Joint Task Difficulties Estimation and Testees Ranking for Intelligence Evaluation

摘要: In this paper, we study the testing tasks evaluation and testees ranking problem, in which have different difficulty levels, capabilities.We assume that a testee may probability to pass certain task so as allow uncertainty. The goal of problem is simultaneously determine relative level each capability every testee, purely based on test outcome. We design two models solve problem. first one assumes outcome follows Bernoulli distribution; while second distribution with beta distribution-type priori knowledge. Then, form original into likelihood estimation problems them by using coordinate descent algorithms. show knowledge needed, when only carry out limited number tests due time financial budgets. All these findings are useful intelligence tests. Finally, discuss how extend statistical learning model for more general cases well specific case field Computational Social Systems like artificial social cognition evaluation.