A0301
Title: Understanding differential item functioning using process data
Authors: Ling Chen - Columbia University (United States) [presenting]
Jingchen Liu - Columbia University (United States)
Abstract: Differential item functioning (DIF) is an important concept in testing fairness. It occurs when items function differently among different subgroups. Previous research on DIF has mainly focused on statistical detection, yet understanding why DIF occurs remains a challenge. Process data obtained from respondents interacting with a computer-based assessment item provides a unique opportunity to understand DIF as it contains rich information about the progress and strategies towards problem-solving for each respondent. Using features extracted from process data, a variable that alleviates the DIF effect is constructed, which helps in detecting behavioural patterns that could lead to DIF and thus provides a deeper understanding of the underlying mechanism of DIF.
