The mathematical structure of a bias-minimized assessment algorithm

We propose an assessor reallocation algorithm that aims to objectively reduce the marking biases of the individual markers based on their earlier marking statistics. The underlying mathematical structure along with a number of pertinent statistical properties and relationships has been analyzed for the model to assure the validity of the proposed methodology. Experiments on simulated data and on the real cases have also been conducted to illustrate the effectiveness on the reduction of the accumulated marking biases over multiple assessment items involving multiple assessors.