Home > Research > Research Overview > 2023 > Office of Academic Affairs

Reducing Academic Testing Time with Artificial Intelligence

Associate Professor, Institute of Arts and Sciences

YASUDA Junichiro

Tests to measure academic achievement take a lot of time. For example, the Common Test for University Admissions takes two days. If we include practice tests and mid-terms and final exams at school, how much of one’s youth is spent on tests?

Computer-adaptive testing, a type of “artificial intelligence,” can alleviate this problem. This approach is expected to reduce the test time by about half by optimizing the questions according to the examinees’ answers (Fig. 1).

In a 2021 study 1), Yasuda et al. showed through numerical simulations that the number of questions can be reduced by half if a 10% loss in accuracy is allowed in a computer-adaptive version of a certain globally recognized test of physics concepts. This result corroborates the efficiency of computer-adaptive testing, and the paper by Yasuda et al. has been selected as a paper of note by a prestigious journal 2).

The University’s “Basic Skills Test” 3) is also conducted in a computer-adaptive manner. The results of research by Associate Professor Yasuda and his colleagues are expected to facilitate further advances in learning assessment based on the Basic Skills Test.

Fig.1 An example of computer-adaptive testing

Analysis of test accuracy and the number of questions

Related Links