Optimizing Test Length with Deep Reinforcement Learning

• In Event: Electronic Board Session I

Abstract

Employing deep reinforcement learning, we investigated whether a reduced item subset adhering to structural constraints could match the precision of the complete set-in estimating student ability. Although combinations yielded similar precision, none fulfilled all additional constraints.

Authors

• James Zoucha, University of Northern Colorado

• Igor Himelfarb, NBCE

• Nai-En Tang, National Board of Chiropractic Examiners

• Bruce L. Shotts