Optimal testing of structured knowledge

  • Authors:
  • Michael Munie;Yoav Shoham

  • Affiliations:
  • Stanford University, CA;Stanford University, CA

  • Venue:
  • AAAI'08 Proceedings of the 23rd national conference on Artificial intelligence - Volume 2
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Adopting a decision-theoretic perspective, we investigate the problem of optimal testing of structured knowledge - the canonical example being a qualifying examination of a graduate student. The setting is characterized by several factors: examinee's knowledge structured around several inter-dependent topics, a limited "budget" of questions available to the examiner, a decision to be made (pass/fail), and an utility for good and bad decisions. The existence of multiple professors brings up additional issues such as committee formation, and the existence of multiple students brings up issues such as fairness.