A Student Self-Grading System
Chao Liang (email@example.com)
Advisor: Dr. Chris Pollett
Committee Members: Dr. David Taylor (firstname.lastname@example.org) and Dr. Mark Stamp (email@example.com).
The main purpose of this project is to develop a grading system for homeworks in which both the students and professor share in the grading process. The hope though is to reduce the total number of homeworks the professor needs to grade in order to accurately assign grades to all of the students. To this end in CS297 an algorithm based on Quicksort, which we call VoteSort, was developed. The idea is that each student as well as the professor is given k randomly chosen homeworks from a total pool of n homeworks to rank from worst to best. These partial orders are then combined into a total order by using the VoteSort algorithm. This algorithm operates like Quicksort, but when a comparison of a homework j with the pivot i is needed, the number of times i appeared in one of the partial orders as better than j is used to decide which is bigger. Providing everyone grades accurately, a hand-wavy mathematical analysis suggests that with k around sqrt (n) (so everyone grading 5 homeworks for a class of 25 students), one can accurately determine a total order. Even if this is not the case it should be possible through simulation to determine a grade-spread which is accurate. To make the grades somewhat correspond to what the "correct grade" should have been for the homework two methods will be considered: (1) weighting the professor's permutation of k homeworks more highly in any vote tally, (2) rewarding students for getting comparisons in alignment with majority decisions, especially when those decisions involved many votes and the spread was large. We will try to analyze the weighting of rewards in terms of game theory. The whole system will also be constructed so as to be an easy add-on for Dr. Pollett PHP based grading program.
Innovations and Challenges
[Blythe92] An Analysis of Search Techniques for a Totally-ordered Nonlinear planner. Jim Blythe and Manuela Veloso. Carnegie Mellon University. 1992.
[Jonsson99] Towards Efficient Universal Planning - A Randomized Approach. Peter Jonsson, Patrik Haslum and Christer Backstrom.Department of Computer and Information Science. Linkoping University of Sweden. 1999.
[Kambhampati96] Unifying Classical Planning Approaches. Subbarao Kambhampati and Biplav Srivastava. Department of Computer Science and Engineering of Arizona State University. 1996.
[Kearns02] A Tutorial on Computational Game Theory. Michael Kearns Computer and Information Science of University of Pennsylvania. 2002.
[Mahajan04] Experiences Applying Game Theory to System Design. Ratul Mahajan, Maya Rodrig, David Wetherall, and John Zahorjan. University of Washington. 2004.
[Minton94] Total-Order and Partial-Order Planning:A Comparative Analysis. Steven Minton, John Bresina, and Mark Drummond. Recom Technologies. 1994.
[Myerson91] Game Theory: Analysis of Conflict. Roger B. Myerson. Harvard University Press. 1991.
[Nielsen00] Designing Web Usability. Jakob Nielsen. New Riders Publishing. 2000.