Using Code Detective Learning Activities to Promote AI Literacy and Code Verification Skills in Python Language

Abstract

At present, the use of Generative AI to support computer programming has become widespread. However, a significant challenge is AI Hallucination, in which AI generates source code that appears credible but contains non-existent methods or logical errors. If learners rely on this technology without critical judgment, it can lead to severe logical fallacies and a regression in self-directed problem-solving skills. This research aims to develop Code Detective learning activities that apply code-tracing processes to combat AI hallucinations and promote AI literacy and code-verification competencies, and to study the ability to detect errors through Trap Tasks. The sample group consisted of 35 students selected through purposive sampling. The research instruments included instructional slides on AI Hallucination, a worksheet featuring five Trap Task missions, pre- and post-tests, and competency evaluation rubrics. Data were analyzed using Mean (x ̅), Standard Deviation (S.D.), and Paired Samples t-test. Through a line-by-line analysis process, students were able to identify the location and type of errors in the Trap Tasks more accurately. The mean scores before learning were 22.31, after learning were 29.77, t = -7.114, p < .001, r = 0.773, and satisfaction with the learning management model was at a high level.