Repository logo
 

Automatic generation of number series reasoning items of high difficulty

Published version
Peer-reviewed

Type

Article

Change log

Authors

Liu, Y 
Luo, F 

Abstract

Number series reasoning items have been frequently used in educational assessment. This study reviewed the previous literature investigating features relating to item difficulty and developed a new automatic generator for number series reasoning items. Ninety-two items were generated and administered to 466 students. The results showed that the test achieved acceptable reliability. Items requiring two arithmetic operations were of particularly high difficulty. All stimulus features implemented in the automatic item generator proved to be significant predictors of item difficulty, explaining 77% of the total variance. This suggests that the automatic number series reasoning item generator was capable of generating items of considerably predictable difficulty. More importantly, the resulting items covered a wide range of difficulty levels, meeting the particular need for items of high difficulty.

Description

Keywords

LLTM, automatic item generation, high difficulty, number series, stimulus features

Journal Title

Frontiers in Psychology

Conference Name

Journal ISSN

1664-1078
1664-1078

Volume Title

10

Publisher

Frontiers Media
Sponsorship
Economic and Social Research Council (ES/L016591/1)
This work was supported by National Science Foundation for Young Scientists of China (Grant number: 61807005).