent and satisfactory assignment • Consistent: assignment breaks no (hard) constraints • Satisfactory: assignment maximizes • number of courses covered • happiness of the GTAs • Usually, number of hired GTAs is insufficient Lim: MS Project defense

4. Motivations • Context • “Most difficult duty of a department chair” [Reichenbach, 2001] • Assignments were done manually, countless reviews, persistent inconsistencies, etc. • Unhappy GTAs, students, instructors • Observation • Computers are good at maintaining consistency • Humans are good at balancing tradeoffs • Our solution • An online, constraint-based system • With interactive & automated search mechanisms Lim: MS Project defense

5. Challenges • We need to • Collect information, every semester • Build & repeatedly revise assignments • Communicate & store decisions • We need a system that • Maintains consistent information & decisions • Is accessible anywhere on the internet • Is robust & user-friendly Lim: MS Project defense

6. Constraint model: CSP [Glaubius ’01] • Variables • Grading, conducting lectures, labs & recitations • Values • Hired GTAs (+ preference for each value in domain) • Constraints • Unary: ITA certification, enrollment, time conflict, etc. • Binary (mutex): overlapping courses, same-TA • Non-binary: capacity, confinement • Objective • longest partial and consistent solution (primary criterion) • that maximizes GTAs’ preferences (secondary criterion) Lim: MS Project defense

7. Previous work • Portfolio of automated search algorithms • ERA search (Environment-Reactive-Agents) [Zou ‘03] • Stochastic local search [Zou ‘03] • Backtrack search (+ various heuristics) [Glaubius ‘01, Guddeti ‘04] • Interactive Search [Thota ‘04] • Web based • Runs on CSE server • Links LISP to PHP4 using network sockets • Vulnerable to CSE downtime Lim: MS Project defense

8. Our contributions • Designed a MySQL databaseto store student applications and class information • Built web-access interfaces • for GTAs and • the department manager • Co-developed a facility for collecting evaluations by instructors for their assigned GTAs • Rebuilt an interactive solver in Java Lim: MS Project defense

9. Outline • Introduction • System implementation • System architecture • Database design • Web-access • GTAs • Manager • Instructor evaluation of GTAs • Java Interactive Solver • Conclusions and future work Lim: MS Project defense

10. System architecture Contributions: • Database • Web-interfaces • GTAs • Manager • Instructor evaluations of GTAs • Interactive solver Lim: MS Project defense

11. Database design • Data originally collected on paper and stored in plain-text files • Pros: simple medium • Cons: Data-entry nightmare, numerous files, inconsistent data, no database ACID properties, difficult to query, no scalability, … • In Spring 2003 we migrated to MySQL Lim: MS Project defense

12. Database design • Imported class schedule data from external database (SIS+, CSE) • 30 database tables for: • Authentication and Logging 2 • GTA 13 • Manager 9 • Auxiliary tables 6 • Instructor names, • Semester identifiers, etc. Lim: MS Project defense

13. Web-access • Password-protected interfaces for • GTAs • the department manager • Data entry directly by the users eliminates inconsistencies from manual data entry • Next: screen shots of the interfaces for • GTAs • the department manager Lim: MS Project defense

14. Web-access: GTA • Each applicant must: • Input & revise academic record • Specify teaching preferences • Further, an applicant may • Report errors: Email webmaster • Provide feedback about the interface: Fill online survey Lim: MS Project defense

15. GTA interface: main page Lim: MS Project defense

16. GTA interface: academic record Lim: MS Project defense

17. GTA interface: teaching preferences Lim: MS Project defense

18. GTA interface: survey results 42 respondents from 2004 to 2006 (present) Lim: MS Project defense

19. Web-access: Manager • Access & manage GTAs information • Set up & manipulate classes information • Perform GTA assignments • Automated solvers • Interactive solvers • Report hiring & assignment decisions for GTAs evaluations Lim: MS Project defense

20. The set of modes For each mode, a set of functionalities Manager interface Lim: MS Project defense

21. Manager interface: GTAs Lim: MS Project defense

22. Manager interface: selective queries Lim: MS Project defense

23. Manager interface: classes Lim: MS Project defense

24. GTAs evaluations by instructors Lim: MS Project defense

25. Java Interactive Solver • Adapted from Thota’s M.S. project [2004] • More robust to CSE downtime • Cross-platform Java application on client-side • Secured authenticated access • Multi-user and simultaneous assignments • Save/delete/restore of scenarios • Printable output on a single sheet Lim: MS Project defense

26. Java Interactive Solver Lim: MS Project defense

27. Java Interactive Solver Lim: MS Project defense

28. Java Interactive Solver Lim: MS Project defense

29. Outline • Introduction • System implementation • Conclusions and future work Lim: MS Project defense

30. Conclusions & future work • We have built the main components of a prototype system for the management and hiring of GTAs at CSE • Designed a MySQL database to record students’ applications and store class information • Built web-access interfaces for GTAs and the department manager • Co-developed a facility for collecting evaluations by instructors for their assigned GTAs • Rebuilt an interactive solver in Java Lim: MS Project defense

31. Conclusions:motivations revisited • “Most difficult duty of a department chair” • Keeps the manager in the decision loop while removing the need for tedious and error-prone manual assignments • Helps producing quick (3 weeks down to 2 days) and satisfactory (stable) assignments • Initially, assignments were manually done on paper • Now, on-line data acquisition process • Enabled department to streamline & standardize GTA selection, hiring, and assignment • Overworked staff, unhappy GTAs • ‘Overjoyed’ staff (relieved from handling application forms and massive paperwork) • Enthusiastic anonymous online reviews from applicants Lim: MS Project defense

32. Current approach & future work • Current approach • Constantly sought & addressed users’ feedback • Systematically & immediately deployed incremental improvements • Maintained the system in use through major transitions & last-minute requests • Future improvements • Reorganize and redesign the web-interface • Reorganize and normalize the database tables • Possibly migrate to a higher performance and a feature-rich database Lim: MS Project defense

33. Acknowledgments • Committee members: Choueiry (Advisor), Daniel, Reichenbach and Srisa-an • Individuals: Clay Stevens, Charles Daniel, Traci Fink, Praveen Guddeti, Chris Reeson, Thota • Institutional support: CSE & NSF CAREER award #0133568 Lim: MS Project defense

34. Thank you. I welcome your questions Lim: MS Project defense