Interpreting aberrant testing events in remote assessments: Innocuous or malicious?

Abstract

Remote assessments are increasingly used in education and professional certification because of their accessibility and convenience. However, ensuring test security in these environments presents significant challenges, particularly in distinguishing between innocuous activity and more malicious behaviors. To address this, we introduce the Continuum of Deception (COD) framework, which ranks aberrant testing events along three dimensions: malicious intent, risk to score integrity, and verifiability. This framework guides empirical investigations across three complementary studies using data from a high-stakes remote assessment of English language proficiency. Study 1 examines expert rankings of aberrant events and finds strong agreement on malicious intent and risk but less agreement on verifiability. Study 2 analyzes score changes among repeat test takers and finds that some flagged events, particularly speaking-related ones, are associated with large score gains. Study 3 evaluates proctoring agreement and finds that events detected with machine or AI assistance, such as plagiarism, are identified more consistently than events based on human observation alone, such as looking away excessively. We use these findings to develop a four-tier system of consequences for test takers, based on the severity of an event and how reliably it can be verified. By integrating theoretical and empirical insights, this research provides a structured approach to balancing test integrity with fairness in remote assessments.